Optimization Problem

Optimization of Slot Layout in Forward Area of Two-Zone Warehouse

Meiliang Zhao
September 2024

1 Problem Description
1.1 Background
In modern warehouse management, warehouses are typically divided into two main areas: the Forward Area
and the Reserve Area, to handle order demands more efficiently.
• Reserve Area: The Reserve Area is used to store large quantities of goods, usually in the form of cases
or pallets. These goods take up a large amount of space but are relatively cost-effective to manage due
to high storage density. The primary function of the Reserve Area is to supply goods to the Forward
Area. Since goods in the Reserve Area are stored in bulk, directly picking from the Reserve Area for
orders is less efficient. Therefore, these bulk goods need to be regularly broken down and replenished in
the Forward Area.
• Forward Area: The Forward Area is used to store small quantities of goods that have already been
broken down from the Reserve Area, typically in the form of single items or small packages after being
broken case from cases or pallets. The purpose of the Forward Area is to quickly respond to small-
batch, multi-variety demands that frequently appear in orders. Due to the limited space and capacity of
the Forward Area, it can only store goods with high demand frequency.

1.2 Core Problem

Given the known order demands over a certain period (such as a day or an hour), how can we optimally arrange
the slot layout in the Forward Area to maximize order fulfilment while adhering to the capacity and inventory
constraints of the Forward Area? We need to decide which goods and in what quantities should be placed in
each slot to maximize order fulfilment under the given capacity of the Forward Area.

1.3 problem Objective

By arranging the slot layout of the Forward Area optimally, I aim to improve order fulfilment, minimize stockouts
caused by insufficient goods in the Forward Area, and enhance the warehouse’s responsiveness to orders within
a specific timeframe.

2 Model Assumptions
2.1 Demand Assumptions
• Demand Stability Assumption: It is assumed that customer demand is stable and known during
the study period. All order quantities are determined before optimization, without considering seasonal
fluctuations or sudden demand changes.
• Order Distribution Assumption: All orders are evenly distributed within the study period (e.g., a
day or an hour), with no peak or low periods in order distribution.

2.2 Storage and Replenishment Assumptions

• Sufficient Inventory Assumption: It is assumed that the inventory in the Reserve Area is sufficient
to support the replenishment needs of the Forward Area, so that no stockouts occur in the Forward Area
due to insufficient inventory in the Reserve Area.
• Timely Replenishment Assumption: It is assumed that each replenishment operation can be com-
pleted within one time cycle, i.e., replenishment is instantaneous and there is no delay.
Optimization Problem

2.3 Forward Area Layout Assumptions

• Forward Area Capacity Assumption: The total capacity of the Forward Area is fixed, meaning that
the total capacity of all slots in the Forward Area does not change with layout adjustments.
• Single Item Per Slot Assumption: Each slot can only store one type of goods, and mixing different
types of goods in the same slot is not allowed.

2.4 Other Assumptions

• Order Priority Assumption: It is assumed that all orders have the same priority, and there is no
specific order that needs to be prioritized.

• Cost Consideration Assumption: Transportation costs, picking costs, and other operational costs are
not considered; only order fulfilment rate and replenishment frequency are focused on.

3 Parameter Definitions

Symbol Description
N Total number of slots in the Forward Area.
M Total number of types of goods in the warehouse (e.g., a, b, c, ...).
Si Capacity of the ith slot (in units), representing the maximum number of units
that can be stored in each slot.
Qj Total inventory of goods j in the Reserve Area, available for replenishment to
the Forward Area.
Dk Demand quantity of the k th order, including the quantity required for each
type of goods.
O Total number of orders.
xij Quantity of goods j stored in the ith slot. For example, xij = 5 means there
are 5 units of goods j stored in the ith slot.
ykj Fulfilled quantity of goods j for order k. For example, ykj = 3 means 3 units
of goods j have been picked from the Forward Area to fulfil order k.
Ctotal Total capacity∑ of the Forward Area, which is the sum of the capacities of all
N
slots, Ctotal = i=1 Si .
ϵ A very small constant is used to prevent division by zero and ensure computa-
tional stability.
Table 1: Parameter definitions

3.1 Decision Variables

• xij �The quantity of product j stored in slot i.
– Explanation� This is the primary decision variable, representing the number of products stored in
each slot in the forward area. Proper allocation of these quantities can help increase the number of
fully satisfied orders within a given time period.
– Value Range�xij ≥ 0, and it cannot exceed the slot’s maximum capacity Si .
• zij �A binary variable indicating whether product j is stored in slot i.
– Explanation�This is an auxiliary variable used to indicate whether product j is placed in slot i. The
configuration of this variable will directly affect order satisfaction, helping to maximize the number
of fully satisfied orders.
– Value Range�zij ∈ {0, 1}, where zij = 1 if product j is stored in slot i; otherwise, zij = 0.

4 Objective Function
∑
O
Maximize Z = Ik (1)
k=1
Optimization Problem

Where:
• Z is the objective function, representing the total number of fully satisfied orders.
• O is the total number of orders.
• Ik is an indicator function for whether order k is fully satisfied:

– Ik = 1 indicates order k is fully satisfied;

– Ik = 0 indicates order k is not fully satisfied.

5 Constraints
5.1 Capacity Constraints
xij ≤ Si ∀i ∈ {1, 2, . . . , N }, ∀j ∈ {1, 2, . . . , M } (2)
Explanation: The quantity of goods j stored in slot i cannot exceed the maximum capacity of slot Si .

∑
N ∑
M
xij ≤ Ctotal (3)
i=1 j=1

Explanation: The total quantity of goods stored in all slots in the Forward Area cannot exceed the total
capacity of the Forward Area, Ctotal .

5.2 Inventory Constraints

∑
N
xij ≤ Qj ∀j ∈ {1, 2, . . . , M } (4)
i=1

Explanation: The total quantity of goods j stored in all slots in the Forward Area cannot exceed the total
inventory of goods j in the Reserve Area, Qj .

5.3 Order Fulfillment Constraints

ykj ≤ Dkj ∀k ∈ {1, 2, . . . , O}, ∀j ∈ {1, 2, . . . , M } (5)
Explanation: The quantity of goods j fulfilled for order k, ykj , cannot exceed the actual demand for goods
j in order k, Dkj .

∑
N ∑
O
xij ≥ ykj ∀j ∈ {1, 2, . . . , M } (6)
i=1 k=1

Explanation: The total quantity of goods j stored in the Forward Area must be greater than or equal to the
total quantity of goods j fulfilled across all orders.

5.4 Non-negativity Constraints

xij ≥ 0 ∀i ∈ {1, 2, . . . , N }, ∀j ∈ {1, 2, . . . , M } (7)
Explanation: The quantity of goods stored in each slot cannot be negative.

ykj ≥ 0 ∀k ∈ {1, 2, . . . , O}, ∀j ∈ {1, 2, . . . , M } (8)

Explanation: The quantity of goods fulfilled for each order cannot be negative.
Optimization Problem

6 Data Simulation Description

In the model, I reflect the order fulfilment situation in actual scenarios by randomly generating order demand
and inventory data. My goal is to maximize the number of fully satisfied orders in each time period and optimize
the layout of slots in the front area and the distribution of goods through metaheuristic algorithms.
• The data in the model is divided into fixed parameters and random variables.
• Fixed parameters include the number of slots in the front area, capacity, and types of goods, which remain
constant throughout the simulation.
• Random variables include order demand and storage inventory, which are generated randomly to reflect
the potential fluctuations in orders and inventory in reality.
• In addition, the time period for order generation is set to simulate the order demand situation within a
day.

6.1 Simulated Data Table

Category Variable Value/Range Description

Number of Slots in Front Area N 10 The front area is set with a total of 10
slots.
Fixed Parameters
Capacity of Each Slot Si 20 items/slot The maximum capacity of each slot is
20 items of goods.
Number of Goods M 10 types of goods The warehouse has 10 different types of
goods.
Total Number of Orders O Multiple occurrences daily Orders are generated multiple times a
day, aiming to maximize the number of
fully satisfied orders through a reason-
able slot layout.
Order Demand Dkj 1 to 20 items/good The demand for each type of good in
Random Variables
each order is randomly generated be-
tween 1 and 20 items.
Storage Inventory Qj 50 to 200 items/good The inventory of each type of good in
the storage area is randomly generated
between 50 and 200 items to supple-
ment the front area.
Working Hours 8 hours The working time is 8 hours a day, from
8 AM to 4 PM.
Time Period
Order Demand Generation Frequency Generated every 2 hours A batch of order demands is generated
every 2 hours, resulting in 4 batches of
order demands per day.
Number of Orders Each Time 5 to 20 orders The number of orders generated each
time is randomly generated between 5
and 20 orders.
Distribution of Order Demand Random The demand for each type of good Dkj
is randomly generated between 1 and
20 items each time an order demand is
produced.

Table 2: Simulated data Table

7 Experimental Design
7.1 Objective
The main objective of this experiment is to optimise the slot layout in the front area of a two-zone warehouse
using a simulated annealing algorithm. The goal is to maximise the number of fully fulfilled orders in a given
time. By adjusting the temperature, the cooling rate and the number of iterations, we aim to analyse how
different parameter settings affect the algorithm’s performance in order fulfilment.

7.2 Data Generation

The data for this experiment is generated randomly to simulate realistic warehouse conditions. There are two
types of data: inventory data and order demand data.
For the inventory data, each product’s stock in the reserve area is randomly generated between 50 and 200
units.
Optimization Problem

For the order demand data, orders are generated in four time periods throughout the day. Each order
contains random demand for each product, with quantities ranging from 1 to 20 units. The number of orders
per period is randomly generated, ranging from 5 to 20 orders per period.
These randomly generated data sets are used as inputs for the optimization algorithm.

7.3 Parameter Settings

In this experiment, we randomly generated three sets of order data and conducted two experiments on each
set with different parameter configurations. This results in a total of six different experimental setups for
comparison.
The three key parameters adjusted in each trial were temperature, cooling rate, and iteration count.

• Temperature (temp): The initial temperature controls the algorithm’s exploration capability. Higher
temperatures allow the algorithm to accept worse solutions at the beginning to avoid local optima. For
each set of data, the initial temperature was set to 1000 and 1200 in two separate experiments.
• Cooling Rate (cooling_rate): This parameter determines how fast the temperature decreases. A
slower cooling rate gives the algorithm more time to explore solutions, while a faster rate leads to quicker
convergence. The cooling rate was set to 0.95 and 0.98 in the two experiments.
• Iteration Count (max_iter): The number of iterations controls how long the algorithm runs. More
iterations allow the algorithm to search for better solutions. The iteration count was set to 1000 and 2000,
respectively, in the two experiments.

In total, six experiments were conducted using three different sets of order data and two different parameter
configurations for each set. These six configurations are used for comparison to evaluate the impact of different
parameter settings on the optimization results.

7.4 Experimental Steps

The experiment was conducted following these steps:

1. Data Generation: I randomly generated three sets of order data. Each set contains order demands
for multiple products over four time periods. The demand for each product in each order was randomly
generated between 1 and 20 units.
2. Initial Inventory Setup: For each set of order data, I also generated the initial inventory for the reserve
area. The stock for each product was randomly generated between 50 and 200 units.

3. Algorithm Execution: For each set of order data, I ran the simulated annealing algorithm twice, once
with the parameter configuration of 1000 (temp), 0.95 (cooling rate), and 1000 (iterations), and once with
1200 (temp), 0.98 (cooling rate), and 2000 (iterations).
4. Recording Results: For each run, I recorded the number of fully satisfied orders, as well as the to-
tal number of goods fulfilled. I repeated this process for all three sets of order data, resulting in six
experimental runs in total.

7.5 Experimental Results

7.5.1 Data Set A Results
The first set of experiments was conducted using Data Set A, which consists of four time periods with varying
order quantities. For each time period, two different parameter configurations were tested: one with a temper-
ature of 1000, a cooling rate of 0.95, and 1000 iterations, and the other with a temperature of 1200, a cooling
rate of 0.98, and 2000 iterations.
Table 3 shows the comparison of fully satisfied orders for each time period under both parameter configura-
tions.
Optimization Problem

Time Period Order Quantity Parameter Configuration Fully Satisfied Orders

1000, 0.95, 1000 7 orders
Period 1 11 orders
1200, 0.98, 2000 9 orders
1000, 0.95, 1000 6 orders
Period 2 9 orders
1200, 0.98, 2000 7 orders
1000, 0.95, 1000 6 orders
Period 3 8 orders
1200, 0.98, 2000 7 orders
1000, 0.95, 1000 7 orders
Period 4 7 orders
1200, 0.98, 2000 7 orders

Table 3: Results for Data Set A under two different parameter configurations.

7.5.2 Data Set B Results

The second set of experiments was conducted using Data Set B, which consists of four time periods with
varying order quantities. For each time period, two different parameter configurations were tested: one with a
temperature of 1000, a cooling rate of 0.95, and 1000 iterations, and the other with a temperature of 1200, a
cooling rate of 0.98, and 2000 iterations.
Table 4 shows the comparison of fully satisfied orders for each time period under both parameter configura-
tions.

Time Period Order Quantity Parameter Configuration Fully Satisfied Orders

1000, 0.95, 1000 8 orders
Period 1 9 orders
1200, 0.98, 2000 9 orders
1000, 0.95, 1000 9 orders
Period 2 10 orders
1200, 0.98, 2000 9 orders
1000, 0.95, 1000 9 orders
Period 3 18 orders
1200, 0.98, 2000 10 orders
1000, 0.95, 1000 9 orders
Period 4 9 orders
1200, 0.98, 2000 8 orders

Table 4: Results for Data Set B under two different parameter configurations.

7.5.3 Data Set C Results

The third set of experiments was conducted using Data Set C, which consists of four time periods with varying
order quantities. For each time period, two different parameter configurations were tested: one with a temper-
ature of 1000, cooling rate of 0.95, and 1000 iterations, and the other with a temperature of 1200, cooling rate
of 0.98, and 2000 iterations.
Table 5 shows the comparison of fully satisfied orders for each time period under both parameter configura-
tions.

Time Period Order Quantity Parameter Configuration Fully Satisfied Orders

1000, 0.95, 1000 6 orders
Period 1 9 orders
1200, 0.98, 2000 7 orders
1000, 0.95, 1000 7 orders
Period 2 14 orders
1200, 0.98, 2000 7 orders
1000, 0.95, 1000 8 orders
Period 3 20 orders
1200, 0.98, 2000 9 orders
1000, 0.95, 1000 7 orders
Period 4 16 orders
1200, 0.98, 2000 7 orders

Table 5: Results for Data Set C under two different parameter configurations.