Received July 28, 2020, accepted August 8, 2020, date of publication August 12, 2020, date of current version

August 21, 2020.

Digital Object Identifier 10.1109/ACCESS.2020.3015976

Surface Optimal Path Planning Using an Extended

Dijkstra Algorithm
MIN LUO 1,2 , XIAORONG HOU 1, AND JING YANG1
1 School of Automation Engineering, University of Electronic Science and Technology of China, Chengdu 611731, China
2 School of Electrical Engineering and Information, Southwest Petroleum University, Chengdu 610500, China

Corresponding author: Xiaorong Hou ([email protected])

ABSTRACT Extensive studies have been conducted on the Dijkstra algorithm owing to its bright prospect.
However, few of them have studied the surface path planning of mobile robots. Currently, some application
fields (e.g., wild ground, planet ground, and game scene) need to solve the optimal surface path. This
paper proposes an extended Dijkstra algorithm. We utilize the Delaunay triangulation to model the surface
environment. Based on keeping the triangle side length unchanged, the triangle mesh on the surface is
equivalently converted into a triangle on the two-dimensional plane. Through this transformation, we set
up the two-dimensional developable passable channel of the surface and solve the optimal route on this
channel. Traversing all the two-dimensional developable and passable paths of the surface, we can get the
shortest route among all the optimal paths. Then the inverse transformation from the two-dimensional plane
coordinates to the corresponding surface coordinates obtains the surface optimal path. The simulation results
show that, compared with the traditional Dijkstra algorithm, this method improves the accuracy of the surface
optimization path in single-robot single-target and multi-robot multi-target path planning tasks.

INDEX TERMS Dijkstra algorithm, path planning, surface, Delaunay triangulation, mobile robots, opti-
mization methods.

I. INTRODUCTION constraints at the same time. This research enabled GRTOP

Dijkstra algorithm is a classical well-known shortest path to reprogram traversal/task frequently, and optimized traver-
routing algorithm in 2D mobile robots’ path planning sal and task security. The authors exploited the diamond
researches. It is a simple algorithm for the single-source square algorithm to create realistic terrain in the surface
shortest path problem, which can effectively calculate the environment. The weight between two consecutive points was
shortest path to all destinations [1]–[4]. taken as the Euclidean distance between them. The Dijkstra
The Dijkstra algorithm was introduced by Dutch computer algorithm was extended to consider multiple targets, and the
scientist Edsger Wybe Dijkstra in 1959. It has been success- weights between adjacent nodes were set as a linear combina-
fully applied in fields like mobile robot 2D path planning, tion of multiple weights. Each weight corresponded to a spe-
computer science, geographic information science, and trans- cific target. Through these methods, the Dijkstra algorithm
portation, etc. Some recent research based on the Dijkstra was applied to the calculation of a three-dimensional optimal
algorithm is shown below [5]–[18]. path and could be extended to multi-objective tasks.
Wolfgang Fink et al. adopted a multi-objective variant of Dong Guo et al. improved the traditional Dijkstra algo-
the Dijkstra algorithm based on terrain data to achieve the rithm and combined it with the vehicle fuel consumption
overall optimal traversal in the 3D surface (2019) [5]. The and emission measurement model to reduce vehicle fuel
gained results were employed in the Global Rover Hori- consumption and emissions effectively during driving (2019)
zontal Optimization Planner (GRTOP) automation system to [6]. This method employed a rectangular area (the smallest
quickly and accurately set up optimized routes for multiple bounding rectangle of the ellipse) to limit the search area,
thereby improving the efficiency of the Dijkstra’s algorithm.
The associate editor coordinating the review of this manuscript and According to the time of one day, the Dijkstra’s algorithm
approving it for publication was Christopher Kitts . and an established database were used to identify the traffic

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/
VOLUME 8, 2020 147827
 M. Luo et al.: Surface Optimal Path Planning Using an Extended Dijkstra Algorithm

situation at the same time. The improved Dijkstra algorithm (2018) [11]. Based on the path planning and driving strategy,
could not only reduce vehicle fuel consumption and emis- the optimal trajectory was generated by curve fitting technol-
sions but also avoid time congestion. This method could ogy, which fully considered the safety and dynamics of the
be used for vehicle path planning based on dynamic traffic driverless bus. This method could improve control accuracy,
networks, reducing fuel consumption and emissions during reduce the computational complexity, and promote driving
driving, and improving urban environmental pollution. efficiency.
Felipe Ribeiro Souza et al. applied the Dijkstra’s method Zheng Zhang et al. adopted an improved Dijkstra algo-
to tree diagram analysis, using mining blocks as nodes of the rithm to determine the initial path of each task in the
tree for analysis, and used to calculate the lowest cost route environment diagram describing the Automatic Guided
to transport mining blocks to their destination (2019) [7]. Vehicle (AGV) in the grid method (2018) [12]. These authors
The transportation cost was reflected in the arc of the graph, proposed a collision-free routing method for AGVs in an
and it could use Euclidean distance or transportation time to automated warehouse based on collision classification. This
calculate the minimum path. The results obtained by the Dijk- method could deal with possible collisions in automated
stra algorithm provided a non-operational path, to overcome warehouses.
this problem, adjustments were made through non-parametric Feristah Dalkilic et al. used the Dijkstra’s algorithm to
equations. In this way, the transportation cost of each block reduce search space and runtime by applying stage-specific
of the model could be determined. Paths based on Euclidean rules and utilized the algorithm in an intelligent itinerary
distance and transit time tended to increase for deeper mines. planning system to assist passengers in itinerary planning
Identifying the areas with the largest growth and quantifying (2017) [13]. The paper introduced a progressive path search
their value correctly could improve the efficiency of mining algorithm to settle this problem, taking into account the num-
planning. ber of transmissions and travel time. This method obtained
Afonso Henriques Moreira Santos et al. used graph theory a trip planning system by integrating route and timetable
and the Dijkstra’s shortest path algorithm for vertex location, information from different transportation agencies. The sys-
completed tower positioning based on dynamic programming tem was managed to help users make better use of public
to find the optimal vertex set along the route (2019) [8]. This transportation to simplify trip planning.
solution was used to solve the expansion planning problem Sai Shao et al. applied a dynamic Dijkstra algorithm to
of the new Transmission Line (TL), and its goal was to find a determine the shortest path between any two adjacent nodes
design solution with minimum cost. This method utilized the on the path (2017) [14]. The paper designed an electric vehi-
Dijkstra’s shortest path algorithm to optimize the transmis- cle routing scheme with variable charging time and travel
sion line vertices and calculate the total cost from the source time. Its purpose was to solve the dilemmas of electric vehicle
node to the sink node. The results showed that this method mileage limitation and charging demand.
has a lower design cost than the original TL. Georgios K.D. Saharidis et al. combined the Dijkstra’s
Jesús Balado et al. applied the Dijkstra pathfinding algo- algorithm with a Mixed Integer Linear Programming (MILP)
rithm to the developed urban scene graph, and realized the model to gain the optimal trip (2017) [15]. This multi-mode
task of directly using point clouds in the urban environ- path solution could make people prefer to accept the mini-
ment for pathfinding (2019) [9]. The method proposed in mum GreenHouse Gas (GHG) emission in various modes of
the paper could automatically set up a graph representing transportation when traveling. This method could be used in
pedestrian navigable urban space, on which the safe and real the construction plan of the public transportation operation
routes of pedestrians under different movement skills could platform to achieve the best travel route for cutting emissions.
be calculated. The Dijkstra algorithm was utilized to develop Tan Zhi et al. presented an improved ant colony algorithm
safe routes in real-time graphics. The generated paths could to balance the energy consumption of wireless sensor net-
be employed to make valid obstacle avoidance routes for works by studying the theory of Dijkstra’s algorithm (2015)
pedestrians and wheelchairs. [16]. The improved ant colony algorithm could increase the
Jinchuan Tang et al. studied the optimal path selection life cycle of wireless sensor networks.
method based on the Dijkstra algorithm and combined with W. C. Lu et al. used the Dijkstra’s algorithm to settle
three probabilistic results for the design of Mission-Critical the feasible air route planning issue, and the shortest path
Push-To-Talk (MCPTT) system for 5G public safety disaster between the earned airport and the training area could min-
relief network (2019) [10]. The Dijkstra algorithm was used imize the impact on the crowd and the threat to the aircraft
to select the best connection, delay, and trust routing. Aiming (2013) [17]. This paper aimed to explore the feasible air
at the MCPTT system, the thesis proposed a routing method routes for light sports aircraft to minimize the impact on
based on connection, delay, and trust to provide the best residential areas and the threat of terrain obstacles to the
connection delay trust performance. aircraft.
Based on the ArcGIS analysis tool, the Dijkstra algorithm Deepak Gautam et al. applied the Dijkstra’s algorithm to
was employed by Lingli Yu et al. for global path plan- bypass obstacles and locate the shortest path from a given
ning to accomplish the path planning and navigation con- initial position to the final position (2013) [18]. The pur-
trol system design of 12 meters long driverless electric bus pose of the paper to select the Dijkstra’s algorithm was

147828 VOLUME 8, 2020

M. Luo et al.: Surface Optimal Path Planning Using an Extended Dijkstra Algorithm

to study the path planning of quadrotor helicopters in a closed path attained by the method proposed in this paper is more
known environment. accurate, shorter, and smoother when calculating the optimal
The earlier improved Dijkstra algorithm mainly focused path of the surface. Especially for the case of a large number
on the improvement of the time complexity of the traditional of nodes, a large per-unit scale, and a large surface rugged-
Dijkstra algorithm and the promotion of the Dijkstra algo- ness, the optimal path obtained by the method proposed in
rithm to different fields [19]–[26]. this paper has obvious advantages.
There are three main ways to improve the traditional Dijk- This article first presents the basic principle of the tra-
stra algorithm. One is to analyze and improve the space ditional Dijkstra algorithm (Section 2). Second, it explains
complexity of the algorithm to improve storage efficiency and the improved Dijkstra algorithm theory (Section 3). In this
save space. The second is to analyze and improve the time part, we utilize the Delaunay triangulation method to con-
complexity of the algorithm. The traditional Dijkstra algo- struct the surface map. The key to extending the classical
rithm has low efficiency and long running time. To improve Dijkstra algorithm is the invariance of the triangle side length
operating efficiency and reduce time complexity, many doc- when transforming a triangle on a surface into an equiva-
uments have done a lot of work to improve this. The third lent triangle on a 2D plane. We convert the undevelopable
aspect is to apply the algorithm to different fields, open up the surface channel into an equivalent two-dimensional passable
application space of the algorithm and enrich the application channel, then solve the best path on the all two-dimensional
field of the algorithm [5]–[26]. This article is to research passable channels, and finally obtain the best path of the
in the third aspect, extending the Dijkstra algorithm to the surface through equivalent inverse transformation. We com-
optimal path on the curved surface to obtain a more accurate pare the simulation results of different surface examples to
shortest path. verify the effectiveness of the extended Dijkstra algorithm
All the above studies are based on the Euclidean distance (Section 4). At the end of this paper, a conclusion is given
between neighbor nodes in the Dijkstra algorithm path plan- (Section 5).
ning [6]–[26]. The above papers rarely involve the optimal
path of the surface [5]. Wolfgang Fink et al. utilized the II. THE TRADITIONAL DIJKSTRA ALGORITHM
Dijkstra algorithm to solve the 3D surface path planning task, The initial work of the Dijkstra algorithm is only dealing
they still used the Euclidean distance between the two nodes with the shortest path between two points. Mathematically,
to calculate the optimal path, and their improvement was to these points must be represented by nodes in the graph net-
take the weight as a composite factor, which comprehen- work. Bellman Ford implemented the possibility of fixing a
sively considered the effects of three-dimensional Euclidean point and determining the shortest path to all other points in
distance, smoothness, roughness, height change, and other the graph. Paul carried out a common practical application,
factors. using this algorithm to figure out the shortest path between
Many practical problems can be abstracted and trans- two cities, considering the street and highway to the destina-
formed into the optimal surface path issue of mobile robots, tion. Sniedovich proposed a clear and structured step division
such as the field rescue and material transportation route to determine the minimum path between two points in a node
planning, the planetary ground exploration and development network [1]–[4].
path planning, 3D game ground travel and war path planning, The Dijkstra algorithm (Algorithm 1) has a simple proce-
etc. The disadvantage of employing the traditional Dijkstra dure. The essence of the traditional Dijkstra algorithm is to
algorithm directly in surface path planning is that the cal- find out the shortest path between two nodes on a digraph
culation of each intermediate path weight is based on the D = (N , W ), where N is the set of all nodes and W is the set of
Euclidean distance between adjacent nodes, which will bring weighted edges of connected nodes. The Dijkstra algorithm
about errors that cannot be ignored in the surface path plan- separates N into two sets, Ne and Nu . Ne is a set of all the
ning. Because the optimal path obtained by calculating the end nodes of the determined shortest path to source node ns .
Euclidean distance between nodes may not be on the surface, In the first step, Ne contains only ns . Nu is the set of nodes
it is one of the most fundamental reasons for the optimal to ns with the undetermined shortest path. The nodes in Nu
path error of the surface. All of these represent that new tools will be moved to Ne in ascending order of the shortest path
and methods are required to improve the surface optimal path length of the source node ns until there are no nodes in the
planning process based on the traditional Dijkstra algorithm. set Nu . The path that sequentially connects the source node
The purpose of this paper is to establish a new general solu- ns to all edges of any node nti is the shortest path from ns to
tion method with a higher precision, which is more suitable nti . The sum of the corresponding weights is the length of the
for the solution of the optimal path of the surface than the Dijkstra algorithm’s shortest path. So, the shortest path from
traditional Dijkstra model. the source node ns to the target node nt can be won.
The main contributions of this paper are as follows: From Algorithm 1, we can see that the logic of the tra-
1. Propose an approach for calculating the optimal path on a ditional Dijkstra algorithm is to find the Euclidean distance
curved surface; 2. Improve the traditional Dijkstra algorithm between nodes, so it can be used to obtain the shortest path
for calculating the optimal path on a curved surface; 3. When from the start point to the endpoint of a two-dimensional
compared with the traditional Dijkstra algorithm, the optimal graph. However, when Algorithm 1 is directly applied to

VOLUME 8, 2020 147829

 M. Luo et al.: Surface Optimal Path Planning Using an Extended Dijkstra Algorithm

FIGURE 1. Surface Delaunay triangulation model. (a) The original surface. (b) 169 nodes Delaunay triangulation model. (c) 625 nodes
Delaunay triangulation model.

calculate the shortest path between two points on a curved The accuracy and resolution of the surface map depend on
surface, the greater the curvature of the surface, the greater the the size of the triangle mesh. The smaller the mesh, the higher
error. The Euclidean distance between nodes and the distance the accuracy. The choice of mesh size depends on the needing
between two points on the surface are two different issues. resolution requirement, as shown in Fig. 1.
For a particular deterministic surface map, choosing a
III. THE EXTENDED DIJKSTRA ALGORITHM larger mesh size map grid cell means fewer nodes and less
In this section, the extended Dijkstra algorithm is intro- resolution. Fig. 1 (a) shows an original surface map. And
duced to resolve the surface optimal path planning task. Fig. 1 (b) shows an example of the large mesh size surface
The extended Dijkstra algorithm is an algorithm to transfer map model of Fig. 1 (a), Fig. 1 (b) has 169 nodes to repre-
surface terrain map into a 2D map along the passable paths sent the surface characteristics information. Similarly, Fig. 1
based on the invariance of the triangle side length for surface (c) shows an example of the small mesh size surface map
Delaunay Triangulation grid map. model of Fig. 1 (a), Fig. 1(c) has 625 nodes. Fig. 1 (c) has
The first step we shall do is to model a surface map by a higher resolution and can better represent the fluctuation
the Delaunay triangulation method. Delaunay triangulation characteristics of the surface of Fig. 1 (a). Fig. 1 (b) has fewer
is one of the most commonly used triangular mesh modeling nodes, less mathematical complexity, and less program cost
methods. Compared with the square grid map, the Delaunay time. Therefore, in the actual design, the mesh size can be
triangulation algorithm can provide more accurate surface reasonably selected according to the required accuracy and
information, and generate a smoother path [27], [28]. Com- time cost.
pared with the digital point cloud method, it is simpler and In Fig. 2 and Fig. 3, the red polygons are the obstructions
more convenient. that the robot cannot travel, and the other areas inside the map
The expression of the Delaunay triangle mesh subdivision boundary are passable. The robot is R and the target is T. The
algorithm makes each triangle unit has 12 adjacent units. numbers denoted in the Figs are the node serial numbers.
Therefore, one non-boundary node can provide 12 feasible Fig. 2 (a) shows an original surface map example.
motion directions, thus this method can provide a smooth The side length of each square (per-unit scale) is taken
motion planning of the path. The Delaunay triangular grid as 0.25 in Fig. 2. Using the traditional Dijkstra algo-
map method [29], [30] is a map representation method using rithm logic (Algorithm 1), we can get the black color
the triangular mesh as the cartographic unit, which can well curve optimal path lts(1,6,11,16) . Because the traditional
express the fluctuation characteristics of the surface. Dijkstra algorithm is based on the Euclidean distance

147830 VOLUME 8, 2020

M. Luo et al.: Surface Optimal Path Planning Using an Extended Dijkstra Algorithm

FIGURE 2. The extended Dijkstra algorithm theory diagram. (a) The original surface. (b) 16 nodes Dijkstra algorithm optimal
path (black color) and the extended Dijkstra algorithm equivalent 2D passable path (green color). (c) 16 nodes the extended
Dijkstra algorithm optimal path (magenta color).

FIGURE 3. Four algorithms optimal path length comparison diagram. (The extended Dijkstra algorithm proposed in this paper is
magenta solid line. The traditional Dijkstra algorithm is black solid line. The extended Dijkstra algorithm proposed in literature [5] is
green dotted line. The A* algorithm is blue dotted line.) (a) 169 nodes example. (b) 196 nodes example. (c) 225 nodes example.

between adjacent pnodes, the optimal path obtained is Dsg = (Nsg , Wsg ), where Nsg is the set of all nodes and Wsg
lp
ts(1,6,11,16) = (x1 − x6 )2 + (y1 − y6 )2 + (z1 − z6 )2 + is the set of weighted edges of connected nodes.
2 2 2 Then, we are going to introduce how to apply the invariant
p(x6 − x11 ) + (y6 − y11 ) + (z6 − z11 ) +
(x11 − x16 ) + (y11 − y16 ) + (z11 − z16 )2 . Obviously, we
2 2 principle of triangular side length to convert a surface triangle
can see from Fig. 2 (b), the black color curve optimal path into a triangle on the two-dimensional plane.
lts(1,6,11,16) maybe above, below, or on the surface. This Shown in Fig. 2 (b), the surface nodes 1, 2, 5, 6 are
causes errors in the optimal path of the surface. forming two three-dimensional triangulations 1s(1,2,5) and
Next, we showed the extended Dijkstra algorithm logic 1s(2,5,6) in white color. Keeping the invariance of the triangle
theory in Algorithm 2. The core idea of the extended Dijkstra side length, using Algorithm 2, the equivalent triangulation
algorithm is to find out the shortest path on a surface digraph 1p(1,2,5) and 1p(2,5,6) in green color on the 2D plane can be

VOLUME 8, 2020 147831

 M. Luo et al.: Surface Optimal Path Planning Using an Extended Dijkstra Algorithm

Algorithm 1 The Traditional Dijkstra Algorithm Framework Algorithm 2 The Extended Dijkstra Algorithm Framework
Input: Input:
A bidirected graph with weights D = (N , W ), the set A surface graph Dsg = (Nsg , Wsg ), the set of all nodes
of all nodes N , the set of weighted edges of connected Nsg , the set of weighted edges of connected nodes Wsg .
nodes W . The source node nssg , R. The target node nstg , T.
The source node ns . Output:
The target node nt . Shortest surface paths and their lengths from the source
Output: node nssg to the target node nstg .
Shortest path and the length from the source node ns to 1: Convert the triangle which has the R robot’s start point
the target node nt . on the surface into an equivalent triangle on a 2D plane
1: Initialization. based on keeping the triangle side length unchanged.
Visited nodes Ne = ns . Continue this process to find a two-dimensional equiv-
d(ns ) = 0, alent triangle adjacent to the two-dimensional equivalent
where d(nti ) is the minimum cost of the shortest path triangle obtained in the previous step along a passable
from ns to nti . path from R to T until the last one tirangle with T node
Nu = N − Ne , as the triangle vertex found.
where Nu is the set of nodes to ns with the undetermined 2: Get the 2D equivalent passable passage graph Ddgpi =
shortest path. (Ndgpi , Wdgpi ), where Ndgpi is the set of all nodes and
d(nti ) = min(W (ns , nti )), if nti is a successor to ns , or else, Wdgpi is the set of weighted edges of connected nodes
d(nti ) = ∞, on the 2D equivalent passable passage graph.
where W (i, j) is the cost between node i and node j. 3: Get the 2D source node ndsg , and the 2D target node ndtg .
2: Repeat the following steps until there are no nodes in the 4: Compute Wdgpi on the passable path in the two-
set Nu . dimensional plane.
2.1 Put the node in Nu that have the minimum cost to the 4.1 Initialization.
old ns as the new source ns . Move the new source node Visited nodes Nedgpi = ndsg .
ns to Ne ; d(ndsg ) = 0, where d(ndtgpi ) is the minimum cost of the
2.2 Set d(nti ) = d(ns ) + min(W (ns , nti )). shortest path from ndsg to ndtgpi .
3: Find the shortest path from the source node ns to the Nudgpi = Ndgpi − Nedgpi , where Nudgpi is the the set of
target node nt . nodes to ndsg with the undetermined shortest path.
4: return d(nt ), Ne . d(ndtgpi ) = min(W (ndgpi , ndtgpi )), if ndtgpi is a successor
to ndsg , or else, d(ndtgpi ) = ∞, where Wdgpi (i, j) is the
cost between the node i and j.
4.2 Repeat the following steps until there are no nodes in
found. In this way, the two-dimensional equivalent triangle the set Nudgpi :
adjacent to the two-dimensional equivalent triangle obtained Set the node in Nudgpi that have the minimum cost to
in the previous step continues to travel along the passable the old ndsg as the new source ndsg . Move the new
passage from R to T until the last triangle with T node as source node ndsg to Nedgpi . Set d(ndtgpi ) = d(ndsg ) +
the vertex of the triangle is obtained. min(W (ndgpi , ndtgpi )).
According to Algorithm 2, holding the length of each side 5: Find the other passable passage from R to T in a different
of the triangle unchanged, the triangular meshes of the curved order, and repeat Step 1 - Step 4 until find the shortest
surface can be sequentially transformed into triangles on the paths for all passable passages.
two-dimensional plane one by one, to achieve an equivalent 6: Compute d(nstg ) and Nesg of the surface by equivalent
passable channel on the two-dimensional plane. Solve the inverse coordinate transformation from the shortest 2D
best path of the two-dimensional equivalent passable tri- path.
angular channel, and then traverse all the two-dimensional 7: return d(nstg ), Nesg .
passable channels to attain the optimal path from the starting
point to the endpoint of the equivalent two-dimensional plane.
Finally, a simple inverse conversion of equivalent coordinates
can secure the optimal path of the equivalent surface. be moved to Nedgpi in ascending order of the shortest path
Ndgpi is the set of all 2D nodes and Wdgpi is the set of length of the source node ndsg until there are no nodes in
weighted edges of connected nodes on the 2D passable pas- the set Nudgpi . The path that sequentially connects the source
sage. Then, we separate Ndgpi into two sets, Nedgpi and Nudgpi . node ndsg to all edges of any node ndtgpi is the shortest path
Nedgpi is a set of all the end nodes of the determined shortest from ndsg to ndtgpi . The sum of the corresponding weights
path to the equivalent 2D source node ndsg . In the first step, is the length of the 2D passable passage’s shortest path. So,
Nedgpi contains the only ndsg . Nudgpi is the set of nodes to ndsg the shortest path from the source node ndsg to the target node
with an undetermined shortest path. The nodes in Nudgpi will ndtg can be won.

147832 VOLUME 8, 2020

M. Luo et al.: Surface Optimal Path Planning Using an Extended Dijkstra Algorithm

Employing this method and by extension, we can eas- factor parameters α = 1/3, β = 1/3, γ = 1/3 that can
ily take the 2D equivalent expansion channel of a surface obtain a shorter path. For the A* algorithm, we compared the
passable path from the robot R’s initial position to the tar- simulation results of a variety of heuristic functions H (n) and
get T’s position. Thus, we can adapt the extended Dijkstra selected the H (n) that can obtain a shorter path, taking the
algorithm (Algorithm 2) to work out the optimal path on Euclidean distance and the Manhattan distance into account.
the acquired equivalent green 2D path below in Fig. 2 (b).
A graph may have multiple passable channels. Computing TABLE 1. Four algorithms optimal path length comparison.
all passable equivalent path on the 2D plane, we can get the
shortest optimal path, the magenta color curve lp(1,6,11,16)
shown in Fig. 2 (b). Finally, we use the equivalent coordinate
transformation to get the surface optimal path from the robot
R’s initial position (nssg ) to the target T’s position (nstg ),
the magenta color curve les(1,6,11,16) shown in Fig. 2(c). And
the magenta color curve is on the surface, not above or below
the surface.
From Algorithm 2, we can see utilizing the extended Dijk- It can be seen from Fig. 3 that only the optimal path
stra algorithm, the surface smooth optimal path can be gotten. obtained by the extended algorithm proposed in this paper
Therefore, in principle, the surface optimal path solved by this is smooth and the optimal paths obtained by the other three
proposed method is more accurate and shorter than the sur- algorithms are all non-smooth paths composed of polyline
face optimal path solved by the traditional Dijkstra algorithm segments. The optimal path obtained by the algorithm pro-
with the same nodes number. posed in this paper is based on a triangular mesh surface,
Furthermore, the traditional Dijkstra algorithm optimal while the optimal path obtained by the other three algo-
path is not a true path that falls on a triangulation mesh, and rithms is the space Euclidean distance or Manhattan distance
may often have path segments above or below the triangula- between the nodes. Therefore, the paths obtained by the other
tion mesh. When the robot tracks along with the traditional three algorithms all have line segments higher or lower than
Dijkstra algorithm optimal path, it cannot travel higher or the triangle surface. On the one hand, it causes errors in path
lower than the curved surface and must travel against the calculation, and on the other hand, it will cause problems
curved surface (Refer to the detailed diagram in Fig. 2 (b), such as malfunction or loss of direction when the robot is
lts(6,11) is the Euclidean space distance between node 6 and walking. More accurate paths have practical significance and
node 11, which is higher than the equivalent triangle blocks value for occasions with higher precision requirements, such
1s(6,7,10) and 1s(7,10,11) , not on these two triangles). Accord- as surgical operations on fine surfaces or finishing on surfaces
ingly, the actual path length with the traditional Dijkstra and other occasions with high error requirements.
algorithm optimal path will be much larger than the theory From Fig. 3 and Table 1, the optimal paths acquired by the
or simulation path length. The further the path is traveled or extended Dijkstra algorithm introduced in this paper are the
the greater the curvature of the curved surface, the greater shortest. As the number of nodes and the distance per-unit
the actual travel path is larger than the theoretical simulation length increase, the path difference between the traditional
path. And this path error is caused by a sharp bump or a Dijkstra algorithm and the algorithm proposed in this paper
large depression on the map, which may cause malfunctions also increases, and the maximum error is about 4.91%. The
such as the robot loses the direction of travel or walks on the maximum error of the path difference between the algorithm
wrong route. The improved Dijkstra algorithm optimal path proposed in [5] and the algorithm in this paper is about
is a true path that falls on an equivalent angulation mesh, so it 7.77%, and the maximum error of the path obtained by the A*
is only slightly little different from the surface path of the real algorithm and the algorithm in this paper is about 4.92%. The
robot. On the same surface with the same number of nodes, side length of each square (per-unit scale) is taken as 0.25 in
the extended algorithm proposed in this paper can get a more Fig. 3 (a). In Fig. 3 (b), the side length of each square is 0.5,
accurate and smoother optimal path. and in Fig.3 (c), it is 1.5.
To further illustrate the characteristics of the extended When comparing these four algorithms in this paper,
algorithm in this paper, here is a comparative analysis of we can see that the purpose of the traditional Dijkstra algo-
four algorithms, as shown in Fig. 3 and Table 1. The rithm is mainly to find the shortest path between nodes based
four algorithms for comparative analysis are the extended on the Euclidean space distance. Literature [5] introduces
Dijkstra algorithm presented in this paper (magenta solid the extended Dijkstra algorithm to select a flatter optimal
curve), the traditional Dijkstra algorithm (black solid curve), path based on factors such as the shortest path, elevation
the extended algorithm introduced in literature [5] (green difference, and ground smoothness. The A* algorithm is
dotted curve), and the A* algorithm (blue dotted curve). For based on a reasonable selection of heuristic functions, and the
the extended algorithm proposed in [5], we comprehensively search time for the shortest path is shorter than the traditional
compared the simulation results of various factor parameter Dijkstra algorithm. And the extended algorithm proposed in
schemes proposed in the article [5], and finally selected the this paper can get a smoother and shorter optimal path of

VOLUME 8, 2020 147833

 M. Luo et al.: Surface Optimal Path Planning Using an Extended Dijkstra Algorithm

TABLE 2. 1-robot 1-target surface optimal path comparison.

FIGURE 4. 1089 nodes example. (a) The extended Dijkstra algorithm

method. (b) The traditional Dijkstra algorithm method.

the surface than the shortest path obtained by the traditional

Dijkstra algorithm. Both the curves in Figure 3 and the data
in Table 1 confirm the effectiveness of the algorithm in this
paper.

IV. MATLAB SIMULATION VERIFICATION FIGURE 5. 2401 nodes example. (a) The extended Dijkstra algorithm
method. (b) The traditional Dijkstra algorithm method.
In this section, the MATLAB examples will further verify
the effectiveness of this extended Dijkstra algorithm. First, 1- The magenta color curve in Fig. 4 (a) shows the optimal path
robot 1-target surface path planning simulation experiments calculated by the extended Dijkstra algorithm. It is smoother
with large numbers of nodes by the extended Dijkstra algo- than the black curve shown in Fig. 4 (b) obtained by the
rithm are conducted and compared the simulation results with traditional Dijkstra algorithm. And the optimal path length
the traditional Dijkstra algorithm simulation optimal path in magenta color is about 11.812, it is shorter than the black
solutions. Furthermore, the simulation experiments of multi- color optimal curve length 12.582.
robot and multi-target path planning are carried out. After We do a 2401 nodes surface simulation example exper-
many experiments, the results show that the improved Dijk- iment in Fig. 5. The robot’s initial coordinate position
stra algorithm introduced in this paper can provide smoother in Fig. 5 is [0, 1, 1.133], the target position is [6, 7, 1.142],
and shorter paths in 1-robot 1-target and multi-robot multi- shown in Table 2. The magenta color optimal path length
target path planning tasks compared with the traditional Dijk- by the extended Dijkstra algorithm is about 8.798, and it is
stra algorithm. shorter than the black color optimal curve length 10.818 by
In the Figs in this section, the red polygons are obstructions the traditional Dijkstra algorithm, too.
where the mobile robot cannot travel, and the other areas A 4225 nodes surface path planning example is in Fig. 6.
inside the map boundary are passable. The robot is R and Table 2 exhibits the data of the simulation experiment. A con-
the target is T in the Figs. The extended Dijkstra algorithm clusion drawn from the comparison of experimental data
surface optimal path curve is in magenta, the black curve in Table 2 is that the extended Dijkstra algorithm is superior
indicates the traditional Dijkstra algorithm surface optimal to the traditional Dijkstra algorithm in solving the 1-robot
path. 1-target shortest path of the surface with large numbers
of nodes, the extended Dijkstra algorithm optimal path is
A. 1-ROBOT 1-TARGET SURFACE PATH PLANNING smoother and shorter. From Table 2, the maximum differ-
Firstly, we perform 1-robot 1-target surface path planning ence rate between the traditional Dijkstra algorithm and the
Matlab simulation experiments with large numbers of nodes. extended Dijkstra algorithm is about 22.96%. The side length
We use the Delaunay Triangulation Algorithm to build the of each square (per-unit scale) is 0.25 in Fig. 4 and Fig. 6, and
surface environment with the required resolution in this the per-unit scale is 0.125 in Fig. 5.
section. And Algorithm 2’s logic theory is employed to solve
the surface optimal path. B. MULTI-ROBOT MULTI-TARGET SURFACE PATH
In Fig. 4, we do a 1089 nodes surface simulation example PLANNING
experiment. The robot initial position is [0, 1, 0.120], and the Next, we perform multi-robot multi-target surface path
target coordinate position is [8, 9, 1.190], as shown in Table 2. planning Matlab simulation experiments. We extend

147834 VOLUME 8, 2020

M. Luo et al.: Surface Optimal Path Planning Using an Extended Dijkstra Algorithm

FIGURE 6. 4225 nodes example. (a) The extended Dijkstra algorithm FIGURE 8. 289 nodes 1-robot 3-target example. (a) The extended Dijkstra
method. (b) The traditional Dijkstra algorithm method. algorithm method. (b) The traditional Dijkstra algorithm method.

FIGURE 9. 625 nodes 1-robot 2-target example. (a) The extended Dijkstra
algorithm method. (b) The traditional Dijkstra algorithm method.
FIGURE 7. 625 nodes 2-robot 1-target example. (a) The extended dijkstra
algorithm method. (b) The traditional Dijkstra algorithm method.
Fig. 8 shows a 289 nodes 1-robot 3-target surface path plan-
ning example. The R robot’s initial position is [0, 1, 0.12],
Algorithm 2 to multi-robot multi-target surface path planning the T1 position is [8, 9, 1.190], the T2 coordinate position is
tasks and carry out MATLAB simulation experiments in this [5, 4.5, 0.897], and the T3 position is [1.5, 3.5, 0.589], shown
section, keeping the simulation experiment settings consis- in Table 3. The magenta color curve length in Fig. 8 (a) from
tent with the previous subsection. R to T1 is about 11.858, the magenta color curve length from
In Fig. 7, we do a 625 nodes 2-robot 1-target surface T1 to T2 is about 5.615, and the magenta color curve length
path planning example experiment. The R1 robot’s ini- from T2 to T3 is about 4.224. The black color curve length
tial position is [0, 1, 0.569], the R2 robot’s initial posi- in Fig. 8 (b) from R to T1 is about 14.626, the black color
tion is [2, 5.25, 0.789], and the target coordinate position is curve length from T1 to T2 is about 6.398, and the black
[6, 7, 0.742], as shown in Table 3. The magenta color curve color curve length from T2 to T3 is about 4.572. According
length in Fig. 7 (a) from R1 to T is about 8.956, the magenta to the experimental curves and data in Figure 8 and Table 3,
color curve length in Fig. 7 (a) from R2 to T is about 4.398. the optimal paths obtained by the extended Dijkstra algorithm
And, the black color curve length in Fig. 7 (b) from R1 to T is are always shorter and smoother than the optimal paths cal-
about 11.853, the black color curve length in Fig. 7 (b) from culated by the traditional Dijkstra algorithm, too.
R2 to T is 4.732. So, according to the experimental curves and Fig. 9 shows a 625 nodes 1-robot 2-target surface path
data in Figure 7 and Table 3, the optimal paths obtained by planning example. Fig. 10 is a 289 nodes 3-robot 2-target
the extended Dijkstra algorithm are shorter than the optimal surface example. Fig. 11 is a 441 nodes 3-robot 2-target sur-
paths calculated by the traditional Dijkstra algorithm. face example. The experimental data are all filled in Table 3.

VOLUME 8, 2020 147835

 M. Luo et al.: Surface Optimal Path Planning Using an Extended Dijkstra Algorithm

TABLE 3. Multi-robot multi-target surface optimal path comparison.

FIGURE 10. 289 nodes 3-robot 2-target example. (a) The extended FIGURE 11. Comparion graph of 4225 nodes example. (a) The extended
Dijkstra algorithm method. (b) The traditional Dijkstra algorithm method. Dijkstra algorithm method. (b) The traditional Dijkstra algorithm method.

According to the experimental curves in these three fig-

ures and the simulation data in Table 3, it is easy to prove multi-robot multi-object path planning task, the extended
that the extended Dijkstra algorithm proposed by this subsec- algorithm can obtain a more accurate and smoother optimal
tion can obtain a more accurate and shorter surface optimal path than the traditional algorithm, and usually, the path
path than the traditional Dijkstra method. From Table 3, obtained by the extended algorithm is shorter.
the maximum difference rate between the traditional Dijkstra
algorithm and the extended Dijkstra algorithm is 32.35%. V. CONCLUSION
The per-unit scale is 0.25 in Fig. 7, Fig. 9, and Fig. 11, and To improve the error of the traditional Dijkstra algorithm
in Fig. 8 and Fig. 10, the per-unit scale is 0.5. when studying the surface optimal path task, we introduce
The above simulation results show that the improved the extended Dijkstra algorithm. The reason for the error of
Dijkstra algorithm can still effectively calculate the optimal the traditional Dijkstra algorithm in researching the optimal
path of the surface in multi-robot and multi-target scenarios. path task of the surface is that it uses the Euclidean dis-
Moreover, compared with the traditional algorithm, the curve tance algorithm to calculate the path length between adjacent
is smoother and the path is always shorter. nodes. This is not a problem for the optimal path of a two-
The simulation research in this section finds that whether dimensional plane, but it will produce errors for the optimal
it is a single-robot single-object path planning task or a path of a surface. Nowadays, more and more application

147836 VOLUME 8, 2020

M. Luo et al.: Surface Optimal Path Planning Using an Extended Dijkstra Algorithm

scenarios want to find the optimal path of the surface. There- [8] A. H. M. Santos, R. M. D. Lima, C. R. S. Pereira, R. Osis,
fore, it is necessary to expand the research on the traditional G. O. S. Medeiros, A. R. D. Queiroz, B. K. Flauzino, A. R. P. C. Cardoso,
L. C. Junior, R. A. D. Santos, and E. L. C. Junior, ‘‘Optimizing routing and
optimal path algorithm to meet the needs of the new era. tower spotting of electricity transmission lines: An integration of geograph-
In this paper, we first introduce an improved Dijkstra ical data and engineering aspects into decision-making,’’ Electr. Power
algorithm approach to solve the optimal path of the surface. Syst. Res., vol. 176, pp. 1–12, Jul. 2019, doi: 10.1016/j.epsr.2019.105953.
[9] J. Balado, L. Díaz-Vilariño, P. Arias, and H. Lorenzo, ‘‘Point clouds
We utilize the Delaunay triangulation method to model the for direct pedestrian pathfinding in urban environments,’’ ISPRS
surface environment. On the Delaunay triangulation map, J. Photogramm. Remote Sens., vol. 148, pp. 184–196, Feb. 2019,
we keep the triangular side length unchanged when trans- doi: 10.1016/j.isprsjprs.2019.01.004.
[10] J. Tang, G. Chen, X. Li, and J. P. Coon, ‘‘Route selection based on
forming the triangular mesh on the surface into the corre- connectivity-delay-trust in public safety networks,’’ IEEE Syst. J., vol. 13,
sponding triangle in the two-dimensional plane. Through the no. 2, pp. 1567–1576, Jun. 2019, doi: 10.1109/JSYST.2018.2813929.
previous transformation, we attain the robot’s accessible pas- [11] L. Yu, D. Kong, X. Shao, and X. Yan, ‘‘A path planning and navigation
control system design for driverless electric bus,’’ IEEE Access, vol. 6,
sages of the two-dimensional plane. Solve the shortest path in pp. 53960–53975, 2018, doi: 10.1109/ACCESS.2018.2868339.
the two-dimensional plane, and then find the corresponding [12] Z. Zhang, Q. Guo, J. Chen, and P. Yuan, ‘‘Collision-free route plan-
shortest path on the surface through equivalent coordinate ning for multiple AGVs in an automated warehouse based on col-
inverse transformation. lision classification,’’ IEEE Access, vol. 6, pp. 26022–26035, 2018,
doi: 10.1109/ACCESS.2018.2819199.
Compared with the traditional Dijkstra algorithm, this [13] F. Dalkäç, Y. Doäan, D. Birant, R. A. Kut, and R. Yälmaz, ‘‘A grad-
approach extended the Dijkstra algorithm’s research scope ual approach for multimodel journey planning: A case study in izmir,
to the surface path planning field. The improved algorithm turkey,’’ J. Adv. Transp., vol. 2017, pp. 1–14, 2017, doi: 10.1155/32017/
5656323.
is suitable for the single-robot single-target project and the [14] S. Shao, W. Guan, B. Ran, Z. He, and J. Bi, ‘‘Electric vehicle rout-
multi-robot multi-target project. And, compared with the tra- ing problem with charging time and variable travel time,’’ Math. Prob-
ditional Dijkstra algorithm, the extended Dijkstra algorithm lems Eng., vol. 2017, Jan. 2017, Art. no. 5098183, doi: 10.1155/2017/
5098183.
can acquire more accurate and shorter surface optimal path, [15] G. K. D. Saharidis, D. Rizopoulos, A. Fragkogios, and C. Chatzigeorgiou,
which is demonstrated by the different MATLAB simulation ‘‘A hybrid approach to the problem of journey planning with the use of
examples in complex surface environments. mathematical programming and modern techniques,’’ Transp. Res. Proce-
dia, vol. 24, pp. 401–409, Oct. 2017, doi: 10.1016/j.trpro.2017.05.094.
The smoother and shorter path obtained in this paper comes
[16] T. Zhi and Z. Hui, ‘‘An improved ant colony routing algorithm for WSNs,’’
at the expense of increased time cost. In the future, we will J. Sensors, vol. 2015, pp. 1–4, Dec. 2015, doi: 10.1155/2015/438290.
combine the work of reducing time cost in the literature [31] [17] W. C. Lu, M. T. Lee, and M. W. Wang, ‘‘Route planning for light-sport
to further study the method of reducing the time cost of aircraft in constrained airspace,’’ Procedia Eng., vol. 67, pp. 140–146,
2013, doi: 10.1016/j.proeng.2013.12.013.
this algorithm, so that the algorithm can be used in practical [18] D. Gautam and C. Ha, ‘‘Control of a quadrotor using a smart self-
research more conveniently and quickly. tuning fuzzy PID controller,’’ Int. J. Adv. Robotic Syst., vol. 10, pp. 1–9,
Aug. 2013, doi: 10.5772/56911.
[19] T.-H. Kim and I.-C. Park, ‘‘High-throughput and area-efficient MIMO
REFERENCES symbol detection based on modified Dijkstra’s search,’’ IEEE Trans.
Circuits Syst. I, Reg. Papers, vol. 57, no. 7, pp. 1756–1766, Jul. 2010,
[1] K. Wei, Y. Gao, W. Zhang, and S. Lin, ‘‘A modified Dijkstra’s algorithm for
doi: 10.1109/TCSI.2009.2034235.
solving the problem of finding the maximum load path,’’ in Proc. IEEE 2nd
[20] Z. Pan, L. Yan, A. C. Winstanley, A. S. Fotheringham, and J. Zheng,
Int. Conf. Inf. Comput. Technol. (ICICT), Kahului, HI, USA, Mar. 2019,
‘‘A 2-D ESPO algorithm and its application in pedestrian path planning
pp. 10–13, doi: 10.1109/INFOCT.2019.8711024.
considering human behavior,’’ in Proc. 3rd Int. Conf. Multimedia Ubiqui-
[2] A. Alyasin, E. I. Abbas, and S. D. Hasan, ‘‘An efficient optimal path tous Eng., Qingdao, China, Jun. 2009, pp. 485–491, doi: 10.1109/MUE.
finding for mobile robot based on dijkstra method,’’ in Proc. 4th Sci- 2009.86.
entific Int. Conf. Najaf (SICN), Al-Najef, Iraq, Apr. 2019, pp. 11–14,
[21] A. Chen, A. K.-S. Wong, and C.-T. Lea, ‘‘Routing and time-slot assignment
doi: 10.1109/SICN47020.2019.9019345.
in optical TDM networks,’’ IEEE J. Sel. Areas Commun., vol. 22, no. 9,
[3] Yujin and G. Xiaoxue, ‘‘Optimal route planning of parking lot pp. 1648–1657, Nov. 2004, doi: 10.1109/JSAC.2004.833832.
based on dijkstra algorithm,’’ in Proc. Int. Conf. Robots Intell. Syst. [22] Bast, Mehlhorn, Schäfer, and Tamaki, ‘‘A heuristic for Dijkstra’s algorithm
(ICRIS), Huai’an, China, Oct. 2017, pp. 221–224, doi: 10.1109/ICRIS. with many targets and its use in weighted matching algorithms,’’ Algorith-
2017.62. mica, vol. 36, no. 1, pp. 75–88, May 2003, doi: 10.1007/s00453-002-1008-
[4] M. A. Djojo and K. Karyono, ‘‘Computational load analysis of dijk- z.
stra, A, and floyd-warshall algorithms in mesh network,’’ in Proc. [23] M. Noto and H. Sato, ‘‘A method for the shortest path search by
Int. Conf. Robot., Biomimetics, Intell. Comput. Syst., Jogjakarta, IN, extended Dijkstra algorithm,’’ in Proc. IEEE Int. Conf. Syst., Man
USA, Nov. 2013, pp. 104–108, doi: 10.1109/ROBIONETICS.2013. Cybern., Nashville, TN, USA, Oct. 2000, pp. 2316–2320, doi: 10.1109/
6743587. ICSMC.2000.886462.
[5] W. Fink, V. R. Baker, A. J.-W. Brooks, M. Flammia, J. M. Dohm, and [24] D. Cavendish and M. Gerla, ‘‘On routing with QOS constraints in ATM
M. A. Tarbell, ‘‘Globally optimal rover traverse planning in 3D using networks,’’ in The International Federation for Information Processing,
Dijkstra’s algorithm for multi-objective deployment scenarios,’’ Planet. A. Tantawy, Ed. Boston, MA, USA: Springer, 1994.
Space Sci., vol. 179, pp. 1–9, Dec. 2019, doi: 10.1016/j.pss.2019. [25] R. V. Helgason, J. L. Kennington, and B. D. Stewart, ‘‘The one-to-one
104707. shortest-path problem: An empirical analysis with the two-tree dijkstra
[6] D. Guo, J. Wang, J. B. Zhao, F. Sun, S. Gao, C. D. Li, M. H. Li, and C. C. Li, algorithm,’’ Comput. Optim. Appl., vol. 2, no. 1, pp. 47–75, Jun. 1993,
‘‘A vehicle path planning method based on a dynamic traffic network that doi: 10.1007/BF01299142.
considers fuel consumption and emissions,’’ Sci. Total Environ., vol. 663, [26] R. B. K. Dewar, S. M. Merritt, and M. Sharir, ‘‘Some modified algorithms
pp. 935–943, May 2019, doi: 10.1016/j.scitotenv.2019.01.222. for Dijkstra’s longest upsequence problem,’’ Acta Inf., vol. 18, no. 1,
[7] F. R. Souza, T. R. Cámara, V. F. N. Torres, B. Nader, and pp. 1–15, Feb. 1982, doi: 10.1007/BF00625277.
R. Galery, ‘‘Mine fleet cost evaluation–Dijkstra’s optimized path,’’ [27] M. Liu, ‘‘Robotic online path planning on point cloud,’’ IEEE Trans.
REM—Int. Eng. J., vol. 72, no. 2, pp. 321–328, Jun. 2019, doi: 10.1590/ Cybern., vol. 46, no. 5, pp. 1217–1228, May 2016, doi: 10.1109/TCYB.
0370-44672018720124. 2015.2430526.

VOLUME 8, 2020 147837

 M. Luo et al.: Surface Optimal Path Planning Using an Extended Dijkstra Algorithm

[28] A. Stumpf, S. Kohlbrecher, D. C. Conner, and O. von Stryk, ‘‘Super- XIAORONG HOU was born in Shanxi, China,
vised footstep planning for humanoid robots in rough terrain tasks in 1966.
using a black box walking controller,’’ in Proc. IEEE-RAS Int. Conf. He is currently a Professor with the School of
Humanoid Robots, Madrid, India, Nov. 2014, pp. 287–294, doi: 10.1109/ Automation Engineering, University of Electronic
HUMANOIDS.2014.7041374. Science and Technology of China. He has pub-
[29] C.-C. Sun, G. E. Jan, S.-W. Leu, K.-C. Yang, and Y.-C. Chen, ‘‘Near- lished over 90 research articles and two mono-
Shortest path planning on a quadratic surface with O(n log n) time,’’ IEEE graphs. His research interests include control
Sensors J., vol. 15, no. 11, pp. 6079–6080, Nov. 2015, doi: 10.1109/JSEN.
theory, intelligent systems, symbolic computation,
2015.2464271.
and real algebraic geometry.
[30] Z. Shiller and J. C. Chen, ‘‘Optimal motion planning of autonomous
vehicles in three dimensional Terrains,’’ in Proc. Int. Conf. Robot.
Autom., Cincinnati, OH, USA, 1990, pp. 198–203, doi: 10.1109/
ROBOT.1990.125972.
[31] M. Luo, X. Hou, and S. X. Yang, ‘‘A multi-scale map method
based on bioinspired neural network algorithm for robot path plan-
ning,’’ IEEE Access, vol. 7, pp. 142682–142691, 2019, doi: 10.1109/
ACCESS.2019.2943009.

MIN LUO was born in Sichuan, China, in 1978.

She received the B.S. degree in automation from
Chongqing University, in 2001, and the M.S.
degree in electrical engineering and information JING YANG was born in Sichuan, China, in 1989.
from Southwest Petroleum University, in 2009. She is currently a Postdoctoral Researcher with
She is currently pursuing the Ph.D. degree in the School of Automation Engineering, University
automation engineering with the University of of Electronic Science and Technology of China.
Electronic Science and Technology of China, Her research interests include nonlinear control
Sichuan. theory and fractional-order systems.
Since 2001, she has been a Lecturer with the
School of Electrical Engineering and Information, Southwest Petroleum
University. Her research interests include robot path planning and control
theory and application.

147838 VOLUME 8, 2020