UC Berkeley

Course Notes

Title
Problem Sets: Fundamentals Of Transportation And Traffic Operations

Permalink
https://escholarship.org/uc/item/1858c1w6

Author
Daganzo, Carlos F.

Publication Date
1998

eScholarship.org Powered by the California Digital Library

University of California
 PROBLEM SETS:
Fundamentals of Transportation
and Traffic Operations
By Carlos F. Daganzo

Institute of Transportation Studies

University of California at Berkeley

UCB-ITS-CN-98-1

February, 1998
 Fundamentals of Transportation and Traffic Operations
Problem Sets

Preface

These problem sets comprise a supplement to Fundamentals of

Transportation and Traffic Operations (C. Daganzo, Pergamon, 1997).
Academicians can also obtain a companion set of solutions by writing to
"Institute of Transportation Studies, Publications Office, 109 McLaughlin
Hall, University of California, Berkeley, CA 94720" or by sending e-mail to
[email protected].

This collection of problems is a result of the author's teaching experience over

numerous years. Some of the problems have been created by G. F. Newell,
and this is acknowledged on a case by case basis. The majority of the
problems have been used in the classroom. Others have been added recently
to supplement the newest portions of Fundamentals. Organized by chapter,
the exercises range in difficulty from elementary, as would be appropriate for
undergraduates, to advanced as may be appropriate for a Ph.D. qualifying
examination. The latter are marked by an asterisk. Most exercises are
intermediate in difficulty, however.

Errata for Fundamentals of Transportation and Traffic Operations are

included in paper copies of this document, and are also available on the
publications section of Carlos Daganzo’s world wide web home page at
http://www.ce.berkeley.edu/~daganzo.

The compilation would not have been possible without the help of the three
authors of the companion solution set. Robert Bertini (U.C. Berkeley)
managed the overall preparation of the problem sets and solutions. He
improved the wording, graphics and notation of many of the problems, and
facilitated the internet availability. David Lovell (University of Maryland)
provided improved problem statements for several significant problems and
Wei Lin (Virginia Polytechnic Institute and State University) also provided
helpful comments on several problem statements. Thanks also go to Ms.
Alison Andreas for her excellent word processing.

Carlos Daganzo
Berkeley, California
February, 1998
 1-1

Fundamentals of Transportation and Traffic Operations

Chapter 1 Problems

Problem 1.1

Consider that the maximum allowable value of the jerk for a train with standing
~
passengers is j (m/s3), its maximum acceleration is a~ (m/s2), and its maximum
velocity is v~ (m/s). First, find an expression describing the minimum distance
dmin between stops such that v~ is just developed. Then, determine the time, t,
needed for the train to travel any given distance, d (d>dmin), from a complete
stop to a complete stop. Assume that the maximum acceleration can be
considered to be independent of speed.

Problem 1.2

If the typical acceleration/deceleration of a car moving up in position in a queue

at a stop sign is a = 0.1g, determine the time that it takes for a car to advance
from the second position into the first. Assume a reasonable reaction time and
car spacing for the vehicle in question and discuss how this calculation may
relate to the maximum number of vehicles that may advance past a stop sign in
1 minute. How do you think variations in a across vehicles would affect this
number?

Problem 1.3

A train descending a 2.0 percent grade at 40 mi/hr applies the brakes at t = 0.

If the friction factor that is developed is 0.05, then find:

(a) The speed of the train after it has traveled 1,000 ft.
(Note: distances along the track and along its horizontal projection are
virtually the same because of the small angle.)

(b) The elapsed time after the train has traveled 1000 ft.

Problem 1.4

A railroad car with an initial speed vo is pushed over the hump of a railroad
yard that is h feet above its final track. The car should be retarded so as to

March 16, 1998

 1-2

reach its final destination x ft. away from the hump also at v = vo. Because the
speeds reached are low, the acceleration of the car is assumed to be the sum of
a constant, -fg, (independent of speed) that captures friction losses and a term
that is proportional to the grade -gs. If the car weighs W Newtons, determine
the magnitude of the retarding force that needs to be applied while the wheels
of the car travel over a retarder of length L if the retarder is located at the level
of the final track, so as to achieve the desired effect.

Problem 1.5

This problem examines the motion of an automobile traveling on a crest vertical

curve with the following profile:

y = 100 + 0.10x - 0.0005x2,

where y and x represent respectively the roadway elevation (in feet) and the
horizontal distance (also in feet). Assume that for the small grades that occur
in roadways, horizontal distances are nearly equal to those actually traveled.
The automobile sets the brakes when x = 0, while traveling at 45 mi/hr. The
friction factor is 0.3, and air resistance is negligible. Then:

(a) Find the stopping distance ignoring the centrifugal acceleration.

(b) Repeat the calculations considering the levitating effect of the

acceleration. (Hint: Solve this part numerically with a spreadsheet).

Problem 1.6

The state of Nevada allows 3-trailer truck combinations, while neighboring

states allow a maximum of 2 trailers per truck. Sketch a (t,x) diagram that
depicts the trajectories of several truck cabs and trailers, assuming that every
trailer-truck combination is stopped at the borders to ensure that every
combination always carries the maximum possible number of trailers.

If westbound 2-trailer trucks arrive regularly at the Nevada/Utah border every

15 minutes and eastbound 2-trailer trucks arrive every 20 minutes at the
Nevada/California border:

(a) Use the (t,x) diagram to sketch the trajectories of cabs and trailers for one
day, assuming that it takes 8 hours to cross Nevada and 30 minutes to
decouple or re-couple a trailer-truck combination. You should use an

March 16, 1998

 1-3

inset to depict in detail what happens at each border.

(b) Superimpose on your picture the trajectories of the drivers for a strategy
that would use them efficiently.

Problem 1.7

Devise a two-way bus service scheme on the (regional) route {A6 B6 C:C6 B6 A}
that will have regular departures from A, but buses continuing on from B to C
will depart less frequently. Some buses will be turned back at B and will return
to A without continuing on to C. Passengers from A to C should not transfer at
B but they may be delayed at B.

If the target service frequencies on A : B are 12 buses/hr in each direction, and

on B : C, 8 buses/hr, and if the trip times from A6 B, B 6 A, C 6 B, and B 6 A
are each 30 minutes, then depict a set of bus trajectories that will meet these
requirements with regular headways on all legs. You need to determine how
much to delay each bus at B, and if a bus just came from A, whether to turn it
around at B.

What size of bus fleet is required to accommodate your scheme?

Problem 1.8 (Courtesy of G. Newell)

If all trucks travel at speed v', cars at speed v, and the fraction of vehicles
passing a stationary observer that are trucks is p, what fraction of vehicles
would be trucks on a photograph? If the length of a truck is twice that of a car,
how many car lengths is the average vehicle on a photograph?

Evaluate these for v' = 50 km/hr, v = 80 km/hr, and p = 0.7.

Problem 1.9 (Courtesy of G. Newell)

If the speed limit on a highway network is changed so that the average speed
drops from v to v' (100 km/hr to 80 km/hr, for example), what happens to q and
k if people continue to make the same trips every day?

March 16, 1998

 1-4

Problem 1.10 (Courtesy of G. Newell)

From two consecutive film frames of traffic flow along a road, one observes that
there are ten cars per kilometer having zero velocity (they are parked), 20 cars
per kilometer traveling at 10km/hr and 40 cars traveling at 20km/hr.

(a) Determine the space mean speed and the time mean speed of traffic.

(b) If individual cars can maintain these speeds even while passing other
cars, how many cars will a driver traveling at 20km/hr pass while
traveling one kilometer?

Problem 1.11 (Courtesy of G. Newell)

From a photograph one observes that on a level section of highway 10% of the
vehicles are trucks, 90% are cars, and that there are 50 vehicles per mile of
highway. The trucks travel at 40 mi/hr, and the cars at 50 mi/hr. This highway
also has a section with a steep grade on which the speed of the trucks drops to
20 mi/hr, and the speed of the cars to 40 mi/hr. No vehicles enter the observed
sections of highways (except at the ends), and the flows are (nearly) stationary.

Determine:

(a) The flow of vehicles on the level section.

(b) The density of vehicles on the grade.

(c) The percent of trucks on the grade as seen on a photograph.

(d) The percent of trucks as seen by a stationary observer on the grade.

Problem 1.12

A one way road is shared by buses and cars. Buses travel at speed vK and carry

nK people. Cars travel at speed v and carry n people. The fraction of vehicles
that are buses passing a stationary observer is p. Answer the following
questions:

(a) What fraction of the vehicles on a photograph would be buses?

March 16, 1998

 1-5

(b) What is the average vehicle occupancy seen by a stationary observer?

(c) What is the average vehicle occupancy seen on a photograph? Evaluate

both occupancy averages for vK = 50 km/hr, v = 80 km/hr, n = 1, nK =20 and

p= 0.2.

(d) Consider the numerical data of part (c). If a bus emits 2 2 times the
pollutants emitted by a car per unit time, what fraction of the total
pollution is generated by the cars along one kilometer of road in one
hour?

Problem 1.13*

The distribution of vehicle speeds measured by a stationary observer is uniform

in the interval [vmin, vmax]. The speed limit is v~ 0 (vmin, vmax). Determine the
fraction of speeding vehicles seen by a police car traveling at speed v0. Plot for
-4#v0#4. Discuss.

Problem 1.14

The wheel base of a bicycle is D and the friction factor is f. The center of gravity
of the bicycle-rider assembly is h distance units above the ground, and is
centered between the wheels.

Find the maximum deceleration that can be achieved by using the rear brake
alone. Do not forget to include the inertial pseudo-forces that cause the normal
reaction between the ground and the wheels to be biased toward the front under
deceleration.

March 16, 1998

 2-1

Fundamentals of Transportation and Traffic Operations

Chapter 2 Problems

Problem 2.1

Sketch a cumulative flow diagram that represents the growth and dissipation of a rush
hour period at a toll bridge with time-independent capacity.

(a) Identify on the diagram: the arrival curve A(t), the maximum queue, the total
delay W, the duration of the saturated period, T, and the service rate F.

(b) Determine the change in delay if the service rate is increased by a small amount
∆F during the last half of the saturated period.

(c) Determine the change in the delay if the service rate is increased by an amount
∆F/2 for the whole duration of the period (involving the same number of
employee hours) and/or by an amount ∆F for the first half of the period.
Discuss.

Problem 2.2

A ramp meter is being considered at an entrance to a freeway. Currently, rush hour

traffic arrives at the on-ramp at a rate q1 from time t = 0 to time t = t*. After t = t*,
vehicles arrive at a (lesser) rate q2.

(a) Assuming that drivers will not change their trips, draw and label an input-
output diagram showing a metering (i.e., departure) rate of F (q2 < F < q1).
Label the maximum delay experienced by any vehicle (wmax).

(b) If an alternate route is available to drivers, and it is known that they will take
this route if their expected delay at the ramp meter is greater than wmax/2, add
this new scenario to your diagram. Now, show graphically the following:

(1) The number of vehicles which will divert.

(2) How much earlier the queue will dissipate (compared to part (a)).

March 16, 1998

 2-2

Problem 2.3

A vehicle (A) traveling on a freeway joins a 2-mile queue that contains 100 vehicles
at a time t = 0 min. Vehicles in this queue pass through the bottleneck at a rate of 50
veh/min. When there is no queue, vehicles travel (in free flow) at a rate of 1 mile/min,
provided that the flow satisfies: q < qmax = 100 veh/min. Do the following:

(a) Determine the delay and the time in queue for our hypothetical vehicle.

(b) Determine the (average) density of vehicles in the queue.

(c) Plot a triangular flow-density relation for our freeway that will be consistent
with the given data.

(d) If the (free) flow upstream of the bottleneck is 80 veh/min, determine the
location of the end of the queue (in miles upstream of the bottleneck) 1 minute
after the arrival of vehicle A. Solve this with the help of a picture, drawn to
scale, as follows:

(1) Construct the virtual arrival curve, the bottleneck departure curve and
the back-of-queue curve, starting with vehicle A.

(2) Identify on the picture the vehicle (B) that joins the queue at t = 1 min.

(3) Determine the distance in queue for such vehicle.

Problem 2.4

Suppose that the two approaches to a freeway merge, labeled 1 and 2, can send a
combined maximum flow F = 4,000 veh/hr through the merge without queues forming.
Suppose as well that if both approaches are queued, vehicles enter the merge in a 3:1
ratio; i.e., three vehicles from approach 1 for each vehicle from approach 2.
Furthermore, if only approach i is queued then the vehicles from this approach are
assumed to enter in a higher ratio (with F = 4,000 veh/hr). Then describe what would
happen if there are no queues and two surges of traffic, with 3,500 veh/hr on approach
1 and 700 veh/hr on approach 2, hit the merge simultaneously. Use a diagram and
explain.

March 16, 1998

 2-3

Problem 2.5

A minor street forms a tee-intersection with a major through-street. The minor street
traffic must stop at a stop sign. Traffic arriving at the stop sign must either turn right
(at rate F1) or turn left (at rate F2 n F1). If the right-turning and left-turning virtual
arrival curves, V1(t) and V2(t), are given, determine the total delay.

Problem 2.6

This is an example of a time-dependent O-D table estimation problem in a congested

network. If curves D1 and D2 are observed downstream of a congested diverge or
intersection, and an upstream curve A(t) is given that is consistent with D1 and D2,
then determine curves A1(t) and A2(t) assuming a FIFO discipline.

Problem 2.7*

Suppose that the service times of a FIFO freeway bottleneck when vehicle n is
discharging depends on its waiting time, wn, in the following way:
~
s(n) = s1 + (s0 - s1) e- wn /w (s1>s0>0)
~
where s1 and s0 are the maximum and minimum service times and w is a characteristic
waiting time. (This type of model has been proposed to explain the drop in capacity
observed in some Japanese motorways subject to severe congestion.).

(a) Assuming such a system reaches an equilibrium, what is the equilibrium queue
length?

Suppose now that at the onset of the rush hour prior to which there is no queueing and
no delay, the (virtual) arrival curve at the bottleneck jumps suddenly by n0 vehicles
and then increases steadily with a slope

1 1
λ0 0( , )
s1 s0

March 16, 1998

 2-4

(b) Sketch a departure curve and the growth of delay during the rush hour, as you
would expect it to happen. Would you expect the system to reach an
equilibrium? Why? Would the value of λ0 matter?

(c) Derive an expression for the critical value of λ0 (λcrit) beyond which the system
cannot reach an equilibrium.
(d) Discuss the dependence of λcrit on n0 and the effect that changes in n0 have on
our system (with a fixed λ0) for the duration of the rush hour. Suppose for
example that n0 varies from day to day.

Problem 2.8

Suppose that the customer cumulative (virtual) arrival curve at a bottleneck is of the
form:

V(t) = middle (0, (Ni /T) t, Ni) for Ni, T > 0.

where Ni is the total number of customers served during day i. If the bottleneck serves
customers at a uniform rate (F) every day, and Ni varies across days so that 50% of the
days (Ni /T) = 2F and 50% of the days (Ni /T) = 3F, determine:

(a) The total delay accumulated in the system after N days (N large), the average
delay per day, and the average delay per customer over all the customers.

(b) Vavg, Davg, the area between these curves, and the average delay per customers
one would estimate if all days were like the average day. Compare with (a).

(c) Derive the average across days of the average delay per customer ( w )avg, and
show that it is different from (a) and (b). Explain why.

Problem 2.9

The (t,x) trajectories of N vehicles (j=0,...,N) are given by the following equation:

xj(t) = min {4t-4j, 2t-j+10}, j=0,...,N

(a) Draw a (t,x) diagram showing several of these trajectories and discuss
qualitatively what would be happening (label some of the trajectories).

March 16, 1998

 2-5

(b) Draw cumulative plots of vehicle number (j) vs. t at the locations x=0 and x=10
of two observers.

(c) Calculate the total trip time experienced by vehicles 0 to 10 between x=0 and
x=10:

(1) From the diagram (a).

(2) From the diagram (b).

(d) Repeat (c) for the maximum number of vehicles between the observers

(e) Draw on sketch (b) the loci of points where vehicles are observed to change
speeds.

Problem 2.10

Suppose that the two observers of Problem 2.9 record the following N-curves A(t) and
D(t):

A(t) = min {t, 2t-2}

D(t) = min {t-1, 2t-8}

(a) Plot these curves and then a set of (t,x) vehicle trajectories that would be
consistent with these data (use two families of parallel lines).

(b) Describe qualitatively what you think is happening, and based on your
conjecture:

(c) Plot N(t,x) as a function of x for t=5 from x=-30 to x=30.

March 16, 1998

 3-1

Fundamentals of Transportation and Traffic Operations

Chapter 3 Problems

Problem 3.1 (Original version courtesy of G. Newell)

The cumulative number of cars, A, arriving at a toll plaza exhibits a daily peak such
that A is approximately given by the following:

A = α1t, if 0 < t < 7:00 hrs,

A = 7 α1 + α2 (t-7), if 7:00 < t < 9:00 hrs, and

A = 7 α1 + 2α2 + α1 (t-9) if 9:00 < t < 24:00 hrs,

where α1 and α2 are arrival flows at different times of the day (in cars/hr), and t
denotes the time of day in hours. Then, if the cost of providing round-the-clock service,
C ($/day), is proportional to the toll plaza capacity, F (cars/hr), C=βF, and if each
customer hour of delay is valued at δ ($/hr), find an expression for the capacity level
that will minimize the total cost per day.

The above scenario would be more realistic if the capacity could be temporarily reduced
during the off-peak period (with only a few booths being operated) to a value F1 : α1
< F1 < F. We assume that this is only done when there is no queue. Repeat the

exercise if the savings from operating at the lower level for T hours are β] (F - F1 ) T

$/day (24β] < β); i.e. if the daily cost is:

C = βF - β] (F - F1) T.

(Hint: find first the best F1 and T for any given F.)

Problem 3.2

A corridor of length L is to be served by bus transportation during the evening

commute (all trips start at x=0 and end somewhere in the interval (0, L)). The demand
is uniform in this interval, with density λ (pax/day/mile). Because the corridor is too
long, it has to be served by two separate routes, i=1 operating between x=0 and x=d1

March 16, 1998

 3-2

with headways H1, and i=2 carrying passengers to x0 (d1, L) with headways H2.

The cost of operation is assumed to be directly proportional to the bus-miles traveled.

Then:

(a) Write an expression for the total number of bus-miles traveled per unit time (e.g.
day) as a function of our parameters.

(b) Explain why the total number of passenger-hours of delay incurred during
boarding and the total number of passenger-hours of delay incurred during
offloading per unit time (e.g. per day) should depend on the decision variables
as follows:

Boarding delay per day = α ∑ H i d i

i

Offloading delay per day = β ∑ H i d i2

i

Here we assume that d2 = L - d1.

Boarding delay is incurred out-of-vehicle. Assume that passenger boarding time

is negligible compared to headways, therefore delay can be counted as the
difference between a passenger=s arrival time at the station and the departure
time of the bus carrying that passenger. Offloading delay for any passenger is
incurred in-vehicle, and represents the time that passenger waits on the bus
while other passengers disembark. Assume that every alighting passenger
delays the bus by γ time units (i.e. assume only one passenger disembarks per
stop, as in a suburban setting where passengers request stops by ringing the
bell).

Interpret the coefficients α and β.

(c) Write a mathematical program (MP) to help the transit agency decide what to
do if the agency has a fixed number of buses, which essentially limits the
number of vehicle-miles traveled to a maximum amount, M. Assume the bus
company would like to minimize its own costs, plus the delay costs incurred to
the user. (This is a politically loaded endeavor, as determining the trade-off
between the two would depend drastically on perspective. Assume that the cost
per vehicle-mile of travel is reflected in a parameter δ, with units of (vehicle-
hours per day) per (vehicle-mile)).

March 16, 1998

 3-3

(d) Classify the MP if the Hi are held constant (route design problem) and also if the
di are held constant (scheduling problem).

(e) Classify the MP if all of the variables are decision variables.

Problem 3.3

Problem 3.2 has 5 parameters (L, M, λ, γ, and δ) in its formulation and 4 decision
variables. Is it possible to perform a change of variable that will eliminate all the
parameters? If so, write the dimensionless version of the problem. If not, write a
version that has the fewest possible free parameters.

March 16, 1998

 4-1

Fundamentals of Transportation and Traffic Operations

Chapter 4 Problems

Problem 4.1

The flows of cars and trucks on a very long upgrade are 10 and 2 vehicles per minute
respectively. Trucks travel at 0.8 mi/min, and cars at 1.0 mi/min when not trapped
behind a truck (otherwise they travel at 0.8 mi/min). The system is stationary and
passing is possible. If the time used to climb a 1 mile section of the grade averaged
across all the vehicles observed over a long time is 1.1 min, determine:

(a) The density of vehicles on the upgrade.

(b) The average speed on the grade, taken across cars only.

(c) The proportion of time that each car spends behind trucks.

Problem 4.2

From two consecutive film frames of stationary traffic along some road, one observes
that there are 20 cars per kilometer traveling at 100 km/hr and 30 cars per kilometer
traveling at 120 km/hr.

(a) If individual cars can maintain these speeds even while passing other cars, how
many cars will one of the drivers traveling at 120 km/hr pass while traveling
one kilometer?

(b) How many passing maneuvers are executed by all drivers in 1 km of road during
one hour? (This type of information is directly relevant for accident analysis and
prevention.)

Problem 4.3

Show the following:

(a) That if the q vs. k diagram is concave, the v vs. s diagram is also concave.

(b) That if the q vs. k diagram is piecewise linear then v vs. s diagram is piecewise

March 16, 1998

 4-2

linear.

(c) Devise the q(k) relation that would correspond to a shifted exponential v(s)
curve.

Problem 4.4

A 3-lane freeway drops a lane at location x=0 km. If the maximum flow per lane is
1,800 vph, the jam density per lane is 108 veh/km, and the free flow speed is 100
km/hr, then:

(a) Plot to scale on a single set of axes the q vs. k relations for the 3-lane and 2-lane
sections of the road, as well as the curve corresponding to a single lane. Assume
that the Q(k) curve(s) are smooth and explain in words the geometrical feature
that relates the three curves.

(b) Plot the v(s) curves that correspond to the above Q(k), and again explain their
relationship. (Note here that the spacing of cars in a given lane is not the same
as 1/k, and that the s in the v(s) relation should be the latter.)

Problem 4.5

Repeat Problem 4.4(a) if the Q(k) relation is triangular. Determine the trip time that
will hold from x = -1 to x = +2 km as a function of flow when conditions are time-
independent, as in Fundamentals Figure 4.4d. Recognize that the section under
consideration is inhomogeneous and assume that the bottleneck transition is much
smaller than the length of the section (3 km). Sketch the answer graphically and
express it mathematically.

Problem 4.6

Explain qualitatively what would happen to the solution of Problem 4.5 if the freeway
was 3-lanes wide throughout but a point bottleneck (e.g. an incident) that effectively
blocked 1 lane existed at x=0 km.

March 16, 1998

 4-3

Problem 4.7

A homogenous freeway section of length L is observed to exhibit a concave Q(k) relation

under all stationary traffic conditions; i.e. there is a reproducible slope q between the
roughly straight N(t) curves at the two detectors for every possible separation, -. When
the observed slope qo(-) and/or the separation between the observed curves N0(t) is not
consistent with that expected, an incident may be suspected. Write an expression for
the position of the (point) incident within L in terms of qo and the observed N 0(t)
curves.

Problem 4.8

The purpose of this problem is two-fold: familiarizing you with the organization of the
Highway Capacity Manual (HCM) and encouraging you to view its recommendations
as just that—recommendations.

A two-lane approach to a suburban intersection is located on level ground. There are

no left turns or special turning phases. Twenty percent of the vehicles turn right,
yielding the right of way to the moderate pedestrian traffic on the cross street. With
10 ½ ft lanes, no buses and only 10% heavy vehicles, parking is allowed because parked
vehicles do not encroach on the moving lanes and parking movements are infrequent.

(a) Determine the saturation flow rate according to the 1994 HCM, and the
minimum proportion of time that the cycle should be effectively green if the
approach flow is 1,170 vph.

(b) Critique the procedures for determining the adjustment factors in the HCM
formula. (Hint: do you think all the factors in the formula should be
independent?)

Problem 4.9

Platoon dispersion (zeroth order approximation): A signal turns green at t=0 and traffic
discharges from it at a rate of 3,000 veh/hr for 30 secs until it turns red. If all vehicles
reach their cruising speed shortly after the signal threshold (x=0) (free passing) and the
distribution of desired vehicular speeds is as follows: 20% at 50 mph and 80% at 60
mph, then draw the cumulative curve of vehicular labels, N(t), at a location x=1 mile
downstream of the signal. Solve graphically, and then express the solutions
analytically. As a more advanced exercise, express N(t) as a function of x for an

March 16, 1998

 4-4

arbitrary distribution of desired velocities and the number of passing maneuvers per
unit time and unit distance (a measure of accident exposure).

Problem 4.10*

Platoon dispersion (first order approximation): If the fast vehicles of Problem 4.9 are
delayed by 5 seconds every time they overtake a slow vehicle (they would follow the
slow vehicle for a longer time, of course), but we neglect the infrequent interactions of
more than two vehicles due to queuing, then determine the approximate time to and
location xo at which all the passing maneuvers will have been completed. Then plot
N(t) at locations x = xo/2, x = xo and x = 3xo/2. (In solving this problem you have
developed and used a first order theory of traffic dynamics.) Explain any shortcuts
that you use in order to obtain the solution; there are some rather useful ones.

Problem 4.11

Explain qualitatively what part(s) of the answer to Problem 4.10 are independent of
the speed of the slow vehicles.

Problem 4.12

Suppose that the q(t,x) and k(t,x) on a one-directional road are given by the following
(in an unspecified system of units):

(a) q(t,x) = qoe(t/to - x/xo) and k(t,x) = koe(t/to - x/xo)

(b) q(t,x) = f (x/t)/t and k(t,x) = f (x/t)/x

(c) q(t,x) = f (x/t)x and k(t,x) = f (x/t)t

for (t,x)  (t0,x0) > (0,0)

Identify which of these formulas is physically possible, without traffic generation in the
vicinity of (to,xo) and which is not (if any) and explain why. For the physically possible
case(s), determine N(t,x) (up to an additive constant) and explain qualitatively what
is happening.

March 16, 1998

 4-5

Problem 4.13

Relative flow measured by a moving observer in non-stationary traffic: If the conditions

of Problem 4.12(b) hold for all (t,x) > (0,0), and we assume that f(x/t) = 1, then
determine the total number of vehicles seen by an observer that departs location x=1
at t=1 with a speed of v<1, as a function of time and the total seen for t.

Problem 4.14

Discuss qualitatively how one might go about deriving the trajectory of an interface
separating non-stationary states A (qA (t,x), kA (t,x) ) and B (qB (t,x), kB (t,x) ) if we know
that the interface passes through a particular (t,x) point (e.g. the origin). As an
example, find the interface that passes through point 2,4 and separates (qA, kA) = (1/t,
1/x) from (qB, kB) = (1,1).

Problem 4.15

Prove that if Q(k) is strictly concave, the relative flow seen by an observer moving with
a shock is maximized when the shock is a wave and that the relative flow seen by the
wave increases with k. (Show pictorially and prove analytically).

Problem 4.16*

Sketch the trajectory of the vehicle passing through (t,x) = (0, -1) for an initial value
problem of a homogenous highway in LWR theory with the following initial data:

k(0,x) = 1 + x (for -1<x<4) and Q(k) relation:

March 16, 1998

 4-6

Derive its equation and find the times and locations where it crosses the shock and
then comes to a stop.

Problem 4.17*

For the data of problem 4.16, derive the equation of the shock path, and then use the
minimum principle to write an expression for N(t,x).

Problem 4.18*

Suppose that the flow past an active bottleneck caused by an incident during the
recovery after a total closure is given by a monotonically increasing function qb(t). If
the queue upstream of the bottleneck obeys the LWR model with a concave Q(k)
relation, explain how you would construct graphically the curve N(t,x) for a (queued)
location that is 1 mile upstream of the bottleneck. Express your procedure precisely.

Problem 4.19*

Solve problem 4.18 for the following data:

qb (veh/min) v (mi/min)
1-exp{(600-1-s)/(600-1)}
1
100
100{1-e-t/10}

150-1

t (min) 600-1 s (mi/veh)

Problem 4.20

Consider an imaginary one-directional highway with the following attributes:

qmax=2,000 vph, ko=50 vpm and kj=200 vpm. We assume that a very long queue is

March 16, 1998

 4-7

discharging past an obstruction at x=0 (miles) with flow q=1,000 vph and that the
obstruction is removed at time t=0. If a q(k) relationship (of the two types specified
below) holds at all points in time-space, plot the density and flow profiles at x=+½ and
x=-½ miles, from t=0 until the flow at both locations is close to qmax. Assume:

(a) Two wave-speed relation.

(b) Linear q(k) relation for k<ko and quadratic for k>ko. The negative wave
velocities range from 0 at ko to -50 at kj. (These are not necessarily reasonable
numbers but will suffice for the purposes of testing your skill).

Make sure that all 4 density profiles are plotted on the same scale. Do the same for the
flow profiles.

Problem 4.21*

At the one-mile marker upstream of a point bottleneck with capacity 50 veh/min, traffic
flow increases with time as:

q(t) = 100{1 - e(-t/10)} if t  0

0 if t < 0

where t is in minutes.

If the highway is homogenous upstream of the bottleneck and LWR theory holds with
a triangular q - k relation with vf = 1 mi/min, w = 1/6 mi/min and qmax = 100 veh/min,
then determine (graphically or analytically):

(a) The cumulative number of vehicles to have joined the queue by time t.

(b) The distance spanned by the queue as a function of t, and the time when the
queue spills over the 1-mile marker.

(c) The cumulative number of vehicles to have crossed a location, xo, ½ mile
upstream of the bottleneck by time t.

(d) Check the consistency of answers (b) and (c).

March 16, 1998

 4-8

Problem 4.22

Traffic leaving a major event enters a 3-lane freeway at a major on-ramp located 2
miles downstream of the nearest off-ramp. The event traffic reduces the available
capacity to the through vehicles as follows:

q̂ max = 100 veh/min for t < 0 min

= 60 veh/min for 0  t < 10 min
= 80 veh/min 10  t < 20 min
= 100 veh/min for t > 20 min

If the approaching freeway traffic is time independent with q = 100 veh/min and 15%
of the traffic exits at the off-ramp (leaving 85 veh/min of through traffic), determine
whether or not the freeway queue will back up to the off-ramp. Assume that the q-k
relation of Problem 4.20 holds for each lane. Explain what would happen after the
spillover. Repeat the exercise assuming that traffic is metered on the on-ramp so that
the q̂ max will never dip below 76 veh/min (i.e. so that the q̂ max values would be 100,
76, 76 and 100 if we assume that the same number of vehicles enter). Solve this
problem with (t,x) diagram and with Newell’s method. Check the consistency of both
results.

Problem 4.23

A freeway exhibits a triangular flow-density relation with parameters: vf, qmax and kj,
where vf is the free-flow speed, qmax is the capacity and kj is the jam density.

(a) Plot this curve and derive an expression for the function that gives the (space-
mean) speed as a function of density inside a queue. Don’t forget to specify the
range of k to which the equation applies.

(b) If kj = 600 veh/mile, vf = 1 mile/min and qmax = 100 veh/min, determine the delay
experienced by a vehicle that joins a 2 mile queue caused by a bottleneck that
flows at q = 50 veh/min.

Problem 4.24

The cumulative arrival curve at a location 5 miles upstream of a bottleneck A(t) is

given and so is the departure curve at the bottleneck D(t):

March 16, 1998

 4-9

A(t) = 50t for t < 0 (min)

A(t) = 100t from t = 0 to t = 30
A(t) = 1500 + 50t for t > 30
D(t) = -400 + 80t for 5 < t < 45

(a) Draw these curves.

We assume that there is negligible entering/exiting traffic in the 5 mile stretch and
that the flow-density relation for the stretch is triangular with qmax = 100 veh/min, vf
= 1 mile/min and kj = 600 veh/mile. If vehicles travel at the free flow speed until they
join the queue and from then on they travel at the speed that prevails inside the queue:

(b) Draw the q-k curve and determine the traffic speed (space-mean speed), vq,
inside the queue.

(c) Determine the number of minutes traveled by vehicle number 2000 at speed vq
and the length of the queue (in miles) at the time that vehicle joins it.

(d) Determine the total number of vehicle-hours that are traveled at speed vq and
also at speed vf by vehicles 0 to 3000.

Problem 4.25

Vehicles traveling on a one-lane road find that it widens to 2 lanes and that shortly
after that (1/15th of a mile downstream (352 ft)) there is a traffic signal. The 2-lane
section is used to store the queue caused by the traffic signal and in this way increase
its capacity. The signal operates with a 60 second cycle and 30 second effective green
for this approach. The flow-density diagram for the one-lane section is linear in
between the following break-points:

k (veh/mile) q (vph)
0.00 0
18.75 900
75.00 1800
150.00 0

(a) Assuming that the diagram for the two-lane section exhibits the same speed for

March 16, 1998

 4-10

twice the density, draw both diagrams on the same graph. Use only the top half
of a sheet of paper.

(b) What would be the maximum flow on the one-lane approach if there was no
intersection? What would be the capacity of the intersection if its approach was
2-lanes all the way?

(c) What is the capacity of the intersection now? To solve this part—the main
objective of this exercise—you must draw a shockwave diagram on the bottom
half of the sheet of paper you used for part (a). For simplicity, you may assume
that vehicles accelerate instantaneously to the optimum speed after discharging
from a queue. You must be careful, however, because the traffic states on the
two-lane and one-lane sections of the road come from different q-k curves.

(d) Can you derive a simple formula for the maximum flow as a function of the
length of the two-lane section? This should be a two minute exercise based on
the part (c) diagram.

Problem 4.26

We consider a homogeneous freeway whose flow-density diagram is linear between the

following break-points:

k (veh/mile) q (vph)
0 0
100 5000
600 0

(a) At exactly 4:00 pm. and when the freeway is flowing steadily at q=4,000 vph, an
incident that blocks ½ of the lanes occurs at mile x, (0 < x <1). Determine as a
function of x, and according to LWR theory, the exact time at which the drop in
flow will be sensed at two detectors located at miles 0 and 1.

(b) Derive an expression for the time t and location x of an incident that is known
to have occurred in the road interval (0,1), given the times, U and D, at which
a flow drop from Q is q is sensed at the upstream and downstream detectors.
(This type of analysis can be used to diagnose detected incidents; detection is

March 16, 1998

 4-11

more efficiently achieved from occupancy data, however.)

Problem 4.27

Traffic flows at 2,000 veh/hr in the direction of increasing x on a one-directional road.

At location x=0 the road narrows from 2 to 1 lanes. We assume that a flow-density
curve, triangular in shape, defines the possible traffic states at all points in time-space.
The following parameters apply to the 1 lane section: Free flow speed vf = 80 km/hr,
jam density kj = 75 veh/km, and optimum density ko = 25 veh/km.

(a) Plot the flow vs. density curves for the 1 and 2 lane road sections on the same
diagram.

(b) Sketch the speed vs. density curve for the 1 lane section. Show clearly where
the curve is linear and where it is not.

(c) If a truck located at x = 1 km suddenly slows to v = 40 km/hr (at t = 0), describe

in words (one or two sentences) what you think will happen behind it.

(d) Draw a time-space diagram with the relevant interfaces between traffic states
and a couple of vehicle trajectories. Find when the effect of the speed reduction
is felt at x = 0.

(e) When is it felt at x = -1 km?

(f) If the truck resumes a speed of 80 km/hr at x = 2 km, complete the time space
diagram. Include a few more vehicle trajectories.

(g) What is the delay caused by the disturbance to the 100th vehicle behind the
truck? (The question has a very easy solution if you see the trick.)

Problem 4.28*

Consider a one-directional arterial with closely but evenly spaced signals and no
turning traffic. The signals have been set with identical cycles and phases. The green
phases have been coordinated so that a vehicle traveling with the free-flow speed would
hit the beginning of the green (or the end) at all the signals. This offset is 5 seconds.

Assume that an incident causes a queue to build up past several of these intersections

March 16, 1998

 4-12

and that queued vehicles behave as in the LWR theory with a constant wave speed
that is 5 times smaller than the free-flow speed. The signals have a 30 second green
phase and a 60 second cycle. Then,

(a) Sketch the trajectories of the discharging vehicles for the first 4 signals
upstream of the incident and determine when the vehicle at the head of this
signal’s queue would depart. Show all the interfaces.

(b) Sketch the trajectories if the offset is 0 seconds, and the delay to the same
vehicles of part (a).

Problem 4.29*

Moving bottleneck with an arbitrary trajectory (Proof of the general statement about
N-spans at the end of Fundamentals §4.4.5).

(a) Sketch some unqueued but irregularly spaced vehicle trajectories to represent
free-flowing traffic with vehicle q on a homogenous road. Assume a triangular
q - k curve. Then superimpose on this picture a curved segment (with endpoints
A and B) representing the trajectory of a slower moving obstruction that
interferes with traffic for a finite time. If this (point) obstruction reduces the
width of the road by a fixed amount (as in the case of Fundamentals Figure
4.24), show on a companion sketch how the trajectories of the vehicles in the
first sketch would be modified as a result of the obstructions. Include all the
interfaces.

(b) Show from an analysis of the pictures from part (a) that the vehicular
trajectories downstream from the x - coordinate of point B do not depend on the
path of the obstruction, conditional on the locations of A and B.

(c) Show that if either point A or B (or both) are moved along the vehicle trajectory
on which they lie, the downstream trajectories do not change.

Problem 4.30

Two consecutive intersections that are 100 ft. apart on a one-way street are controlled
by identically set, pre-timed traffic signals. Their cycle is one minute and the effective
green phase for through traffic is 30 secs. Assuming no turning movements, determine
the maximum flow that can get through both intersections as a function of the offset

March 16, 1998

 4-13

if the spatial extent of queues on the street can be modeled with a triangular q(k)
relation with the following parameters: jam density = 3 veh every 80 ft, capacity
(saturation flow) = 0.5 veh per sec, and free-flow speed = 40 ft per sec. Use 0, 10 and
20 second offsets for illustration.

Problem 4.31*

Suppose that the q-k diagram on a given highway was as follows:

Construct a t-x diagram showing the vehicle trajectories and interfaces arising when
a capacity stream of vehicles (q=100) is suddenly stopped (e.g. at t=0) at a railroad
crossing. (Note that the q-k curve is not concave).

How long would be the transition seen by a stationary observer situated ½ km

upstream of the barrier between the free flow state and the stopped state?

March 16, 1998

 5-1

Fundamentals of Transportation and Traffic Operations

Chapter 5 Problems

Problem 5.1

During the rush hour, the desired flows past two neighboring traffic signals in the
critical part of a two-way crowded arterial are large enough to generate lasting queues
upstream of both signals. The signals are ¼ km apart. They are operated on the same
cycle and phases, C=1 min, G=30 secs, and with no offset. The saturation flow of both
signals is 50 veh/min.

A secondary street controlled by a yield sign intersects the arterial in between the two
signals, x km away from one of them. If the critical gap is Ho=7 secs and the move up
time is M=3 secs (both assumed to be the same for everyone), and assuming that right
turns have a dedicated lane, find the capacity of the left turn lane as a function of x,
and the location xo where the capacity would be maximum.

You should assume that if a queue backs up past the intersection, the queue will not
block cross-traffic but it will block any traffic that would want to join the queue. In
order to estimate the extent of queues you should assume that vf=1 km/min and kj=300
veh/km.

Problem 5.2*

The entrances and exits of a roundabout are consecutively numbered from 1 to N in the
direction of travel. Even numbers are exits and odd numbers are entrances. There is
an O-D matrix of (stationary) flows q = {qij}, where qij 0 only if i is odd and j is even.

The roundabout is narrow and each of its links can accept a maximum of 2,000 vph.
Likewise, the exits can accept a maximum of 1500 vph. All the entrances are able to
carry the respective qi’s.

(a) Introduce the necessary notation to express the capacity condition for q as a set
of linear (in)equality equations.

(b) Assume that qij = 0, except for q14 = q32 = 1,500 veh/hr and that there are traffic
signals (meters) at entry points 1 and 3. What metering rates )1 and )3 would
maximize the total system outflow?

March 16, 1998

 5-2

(c) If the O-D table of part (b) becomes effective at t = 0 (min) and lasts until time
t = 30 (min), and from then on we have q14 = q32 = 500 veh./hr, determine the
total vehicle delay that would be experienced if the metering rates are chosen
at all times so as to maximize the total system outflow without exceeding
capacity. Explain the rationale for such a strategy. (We assume that vehicles
traverse the roundabout in a time that is negligible compared with 30 min., our
time scale.)

Problem 5.3 (Courtesy of G. Newell)

A single airport runway is used for take-offs and landings. The minimum time
between a landing and a take-off, a take-off and a landing, or two consecutive take-offs
is one minute each, but the minimum time between two consecutive landings is 3/2
minutes. The strategy for sequencing operations is to alternate take-offs and landings
whenever there is a queue of both.

If qT is the rate of take-off requests and qL the rate of landing requests, determine as
a function of the ratio qT/qL, the maximum rates qT and qL at which one can serve take-
offs and landings without having a queue of either steadily increase: i.e., evaluate the
capacity.

Problem 5.4

Consider a traffic signal at an intersection of two one-way streets with symmetric

virtual arrival curves (V1(t) = V2(t)) and no turns. Using a graphical construction such
as that of Fundamentals Figure 5.3c as a basis for thinking, derive a formula for the
cycle that should be used during the oversaturated period to minimize delay. The cycle
should be such that a slight increase in C with the concomitant increase in )(C), would
induce an increase in the shaded area of the figure that would be exactly offset by the
decrease in the unshaded part. Discuss the result for reasonable values of the
parameters.

Problem 5.5

A vehicle actuated traffic signal serves two Poisson traffic streams with flows q1 = 500
vph and q2 = 300 vph. The saturation flow rate on both approaches is the same s =
1,200 vph, and the lost time each time the signal changes from one phase to the other
is 5 seconds. Then:

March 16, 1998

 5-3

(a) Using a spreadsheet, plot on the same graph the cumulative arrival time vs.
vehicle number curve for the two approaches. Make sure that the graph includes
enough vehicles so that the last vehicle (on both approaches) arrives after 300
seconds.

(b) Using a straight edge and a sharp pencil, construct on such a graph the
cumulative departure time curves for all vehicles to have arrived in the first 300
seconds.

(c) Compare the average cycle length you have simulated with the theoretical.
Then, describe how you would calculate (from the graphs) the mean and
variance of the delay per car for approach 1.

Problem 5.6 (Courtesy of G. Newell)

Along a one-way arterial highway of uniform width, there are two one-way cross
streets. Vehicles arrive at the intersections on the cross streets at uniform rates q2(1)
and q2(2), at intersections 1 and 2 respectively, and at intersectoin 1 on the arterial at
a uniform rate q1(1). There is no turning traffic and negligible platoon spreading
between the intersections on the arterial. The saturation flow is s at both intersections
in both directions, and the effective lost time per cycle is the same at both
intersections. Both intersections are controlled by fully actuated two-phase signals
which hold any signal phase until the queue it is serving has vanished. If, after the
queue vanishes, there is no call for the other phase, the signal phase will continue until
there is a call.

Traffic is sufficiently heavy that the signal will typically serve many vehicles in each
phase, so one can treat vehicles as a continuous fluid, but the intersections are
undersaturated.

Describe how the signals will behave if (a) q2(1)>q2(2) and (b) q2 (1)<q2 (2). Would the
existence of stochastic effects (Poisson arrivals, for example), slight turning traffic, or
slight platoon spreading disrupt the patterns significantly?

Problem 5.7

A traffic signal is set for a 1 minute cycle. The red and green phases on approach A are
as follows: R = 30 secs. (including amber phases) and G = 30 secs. If the flow on
approach A is constant (no random fluctuations) at 20 veh/min and the saturation flow

March 16, 1998

 5-4

is 1 veh/sec calculate the average delay for a typical car.

If work on the street is taking place downstream from the intersection so that only 25
veh/min can pass, calculate the average delay caused by the street work to a typical
vehicle leaving the intersection. Assume that the queue at the downstream restriction
never backs up all the way to the intersection.

If there is a signal downstream of the work zone, with the same (C,G) as the upstream
signal, determine:

(a) The delay per vehicle at the downstream signal if the offset between the two
signals is the free-flow trip time between them.

(b) The offset that will minimize the total delay.

Problem 5.8*

A railroad classification yard serves a periodic train schedule with a period of 1 week.
All the trains are of the run-through type. This means that the block of cars for our
yard carried by each train is dropped off in one of the receiving tracks immediately
after each arrival, and that shortly thereafter (the time lag can be ingnored) the engine
(with rest of the train tagging along) collects the blocks of cars that have been prepared
for the train in question from a departure track.

A train is characterized by the set of blocks of destinations it serves and by its

departure time, which is the same every week. We assume for simplicity that each
block is carried by only one train. The flow of cars through the yard (by block) is
assumed to be reproducible; i.e., to be the same every week.

We also assume that cars can be classified (i.e., moved from the receiving tracks to the
clasification tracks) at a given rate during the classification process, and that they can
also be moved from the clasification tracks to the departure tracks at the same rate
during the train make-up process.

You are supposed to explain as briefly and elegantly as possible (with pictures and
equations as necessary) the following:

(a) An efficient yard operating strategy, which will ensure that no trains are
delayed while using as few receiving and departure tracks as possible.

March 16, 1998

 5-5

(b) How you would identify the number of receiving and departure tracks needed
for your given strategy, assuming:

(1) That the tracks are distinct, and

(2) That they are not.

(c) How would you identify the number of cars that would miss a connection, and
the time when the yard has the maximum number of tracks in use?

You will be evaluated on the precision of your notation and explanation. Therefore you
should be as succinct as possible.

Problem 5.9*

A long freeway with many origins along its way branches off at its very end, when it
hits a congested ring road around a central business district (CBD). Branch 1 can
serve vehicles at rate 1 and branch 2 at an unlimited rate. We assume that if a queue
develops on the freeway it is never long enough to reach back to any of the origins.
(This is not realistic but makes the solution easier). We also assume that vehicles are
stored in the freeway and advance into the diverge in a FIFO order, never exceeding
the service rate of branch 1.

(a) Explain how you would determine with pictures and/or equations the total delay
in the system if we know the cumulative number of vehicles that leave each
origin for each destination as function of time, and the trip time from each origin
to the bottleneck.

(b) Formulate the equilibrium conditions for the morning commute problem, where
everyone must pass the bottleneck by time to, assuming that the number of
people traveling from each origin to each destination is known. The goal is to
find people’s chosen arrival times; i.e. the arrival curves for 1 and 2. You should
assume that people hate time in queue more than the same amount of time
wasted by passing the bottleneck too early. You may also assume that the ratio
of these two quantities is fixed and the same for everybody. (Hint: explore the
possibility of simplifying this problem by studying its equivalence to a simpler
one. Describe the analogy in detail.)

March 16, 1998

 5-6

Problem 5.10*

A closed-loop corridor of length L consists of a closed-loop freeway with a well defined

flow vs. density relation q(k), and the set of arterials within which it is embedded.
These can be viewed as a frontage road on which trips average a speed va. This speed
is both independent of flow and smaller than the free-flow speed of the freeway. We
assume for simplicity that the corridor's origin-destination flows are stationary in time
and space (along the corridor), that they average D distance units per trip and that the
ramp spacing is negligible compared with D. The combined number of trips starting
from all origins per unit time is Q. We imagine that we can control the number of trips
that enter the freeway per unit time without inducing queues on the ramps, e.g. by
pricing, or by a system of stickers whereby people are only allowed to use the freeway
on certain days. The remaining trips use the arterials, resulting in a combined arterial
flow qa across every cross-section of the corridor.

(a) Write a relation among the freeway flow q, the arterial flow qa and the regional
origin-destination flow Q. (Hint: it will involve L and D.)

Note that qa can be freely chosen and that q is then determined by the relation you
derived in (a). Alternatively, you could imagine choosing the density ka and
determining k (or q).

(b) In view of this, and using the relation q(k), rewrite the result of part (a) in terms
of Q and densities k and ka.

(c) Find the (q,k) combination that minimizes the total system delay and then
describe a simple graphical construction to display it on the q(k) diagram.

(d) Repeat (c) under the assumption that there is no control and users reach an
equilibrium where the speeds of both facilities are the same. Calculate the
increase in delay from the solution of part (c), and express the result on the q(k)
diagram.

March 16, 1998

 5-7

Problem 5.11

Imagine a network consisting of three components: a freeway corridor, a set of arterials

and an inexhaustible traffic demand for a “destination zone,” D. Drivers are assumed
to choose to enter the freeway only if its travel time (including any queueing on the on-
ramp) is less than the arterial trip time. The freeway is homogeneous with m lanes,
so that the average speed within a queue is a function V(Q) of the discharge rate, Q,
for any Qqmax. The freeway has some freeway flow speed, v1, and the arterials have
an average speed (including stops), v2. The freeway on ramps are spaced some distance
d apart. We assume that when the freeway and the on-ramp are queued, merging
vehicles and vehicles in the right-most freeway lane combine into one traffic stream in
a one-to-one rate.

Determine the equilibrium queues on each ramp as a function of r if people choose the
least time route. Would a ramp metering strategy change the rate at which vehicles
enter D? What would be the advantages and disadvantages of metering/closing the
ramp closest to D?

Freeway

d Destination
Zone

r r=3 r=2 r=1 r=0

Problem 5.12

Repeat Problem 5.11 if the freeway is the only way to reach D; i.e. if the arterials end
before reaching the destination. Would a ramp metering strategy change the rate at
which vehicles enter D?

March 16, 1998

 5-8

Problem 5.13

A capacity “paradox:” There are two ways of going from A to B. A fast route that can
be accessed after passing through a bottleneck of capacity ) (veh/min) and a slow route
that takes - (min) longer. The bottleneck queue can be seen from the point of decision,
so that drivers can estimate the delay due to the queue, w. We also assume that if the
queue contains more than Qmax vehicles it will have spilled over past the diverge, and
that at the diverge drivers choose the fastest route (e.g. the fast route if w<-). Then,
if the demand during the rush hour (, assumed to be constant during the rush)
exceeds both ) and Qmax/-, determine the rate at which vehicles will advance into the
diverge during the rush if (a) ) < Qmax/- <  and (b) if the bottleneck capacity is
expanded: Qmax/- < ) < .

March 16, 1998

 6-1

Fundamentals of Transportation and Traffic Operations

Chapter 6 Problems

Problem 6.1 (Source: Anonymous)

A seven-story building is served by two elevators. You are at the ground level, trying
to decide whether to use the stairs or wait for an elevator. The elevators operate
independently, and each is equally likely to be at any floor at any given moment.

(a) Sketch the sample space for the position of both elevators. Floors are numbered
from 1 (ground) to 7, and neither elevator is at the ground floor (they could be
at floors 2, 3, 4, 5, 6, or 7).

(b) Use the sketch to determine the probability that at least one of the elevators is
no more than two floors away from the ground level.

(c) Use a new sketch and repeat the above exercise assuming now that you are on
the third floor, and neither elevator is there (they could be at floors 1, 2, 4, 5, 6,
or 7).

Problem 6.2

If X,Y are the coordinates of a randomly selected house in a circular city of radius 1,
0  X 2+Y 2  1, where housing density is uniform:

(a) Are X and Y statistically independent? Identically distributed?

(b) What is the probability density of a house’s euclidean distance to the CBD?

(c) What is the expectation?

If the travel must take place over an infinitely dense square grid network oriented with
the coordinate axes,

(d) What is the expected distance traveled to the CBD?

(e) What is the probability density of the distance traveled to the CBD?

March 16, 1998

 6-2

Problem 6.3

The distance between points A and B of a certain highway is measured as the sum of
124 independent measurements. The random error E in each of the measurements
is uniformly distributed between [-2, 2] cm. If the total distance AB is approximately
2 km and the lengths of the segments are approximately equal, compute the following:

(a) The mean and variance of the total error in the distance AB.

(b) The probability that the total error is not more than 0.01% of the actual distance
AB.

Problem 6.4

Suppose 4,000 containers are to be loaded onto a vessel. Each container is weighed
before being loaded using a heavy-duty scale that indicates their weight accurately as
an integer number of pounds (so that the (round-off) error in each measurement is
uniformly distributed on [-0.5, 0.5]).

(a) What is the maximum possible error in the total estimated weight?

(b) Would it be reasonable to use a normal approximation to estimate the error in

the total weight measured? Explain.

(c) What is the probability that the total weight will be underestimated by 50 lbs
or more?

Problem 6.5

A container crane unloads 1,000 containers from a ship. All of these containers have
an equal probability (1/4) of being 20-footers. The rest are 40-footers. Twenty footers
have a 0.4 probability of leaving by train, while 40-footers have a 0.6 probability.

(a) What is the mean and standard deviation of the number of 20-foot containers
leaving by train?

(b) What is the mean, standard deviation and approximate distribution of the
number of TEU’s (twenty-foot equivalent units) unloaded? Hint: This question

March 16, 1998

 6-3

is very easy but a little tricky because the numbers of containers unloaded of
each type are not independent; their sum is 1000. (Note: one 2-foot container =
1 TEU and one 40-foot container = 2 TEU’s.)

Problem 6.6

A train with 88 railcars is being classified at the railyard. All of these cars have an
equal probability of being sorted onto any of 10 available classification tracks. What
is the mean and standard deviation of the number of cars sorted to track 1?

Each of those cars is carried by the next departing train with a probability p = 0.3.
What is the mean and standard deviation of the number of cars remaining in track 1?

Problem 6.7*

Consider the uphill direction of a very long (infinite for our purposes) two-lane,
bidirectional road that climbs to a mountain pass. Geometric design considerations
preclude passing, except at wider passing sections, of length l (miles), which are
regularly spaced (beginning to beginning) L miles apart. The flow of trucks in the
uphill direction is q and the flow of cars: Q = nq, where n is assumed to be an integer.

Trucks travel at speed v and cars travel at speed V (v < V) unless trapped in a moving
queue behind a truck. The first car in a queue is assumed to initiate a passing
maneuver by accelerating instantaneously to speed V when the car (not the truck)
reaches the passing section's threshold. Passing is only possible inside the passing
sections. Able to accelerate and decelerate instantaneously between speeds v and V,
cars are also assumed to keep a headway of 2 seconds at all times with the car they
follow. (This means that if a car suddenly changes speed, a trailing car will change
speed at the same point in space.) Then:

(a) If l = 0.25, L = 5, q = 2 veh/min, n=5, v = 45 mph and V = 60 mph, find the

average speed of a fast vehicle as it negotiates the uphill section. (Assume that
on approaching the uphill section the trucks are evenly spaced and that each is
followed by the same number of cars.)

(b) As a function of q, n, v and V, determine the smallest L and l that will ensure
that long term average speed of the cars is 60 mph (no delay). Do first for the
data in (a).

March 16, 1998

 6-4

(c) Explain why the number of vehicles trapped behind a truck on approach to the
uphill may vary across trucks like a Poisson random variable. Then, give
reasons why it might not.

(d) If the number in (c) is Poisson, explain how this would change the answer to
part (b).

(e) If, for the data of part (a) with n=0, the headway between trucks are described
by a Poisson process with rate q, determine the average travel speed of an
isolated car. (Hints: A car will always leave a passing section at speed V; and
the time until it meets the next truck should have the forgetfulness property.)

Problem 6.8*

The intersection of two one-lane, one-way streets is controlled by a stop sign. Traffic
flow on the major street is well described by a stationary Poisson process with mean
Q (veh/sec). Queued vehicles in the minor street start looking for gaps M seconds after
the preceding vehicle in the queue has merged. They merge (immediately) upon
finding a gap in traffic greater than T secs (T > M). Find the minor street capacity as
a function of Q, T and M. (Hint: Find first the expected number of merges in one main
street headway.)

Problem 6.9

The headways of successive cars observed at an active bottleneck are uniformly and
independently distributed between a and A secs (a < A). Use the inversion formulas
for processes with positive and independent increments to derive the index of
dispersion of the counting process when the latter is observed on scales including many
vehicles.

Problem 6.10

Use a spreadsheet to do the following:

(a) Generate 500 sets of three random digits (rectangular (0,1) variables), (X, Y, Z)
and their average S = (X+Y+Z)/3.

(b) Calculate the theoretical mean and variance of S. Compare with the sample

March 16, 1998

 6-5

mean and variance for the 500 observations of S.

(c) Use the SORT capabilities of the spreadsheet to plot the empirical c.d.f. of S,
and the theoretical one according to the central limit theorem (this may require
that you integrate numerically with your spreadsheet the normal p.d.f.).
Comment.

Problem 6.11

Export containers arrive at a port marshaling yard, where they accumulate while
waiting for the departing ships. Type a containers are 20 feet long and take 200 ft2 of
yard floor space; type b containers are 40 feet long and take 400 ft2 of yard floor space.
Both container types arrive at the yard as independent Poisson processes; each with
an average arrival rate of 60 containers per day—this arrival pattern holds around the
clock, 7 days a week. Containers of each type enter the port facilities through separate
gates, where they are inspected (one at a time) for a nearly constant 5 minute period.

Answer the following questions:

(a) What is the probability that the next container to arrive at one of the gates
immediately after a container that doesn’t have to queue can also avoid
queuing delay? Explain the logic behind your answer.

(b) What is the probability that more than 75 type b containers arrive in 1 day? Be
explicit about any approximations you use.

(c) The port is considering including a more formal inspection at the gate (including
random customs searches), taking 20 minutes on average, with a standard
deviation of also 20 minutes. Approximately, what would be the steady state
average queue length at the gate with the expanded inspections?

(d) Write an expression for the mean and the variance of the total floor space taken
by the containers of both types together arriving in a time interval of length -
days. Give the answer for - = 7 days.

(e) Assume that the arrival rate of type a containers is not known exactly (it is
known to be between 45 and 55 containers per day) but that in your zeal for
accuracy you want to pinpoint it within a standard error of 0.1 containers per
day. How long would you have to observe the system?

March 16, 1998

 6-6

Problem 6.12

Let {S1, S 2, ..., S n} be a number of independent vehicle speed measurements from an

unknown distribution with mean µ and variance )2. Determine the smallest sample
size n such that the probability of the absolute difference between the true mean µ and
the sample mean S being greater than one tenth of the true standard deviation is less
than 0.05:

P (|S - µ| > )/10) < 0.05.

(You may assume that n is large.)

Problem 6.13

Comparison of N-curves: The true (unknown) one-minute counts, n(t), upstream of a

bottleneck depend on the time in the middle of the counting interval, t, as follows:

n(t) = no(10 + sin t),

where t is in minutes. (A simple expression is used here for simplicity of formulation,

and so that the various error measures can be computed quickly; the qualitative results
arising in this problem do not depend on the ability to express the data in a neat closed
form).

Model A predicts the following interval flows (i.e. minute counts): nA(t) = 10no. Model
B predicts: nB(t) = no(9.8 + sin t).

(a) Plot the curves n(t), nA(t) and nB(t) on the same graph and evaluate the root
mean square error in the counts produced by model A and model B for a long (1
hour) period.

(b) Plot the cumulative curves obtained for 1 hour of data N(t), NA(t) and NB(t) and
again determine the root mean square error in the counts produced by both
models. On the basis of the results in (a) and (b), which model would be
preferable?

March 16, 1998

 6-7

Problem 6.14

In the case of Problem 6.13, the model predictions remain fixed, but the true counts
fluctuate from day to day. On day i they are n(t)i = no (10 + sin t) + it where the it ’s
are i.i.d. observations from a random variable with mean 0 and variance (2no)2. Then,
identify the time to at which the maximum deviation between Ni(t) and NA(t) would be
expected to occur, as well as its distribution. If the model is rejected if its maximum
deviation (which we approximate by the deviation at time t0) exceeds the 95th
percentile deviation that would occur at t = 60 with a true model, determine the
probability of rejection on any given day. Repeat for Model B.

Problem 6.15

Traffic on a short stretch of road (length, L=1 km) which is being studied by time
lapsed photography can be described as a superposition of N (stationary) Poisson
streams (i=1, ...N) with speeds vi and average flow qi. If the time lapse t is short or
comparable with a vehicle’s trip time:

(a) Show that the number of vehicles, nm, on each photograph, m, is an outcome of
a Poisson random variable with a given mean. Find the mean.

(b) Find an expression for c(lt), the covariance in the counts of two frames, l shots
apart. Plot a graph of c(lt) vs. (lt) if N=2, q1=10 veh/min, q2=20 veh/min V1=2
km/min and V2=1.5 km/min. Then evaluate the covariance integral, C, and the
variance of n̄ as a function of time if observation lasts many minutes and t0.

March 16, 1998

 6-8

Problem 6.16

The table below contains vehicle positions and speeds, as observed from two aerial
photographs of a 1 km freeway section.

DATA
Vehicle # Position (m) Speed(m/sec)
1 27.14 23.28
2 121.34 24.94
3 143.16 22.77
4 168.64 25.84
5 204.41 26.87
6 253.93 24.93
7 291.11 26.26
8 382.08 25.07
9 519.82 25.74
10 529.76 22.31
11 531.51 22.21
12 570.41 22.97
13 654.68 22.69
14 665.61 25.42
15 671.45 24.46
16 689.56 25.84
17 707.62 26.41
18 773.21 26.12
19 824.86 26.42
20 839.49 26.45
21 915.17 26.56

Using these data:

(a) Estimate the time mean speed.

(b) Construct the empirical cumulative distribution function of the speeds that
would be seen by a stationary observer. (Assume that the observed pattern is
repeated every km of freeway, and that the vehicle trajectories are straight).
Calculate the mean and variance.

(c) Compare the result of (a) and (b) and try to explain any discrepancies.

March 16, 1998

 6-9

Problem 6.17 (Courtesy of G. Newell)

Suppose that an observer knows that her vehicle detector turns on and off before the
vehicle has completely crossed the detector. From experimental calibration of the
instrument she has verified that the true length of a vehicle lj is always  larger than
the effective length for all vehicles (independent of j). She knows . She observes a
flow q and an effective occupancy r' from her detector. What else must she know in
order to convert her r' into r so that she can compare her observations with someone
else’s? (The vehicles have different lj and different speeds vj.)

Problem 6.18

Paired loop detectors can be used to estimate vehicle length and speed. If the length
of each detector is d and their center to center separation is D >> d, derive an
expression for the speed and length of vehicle i (vi, li) as a function of the times t1u and
t2u when the upstream detector u goes on and off, and the times t 1 d and t 2 d when the
downstream detector does the same. You may assume in your derivation:

(a) That li < D-d for all i.

(b) That the detectors are on when any part of the vehicle is over the detector.

(c) That vehicles maintain their speed over the detectors.

Explain how a bias in the synchronization of the 2 detectors would affect the results.

Problem 6.19*

Explain how you would determine the acceleration of vehicle i with the setup of
Problem 6.18, assuming that vehicles undergo a constant acceleration motion over the
detectors. Ignore detection errors.

Problem 6.20

Consider the numerical data of Problem 6.15 (b) and assume that a moving observer
traveling with the traffic stream at speed vo is used to estimate the (stationary) and
unknown flow (q = q1+q2 = 30 veh/min) on the 1 km road section of interest. She does
this by recording in a single one-way trip the relative speed of every vehicle she passes

March 16, 1998

 6-10

or leaves behind; i.e. the number of vehicles of type 1 and type 2, m1 and m2. Then,

(a) Write a formula for an estimator of q in terms of m1, m2, V1, V2 and vo. (The
observer is aware that there are only 2 vehicle families)

(b) Derive an expression for the variance of estimator as a function of q1, q2 and vo
and plot it as a function of the amount of time spent in the trip (1/vo).

Problem 6.21

This problem examines the effectiveness of license plate travel time studies. An
arterial street carries a flow of q veh/hr for its whole length. However, each individual
vehicle has a probability p = 0.2 of leaving the arterial at each intersection; i.e. at every
intersection 20% of the traffic turns off (and an equivalent number of vehicles turns
onto the arterial). The method used involves recording a vehicle’s license plate and
time of arrival at the two ends of the study section by two separate observers. The
study section contains N intersections. Each observer can record a maximum of ) cars
per hour. The (unknown) average travel time to travel one block is E(to). Its variance
(across vehicles) is E(to), where is a time constant. We assume that these moments
are (approximately) the same for all blocks. If the travel times on succeeding blocks
by the same vehicle are mutually independent, then the variance to mean ratio of the
time to travel the study section is N E[to]/NE[to] = . Under these assumptions, answer
the following questions:

(a) For light traffic (q < )), write an expression for the duration, T, of the study
period that will allow us to estimate the average travel time for the section with
a standard error equal to 5% of the mean. The expression should include as
variables: q, n, , and E[to]. Explain why T is smallest for N 4 or 5, and what
you would do if you had to study a very long arterial with, say, N = 30. What is
the optimal T if q = 400 vph, N = 4, and we believe that E[to] = 30secs. and =
20secs.

(b) Derive an expression for heavy traffic (q > )).

Problem 6.22

A few bus lines (i = 1....N) connect at a given terminal. The station is so small that
walking times between connecting vehicles and or the entrance/exit of the terminal are
comparable to the bus stopping time and much smaller than the headways. Observers

March 16, 1998

 6-11

record the cumulative arrival and departure curves of passengers vs. time at each
berth, but not at the entrance and exit of the terminal (i = 0). Explain how you would
determine from this information (and from the realized schedule of bus arrivals and
departures) the cumulative flows by O-D pair including the terminal itself (i = 0) as an
origin and destination. You may assume that the data are consistent with a world in
which passengers do not waste time moving around.

March 16, 1998

 7-1

Fundamentals of Transportation and Traffic Operations

Chapter 7 Problems

Problem 7.1

Draw a time-space diagram including the trajectories of 2 rapid transit trains that
undergo the same motions while separated at all times by a 1 minute headway. The
two trains pass through and stop at two transit stations that are 1 mile apart. They
stop for ½ minute at each station and travel at 60 mph in between. Label the
headway, show the position of the stations, identify on the figure the time (t*) when the
two vehicles are closest in space, and determine such minimum distance.

Problem 7.2

(a) If arrivals at a bus stop are stationary in time (constant flow) and the headways
alternate between 5 minutes and 10 minutes, determine the probability that a
randomly chosen passenger has to wait more than x minutes Gw(x), for all
possible x. Plot the result. Would the result change if passengers arrived
independently of the schedule as a stationary Poisson process?

(b) If the same number of passengers arrive to meet each bus and they are
uniformly distributed in a 7-minute window of time that extends all the way to
each bus’ departure, determine Gw(x) and the average waiting time assuming
that people will board the first bus that comes.

Problem 7.3 (A similar, simpler version of this problem can be found in Horonjeff &
McKelvey, “Planning and Design of Airports,” 1983)

Two different aircraft use a runway used for landings only. One aircraft has an
airspeed of 120 knots, while the other has an airspeed of 150 knots. The common
approach path is 6 nautical miles long. Along this approach, the aircraft must
maintain a minimum separation distance of 3 nautical miles. Both aircraft occupy the
runway for 60 seconds, and they cannot be on it simultaneously. There is a 20 knot
headwind.

(a) Determine the minimum safe aircraft headway at the runway threshold for each
pair of airplanes.

March 16, 1998

 7-2

(b) If the arriving aircraft are split 50-50 between those with airspeeds of 120 and
150 knots, determine the maximum theoretical (no buffers) runway capacity in
aircraft per hour.

Problem 7.4 (Courtesy of G. Newell)

Each of two buses carries passengers from a depot to various destinations and returns
for another trip with a round trip time very nearly equal to R. The buses are run by
independent drivers, however, who make no attempt to coordinate their schedules.
Actually, one bus runs slightly faster than the other so that over many trips the
fraction of trips that the second bus leaves within a time t after the first bus is t/R,
with 0 < t < R. In effect, the times between departures of the buses are random, with
a uniform distribution over the interval 0 < t < R.

If the passengers arrive at the depot at a constant rate, what is the average time that
a passenger must wait for the next bus? Compare this with the wait if the headways
are controlled so as to be R/2.

Problem 7.5

A bus takes commuters from a CBD to a suburb on a route that (not including stops)
takes on average ½ hour; we also know that the non-stop travel time is normally
distributed with a 2 minute standard deviation. The bus makes 20 stops, and 50
passengers start the trip. Each stop takes a time which is the sum of a constant (30
seconds), plus a negative exponential random variable with mean 10 seconds. In
addition, each alighting passenger extends the time of its stop by a time which is
rectangularly distributed between 2 and 12 seconds. How late after the start of the trip
can the bus company be 95% sure that the bus will be available for other purposes?

Problem 7.6 (Courtesy of G. Newell)

Suppose that two shuttle buses operate between downtown Berkeley and the campus.
One bus can carry 20 people and has a round trip time of eight minutes (including
loading and unloading passengers), the other can carry 25 people but has a round trip
time of 10 minutes. There is always a queue of people waiting at the downtown bus
stop, but people arrive at the campus bus stop at a steady rate of 2 per minute.

March 16, 1998

 7-3

If neither bus is idle at any time, determine:

(a) Flow of people from downtown to campus.

(b) Flow of people from campus to downtown.

(c) Flow of buses from downtown to campus.

(d) Flow of buses from campus to downtown.

(e) The average number of passengers per bus from campus to downtown.

If one insisted on dispatching these buses from downtown to campus at equal

headways, what is the maximum number of people that could be carried per hour from
downtown to campus? Would you expect this strategy to reduce the average waiting
time of passengers going from campus to downtown?

Problem 7.7

Consider a symmetric circular bus route, where every passenger travels exactly on 4
links (he alights on the 4th stop not counting the one he got on) and where the
cumulative flow of arriving passengers at each stop is A(t). If the headway is H and
the trip time between stops (not including the stop) is H/2, depict graphically and
express analytically the occupancy of the bus immediately after departing a stop at
time t*. Repeat the problem if, instead of 4 stops, there is a distribution where pi of the
passengers travel for i stops (i = 1, 2, 3, ....).

Problem 7.8

A 200 double-chair lift at a ski resort (a form of transportation) moves at 8 mph. Such
a chair lift can hold at most 200 skiers, since at any point in time only 100 two-person
chairs are going uphill. The distance traveled by each chair from the bottom to the top
is ½ mile. The downhill run is 1 mile long. On Sunday they expect 500 skiers. Then:

(a) If everybody skis downhill at 16 mph, find the length of the lift line that will
develop (i.e. the number of skiers waiting for a chair), and the average time it
takes to go through the line.

(Note: to solve this problem you should recognize that: (1) the flow of skiers

March 16, 1998

 7-4

going up must equal the flow of skiers going down, and (2) that if a line
develops, the chair must be full to capacity.)

(b) If half the skiers ski back downhill at 24 mph and the other half at 10 mph, then
determine the length of the lift line, and the number of 24 mph skiers it will
contain.

Problem 7.9

A B

D C

Transportation service is to be provided by M=4 buses linking stops A, B, C, and D,

as shown by the route in the figure. The 4 links on this route have the same length so
that, aside from the passenger loading and unloading time a bus will take -=300
seconds to traverse the link between any two consecutive stops. This includes the time
needed to stop the bus and open and close the doors. The passenger flow between
any two different stops (e.g. A and B, D and B, etc...) is q=60 passengers per hour per
origin-destination pair; each boarding and alighting passenger extends the link
traversal time by =5 seconds.

(a) Numerical Analysis: Answer the following questions assuming that the buses
can maintain constant headways operating continuously (without slack time),
and that passengers for any O-D pair arrive at a constant rate over time.

(1) What is the flow on each link?

(2) What is the headway at each stop?
(3) What is the minimum bus capacity that can do the job?
(4) What is the passenger delay before boarding the bus, averaged over all
passengers?
(5) What is the total average time in the system (delay plus riding time)?

March 16, 1998

 7-5

(b) Analytical Modeling:

(1) Derive an expression for the total average time in the system for an
arbitrary number of stops, n, and arbitrary: q,-, and M.

(2) The bus company should decide whether to assign M/2 buses for
clockwise service and M/2 buses for counter-clockwise service. In this
manner, a passenger traveling from A to D could take the counter-
clockwise bus and cut her travel time substantially; on the other hand
this might increase the headways and the average waiting time as a
result. Which of these two strategies yields the least total average time
in the system depends on the variables (q, n,-, and M). Can you state
a condition for choosing? Which of the variables are irrelevant?

Problem 7.10

An entrepreneur is planning to open a long term parking lot d kilometers away from
an airport. To be competitive, he plans to operate shuttle buses between the lot and
the airport at regular H minute headway, 24 hours a day. He plans to use buses able
to carry any number of passengers. The buses would travel at speed v (km/min) and
would spend T1 and T2 minutes stopped at the two ends of their route (the lot and the
airport).

Outbound customers (traveling from the lot to the airport) are foreseen to arrive at an
evenly distributed rate of q cars per minute throughout the day, and the inbound
(return) passengers to behave similarly. We also believe that the probability density
function of the parking times, P, would be:

fP( x ) = (1/p) e-x/p, if x  0, and fP( x ) = 0, otherwise.

Finally, we assume that the operating cost of one bus (driver, wages, lease cost, etc.)
is w ($/min), that the rent for one parking space is r ($/car-min), and that all other costs
are negligible. Then,

(a) Derive a formula in terms of some of the above variables for the rate at which
the entrepreneur’s cost per car served increases with d.

(b) Calculate the required size of the lot, assuming that its surface area should be
double of that which is strictly needed to hold the average number of cars in the
lot.

March 16, 1998

 7-6

(c) Write an expression for a parking fee schedule as a function of the actual
parking durations (pi) experienced by specific individuals (i) if the goal is to
achieve a 50% mark-up over the cost of providing the service.

Problem 7.11

Explain technically how you would go about calculating the total emissions of NOx in
lbs. per hour arising from a 1 mile long freeway segment in which traffic travels at
speed v with flow q, assuming that you are given curves such as those in the
Fundamentals of Traffic Engineering (Homburger, et al., 1996) that relate NOx
emissions by a single vehicle (in lbs per vehicle-mile) to its speed.

Problem 7.12*

Because of tides, an inland port's shallow river access can only be entered during two
3-hour periods each day. This causes a delay to ships waiting to enter the port; and
also to those wishing to leave. The port authority is contemplating a dredging project
that would essentially eliminate all the delays, at a substantial cost. Assuming that
the opportunity cost of ship delay is $ 25,000 per hour waited and that the port expects
200 ships each year, your job as a transportation specialist is to determine the
economic benefit from the dredging project. You are to:

(a) Develop a yearly benefit estimate if ships arrive randomly in time, travel the
river section in one hour (the one way access time is TA=1 hour) and spend
TP=17 hours in port.

(b) Describe how the problem under consideration is analogous to a traffic flow
problem through an arterial street section controlled by two traffic lights.
(Include a brief description of both systems, explain any assumptions needed for
the equivalence to hold, and identify the parameters/variables that are
mathematically equivalent).

(c) Develop an expression for the annual savings as a function of the ship
turnaround time, T = TA+TP, and plot it.

(d) Without any complex calculations (an argument worth two or three lines of text
suffices), plot the curve of part (c) if the ship service times are random with a
standard deviation that is large compared with 12 hrs.

March 16, 1998

 7-7

(e) Explain how your results might change if ships can be scheduled to arrive any
time (e.g. when the tide is up).

Problem 7.13*

Cities A and B are connected by a congested bridge (1) on which there is a steady state
flow of Q cars per hour. There is one person per car and no toll is charged. You believe
that an increasing relationship of the form y = c(x) exists for the time delay to the users
of the bridge y (in hours) when the flow is x veh/hr. Because y is large, we contemplate
the idea of allowing an entrepreneur to build and operate a parallel identical bridge
(2) with the same delay function, c; and to charge any toll, - ($).

We agree with the entrepreneur that the flow from A to B, Q , would not change and
that the flows on the two bridges, x1 + x2 = Q, would be the result of an equilibrium in
which each driver would choose the cheapest bridge, i = 1, 2 , as measured by the two
generalized costs, Zi ($), experienced by the driver.

These costs Zi are defined for each bridge as the sum of the toll,- i = 0 or -, and the
cost-equivalent of delay at equilibrium, yi, where  ($/hrs), is a characteristic of each
driver that varies across drivers. The fraction of the population with   z is given by
a known function, F(z).

(a) Set up the equations for the toll that maximizes the entrepreneur's revenue.
(This is what the entrepreneur would do if the cost of operating the new bridge
is independent of flow).

(b) Find the toll if Q = 10,000, c(x) = 1 + (x/5,000)4, and F(z) = z/10 for 0  z  10.

(c) Calculate the increase in consumers' surplus ($/hour) induced by the new bridge
(1) for the users of the new bridge and (2) for the remaining ones. (Aside from
an irrelevant additive constant, surplus is the negative sum of the generalized
costs across drivers).

(d) Repeat (c) if we choose to express consumers' surplus as the sum of the
generalized delays experienced, defined for each driver as: Zi/. Comment on
the results of (c) and (d).

March 16, 1998