Path Planning Method for Unmanned Delivery Robot Based on EWB-GWO Algorithm

Abstract

The present invention discloses a path planning method for an unmanned delivery robot based on the EWB-GWO algorithm, including the steps: S1, the strategy of generating the initial wolf pack of the GWO algorithm is corrected by introducing a two-dimensional Tent-Sine coupled chaotic map. S2, by introducing a non-linear inertia weight strategy, an opposition-based learning method based on the elite strategy, and a random walk rule integrating the butterfly optimization algorithm, the defects of slow convergence speed, low accuracy, and imbalance between global exploration and local exploitation existing in the GWO algorithm when dealing with high-dimensional complex problems are improved; S3, the line-of-sight detection technology based on the Bresenham line algorithm strengthens the rational generation of the initial population of the EWB-GWO algorithm, reduces the number of iterations of the algorithm, and reduces the time complexity of the algorithm when dealing with path planning problems. The simulation results of the present invention show that the EWB-GWO algorithm has very strong competitiveness among the same type of meta-heuristic algorithms. At the same time, the path length measurement index and the smoothness index in the path planning experiment reach the optimal.

Description

Technical Field

[0001] The present invention belongs to the field of path planning and trajectory optimization of mobile robots, and particularly relates to a path planning method for an unmanned delivery robot based on the EWB-GWO algorithm. Background Art

[0002] Initially, mobile robots were widely used in high-risk areas in the industrial and military industries. UAVs and unmanned boats have gradually become representatives of the backbone military forces of various countries due to their high concealment, flexible operation, and fearlessness of casualties. Rail-guided vehicles and automated guided vehicles (AGVs) have gradually become an indispensable part of the automated industrial chain in the industrial field due to their simple and convenient operation, strong autonomy, and high work efficiency. With the continuous rise of Industry 4.0 and emerging technologies, mobile robots play an important role in search and rescue, cargo transportation, unmanned services, geological exploration, and culture and entertainment. The progress of mobile robot autonomous navigation technology makes it possible to build a ground unmanned delivery network. In 2016, the unmanned delivery robot developed by JD Company entered the road test stage and was about to be put into operation. This technology makes it possible to achieve large-scale and contactless cargo delivery services during the epidemic to meet the living needs of a large number of consumers who cannot travel. Currently, unmanned delivery robots are still strictly restricted in urban areas, especially in environments with complex and congested traffic. The activity ranges of quite a number of delivery robots are limited to relatively fixed areas such as campuses and communities. This still fails to give full play to the advantages of unmanned delivery robots to replace people to achieve efficient, fast, and contactless delivery services in the current epidemic environment. Therefore, reasonably building an unmanned delivery network, widely applying mobile robots to efficient cargo transportation, and solving the problem of "the last mile of cargo delivery" will bring great convenience to consumers.

[0003] Unmanned delivery robots are different from traditional AGVs in terms of navigation methods, motion modes, and usage functions. Traditional AGVs are mainly based on the differential drive model and use radar and laser SLAM for navigation. Their movement speed is relatively slow, and they have a ton-level lifting capacity to meet the needs of transporting heavy goods in factories while ensuring delivery safety. Although unmanned delivery robots have rapid braking, GPS, and a vehicle motion model design structure, which to a certain extent enhances the adaptability to road transportation. However, they still face severe challenges in the delivery tasks of large and complex maps. Traditional path planning algorithms for mobile robots will lead to an exponential increase in the planning difficulty when dealing with complex environments with a large number of uncertain factors. Finding a path that meets the constraint conditions of the mobile robot between two nodes is understood as an NP-hard problem, and there is no general solution. Summary of the Invention

[0004] To solve the above problems, the present invention proposes a path planning method for unmanned delivery robots based on the EWB-GWO algorithm, which realizes the safe and efficient autonomous navigation of mobile robots by virtue of the ability of metaheuristic search algorithms to solve highly nonlinear and complex problems.

[0005] The technical solution adopted by the present invention is as follows: A path planning method for unmanned delivery robots based on the EWB-GWO algorithm, comprising the following steps:

[0006] S1. Based on the two-dimensional particularity of the path solutions generated in path planning, the strategy for generating the initial wolf pack in the GWO algorithm is modified. According to the two-dimensionality of the path solutions of the mobile robot, a two-dimensional Tent-Sine coupled chaotic map is introduced, and the perturbation value generated by Sine is used to solve the small-period characteristic of the Tent central map. While ensuring the diversity of the population of the GWO algorithm, the randomness between the two-dimensional state variables of the path nodes is improved.

[0007] S2. An improved GWO algorithm (EWB-GWO) is proposed. This algorithm combines an adaptive non-linear inertia weight strategy, a reverse learning method based on an elite strategy, and a random walk rule that integrates the butterfly optimization (BOA) algorithm, and improves the defects of the GWO algorithm in dealing with high-dimensional complex problems, such as slow convergence speed, low accuracy, and imbalance between global exploration and local exploitation functions; thereby balancing the search performance of the GWO algorithm and improving the convergence accuracy of the algorithm.

[0008] S3. The EWB-GWO algorithm is combined with the path planning of unmanned aerial vehicles. Through the line-of-sight detection technology based on the Bresenham line algorithm, the rational generation of the initial population of the EWB-GWO algorithm is strengthened, the number of iterations of the EWB-GWO algorithm is reduced, and the time complexity of the algorithm in dealing with path planning problems is reduced.

[0009] Furthermore, in the step S1:

[0010] S11. The specific expression formula of the Tent chaotic map is as follows:

[0011]

[0012] where x k represents the sequence number currently generated by the Sine chaotic map, x k+1 represents the next sequence number generated by the Sine chaotic map, and a represents the system parameter of the Sine chaotic map;

[0013] S12. The specific expression formula of the Sine chaotic map is as follows:

[0014] x z+1 = bsin(πx z)b ∈ (0, 1]

[0015] where x z represents the sequence number currently generated by the Tent chaotic map, and x z+1 represents the next sequence number generated by the Tent chaotic map. b represents the system parameter of the Sine chaotic map.

[0016] S13, the formula of the two-dimensional Tent-Sine coupled chaotic map is as follows:

[0017]

[0018]

[0019] where x n and y n represent the sequence numbers currently generated by the coupled chaotic map, and x n+1 and y n+1 represent the next two sequence numbers generated by the chaotic map. a represents the system parameter of the Tent chaotic map, λ represents the system weight value, and λ ∈ (0, 1) 。

[0020] Furthermore, the step S2 specifically includes:

[0021] S21, introducing an adaptive non-linear inertia weight in the global search stage of the GWO algorithm; as follows:

[0022]

[0023] where W GWO represents the inertia weight of the current grey wolf pack, Fit t,i represents the fitness value of the i-th grey wolf in the t-th iteration process, and Fit t,g represents the fitness value of the optimal grey wolf in the t-th iteration process. Max Iter represents the maximum number of iterations of the GWO algorithm.

[0024] S22, introducing a random walk rule based on the butterfly optimization algorithm in the iteration process of the GWO algorithm; the expression is as follows:

[0025]

[0026] where represents the i-th butterfly individual in the t-th iteration process, r represents a random number satisfying r ∈ [0, 1], and represent the z-th and k-th butterfly individuals in the solution space, and f i represents the intensity of the pheromone of the fragrance emitted by the i-th butterfly, where f iThe expression is as follows:

[0027] f i = cI α

[0028] where f i represents the perceived intensity of the pheromone of the i-th butterfly fragrance, I represents the fitness of the butterfly, that is, the stimulation intensity of the fragrance produced by the current butterfly, c represents the perception modality, and α represents the power exponent dependent on the pheromone of the fragrance.

[0029] S23. In the iterative process of the GWO algorithm, an opposition-based learning method based on the elite strategy is introduced. Specifically: by selecting elites, the opposition wolf pack is merged with the current wolf pack, and the excellent individuals are selected to enter the next iterative cycle for search. Let x i (t) and be the current individual and the opposition individual in the t-th iteration process respectively, Fit i,j (t) and be the fitness values corresponding to x i (t) and on the j-th dimension. The m (2 ≤ m ≤ N) elite individuals are represented as follows:

[0030]

[0031] where {e1(t), e2(t),..., e m (t)} represents the set composed of elite individuals, N represents the number of gray wolf individuals, and the opposition solution is represented by the following formula:

[0032]

[0033]

[0034] where η is a random number and satisfies η ∈ (0, 1), the min function represents returning the minimum value of the input sequence, and the max function represents returning the maximum value of the input sequence.

[0035] Furthermore, the calculation steps of the improved GWO algorithm specifically include:

[0036] (1) Initialize the number of gray wolf individuals N, the number of algorithm iterations Max Iter , the dimension of the solution space Dim, the conversion probability p between global search and local wandering, and the boundary conditions (ub, lb) in the GWO algorithm;

[0037] (2) Generate the initial gray wolf population through the Tent-Sine coupled chaotic mapping in the solution space. Calculate the fitness value of each individual according to the fitness function and initialize the Alpha wolf, Beta wolf, and Delta wolf.

[0038] (3) Calculate the perception intensity generated by the scent pheromone of the gray wolf individuals according to the scent formula, and determine the search behavior of the current algorithm by the size of the random number and the conversion probability.

[0039] (4) When rand < p, calculate the interval distances between the Alpha wolf, Beta wolf, and Delta wolf and the wolf pack respectively according to the gray wolf migration formula, and obtain the migration direction of the overall wolf pack by multiplying with the inertia weight; when rand ≥ p, the current algorithm executes the local wandering rule, where rand represents a random value between 0 and 1.

[0040] (5) Recalculate the Alpha wolf, Beta wolf, and Delta wolf based on the wolf pack individuals after migration.

[0041] (6) If the number of algorithm iterations reaches the termination condition, end the loop and output the optimal individual; otherwise, further generate the next generation population through the reverse learning method based on the elite strategy and return to step (3).

[0042] Further, the S3 specifically includes:

[0043] S31, Discretize and model the working space of the mobile robot through equal-area grids, and set the grid volume to 1.2 to 1.5 times the volume of the mobile robot.

[0044] S32, Generate feasible path nodes through the Tent-Sine two-dimensional chaotic mapping, and detect the visibility reachability between all generated nodes and their parent nodes and the target node through the visibility line algorithm.

[0045] S33, If the current node is simultaneously LOS reachable with its parent node and the target point, output this individual and generate the next gray wolf individual; if the current node is only LOS reachable with its parent node, retain the current node as the parent node and continue to generate the next node; if there is no reachability between the current node and its parent node, directly discard the current node.

[0046] S34, Generate the initial population in an iterative manner until all initial gray wolf individuals meet the feasible solutions, then end the algorithm.

[0047] The present invention also provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, it implements the path planning method for an unmanned delivery robot based on the EWB-GWO algorithm as described above.

[0048] The advantages and beneficial effects of the present invention are as follows:

[0049] For a mobile robot performing an unmanned delivery task, the autonomous navigation of the mobile robot will be a decisive factor in determining the success or failure of the delivery task. The path planning technology based on the autonomous decision-making of the mobile robot, as a key technology for robot navigation, has received extensive attention. Under the condition of ensuring the running safety of the mobile robot, a shorter and smoother path will significantly reduce the time for the robot to execute the task and improve the delivery efficiency. Based on the above objectives, the present invention realizes the path planning design for an unmanned delivery robot through an improved GWO algorithm, and completes the delivery task in a complex urban and suburban environment while ensuring the safety of the mobile robot. The contributions made by the present invention are summarized as follows:

[0050] (1) Due to the two-dimensional particularity of the path solutions generated in path planning, the present invention modifies the strategy for generating the initial wolf pack based on the GWO algorithm. According to the two-dimensionality of the path solutions of the mobile robot, a two-dimensional Tent-Sine coupled chaotic mapping is introduced, and the perturbation value generated by Sine is used to solve the small-period characteristic of the Tent center mapping. While ensuring the diversity of the population of the GWO algorithm, the randomness between path nodes is improved.

[0051] (2) Based on the disadvantages of the GWO algorithm in dealing with complex high-dimensional problems, such as slow convergence speed, poor convergence accuracy, and imbalance between global exploration and local exploitation capabilities, the present invention proposes an improved GWO (EWB-GWO) algorithm. This algorithm combines an adaptive non-linear inertia weight strategy, a reverse learning method based on an elite strategy, and a random walk rule that integrates the butterfly optimization (BOA) algorithm to balance the search performance of the GWO algorithm and improve the convergence accuracy of the algorithm.

[0052] (3) In the path planning design based on the EWB-GWO algorithm, the present invention strengthens the rational generation of the initial population of the EWB-GWO algorithm through a line-of-sight detection technology based on the Bresenham line algorithm, reduces the number of iterations of the EWB-GWO algorithm, and reduces the time complexity of the algorithm. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] Figure 1 is a flowchart of the present invention;

[0054] Figure 2 is the working environment of a mobile robot based on a discrete grid;

[0055] Figure 3 To determine whether the current node needs to be retained through LOS detection;

[0056] Figure 4 is the complete path generated by the mobile robot in the workspace. Specific embodiments

[0057] Next, the technical solutions in the embodiments of the present invention will be clearly and detailedly described in conjunction with the accompanying drawings in the embodiments of the present invention. The described embodiments are only a part of the embodiments of the present invention.

[0058] See Figure 1 , the specific solution adopted by the present invention is:

[0059] S1. Based on the discretization modeling of the mobile robot workspace, the present invention uses equal-area grids to divide the mobile robot workspace and obstacles, as Figure 2 shown.

[0060] S2. Regarding the generation problem of the initial gray wolf population based on the GWO algorithm, a two-dimensional Tent-Sine coupled chaotic map is added:

[0061] S21. The specific expression formula of the Tent chaotic map is as follows:

[0062]

[0063] where x k represents the sequence number currently generated by the Tent chaotic map, x k+1 represents the next sequence number generated by the Tent chaotic map, and a represents the system parameter of the Tent chaotic map.

[0064] S22. The specific expression of the Sine chaotic map is as follows:

[0065] x z+1 = bsin(πx z ) b ∈ (0, 1]

[0066] where x z represents the sequence number currently generated by the Sine chaotic map, x z+1 represents the next sequence number generated by the Sine chaotic map, and b represents the system parameter of the Sine chaotic map.

[0067] S23. The present invention couples the two mapping methods and, through the small perturbation provided by the Sine map, forces the sequence generated by the Tent map to achieve small-period escape. Specifically, the formula of the two-dimensional Tent-Sine coupled chaotic map is shown as follows:

[0068]

[0069]

[0070] where x n and y n represent the sequence numbers currently generated by the coupled chaotic map, x n+1 and y n+1 represent the next two sequence numbers generated by the chaotic map. a represents the system parameter of the Tent chaotic map, and λ represents the system weight value, and λ ∈ (0, 1) is satisfied 。

[0071] S3. The present invention proposes an improved GWO (EWB - GWO) algorithm. This algorithm combines an adaptive non - linear inertia weight strategy, an elite - based reverse learning method, and a random walk rule integrating the Butterfly Optimization Algorithm (BOA) to balance the search performance of the GWO algorithm and improve the convergence accuracy of the algorithm

[0072] S31. Aiming at the disadvantages of the GWO algorithm, such as slow convergence speed for dealing with complex functions and low accuracy of the optimization results, an adaptive non - linear inertia weight is introduced in the global search stage of the GWO algorithm as follows:

[0073]

[0074] where W GWO represents the inertia weight of the current grey wolf pack. Fit t,i represents the fitness value of the i - th grey wolf in the t - th iteration process. Fit t,g represents the fitness value of the optimal grey wolf in the t - th iteration process. Max Iter represents the maximum number of iterations of the GWO algorithm

[0075] S32. A random walk rule based on the Butterfly Optimization Algorithm is introduced in the iterative process of the BOA algorithm; the expression of the random walk rule based on the Butterfly Optimization Algorithm is as follows:

[0076]

[0077] where represents the i - th butterfly individual in the t - th iteration process. r represents a random number satisfying r ∈ [0, 1]. and represent the z - th and k - th butterfly individuals in the solution space. f i represents the intensity of the pheromone of the fragrance emitted by the i - th butterfly, where the expression of f i is as follows:

[0078] fi = cI α

[0079] where f i represents the perceived intensity of the pheromone of the i-th butterfly, that is, the intensity of the fragrance that can be perceived by other butterfly individuals. I represents the fitness of the butterfly, that is, the stimulation intensity of the fragrance generated by the current butterfly. c represents the perception modality, which determines the convergence speed of the BOA algorithm. α represents the power exponent depending on the pheromone of the fragrance, and it jointly determines the movement behavior of the BOA algorithm with the perception modality.

[0080] S33. Introduce an opposition-based learning method based on the elite strategy during the iterative calculation process of the GWO algorithm; the opposition-based learning method based on the elite strategy is specifically as follows: By selecting elites, the opposition wolf pack is merged with the current wolf pack, and excellent individuals are selected to enter the next iteration cycle for searching. That is, to a certain extent, it maintains the diversity of the wolf pack, prevents the algorithm from falling into local optimum, and at the same time fully absorbs the beneficial information of the elite individuals in the current wolf pack, and accelerates the global convergence speed of the algorithm. Let x i (t) and be the current individual and the opposition individual respectively in the t-th iteration process. Let Fit i,j (t) and be the fitness values corresponding to x i (t) and on the j-th dimension. The present invention represents m (2 ≤ m ≤ N) elite individuals as shown below:

[0081]

[0082] where {e1(t), e2(t),..., e m (t)} represents the set composed of elite individuals. The opposition solution is expressed as the following formula:

[0083]

[0084]

[0085] where η is a random number and satisfies η ∈ (0, 1). The min function represents returning the minimum value of the input sequence. The max function represents returning the maximum value of the input sequence.

[0086] According to the above introduction, the calculation steps of the EWB - GWO algorithm specifically include:

[0087] (1) For the number of gray wolf individuals N and the number of algorithm iterations Max in the GWO algorithm IterInitialize the solution space dimension Dim, the conversion probability p between global search and local wandering, and the boundary conditions (ub, lb).

[0088] (2) Generate the initial gray wolf population through the Tent-Sine coupled chaotic map in the solution space. Calculate the fitness value of each individual according to the fitness function and initialize the Alpha wolf, Beta wolf, and Delta wolf.

[0089] (3) Calculate the perception intensity generated by the scent pheromone of the gray wolf individuals according to the scent formula, and determine the search behavior of the current algorithm by the size of the random number and the conversion probability.

[0090] (4) When rand < p, calculate the interval distances between the Alpha wolf, Beta wolf, and Delta wolf and the wolf pack respectively according to the gray wolf migration formula, and obtain the migration direction of the overall wolf pack by multiplying with the inertia weight; when rand ≥ p, the current algorithm executes the local wandering rule, where rand represents a random value between 0 and 1.

[0091] (5) Recalculate the Alpha wolf, Beta wolf, and Delta wolf based on the migrated wolf pack individuals.

[0092] (6) If the algorithm iteration times reach the termination condition, end the loop and output the optimal individual; otherwise, generate the next generation population through the opposition-based learning method based on the elite strategy and return to step (3).

[0093] S4. The specific steps of the initial population generation strategy of the line-of-sight detection method based on the Bresenham line algorithm are as follows:

[0094] S41. Discretize and model the working space of the mobile robot through equal-area grids. The grid volume is set to 1.2 to 1.5 times the volume of the mobile robot.

[0095] S42. Generate feasible path nodes through the Tent-Sine two-dimensional chaotic map, and detect the line-of-sight reachability between all generated nodes and their parent nodes and the target node through the line-of-sight algorithm.

[0096] S43. If the current node is simultaneously LOS reachable with its parent node and the target point, output this individual and generate the next gray wolf individual; if the current node is only LOS reachable with its parent node, retain the current node as the parent node and continue to generate the next node; if there is no reachability between the current node and its parent node, directly discard the current node, as Figure 3 shown.

[0097] S44. The algorithm generates the initial population iteratively until all the initial gray wolf individuals satisfy the feasible solution, and then the algorithm can end. As shown in Figure 4 shown below.

[0098] According to the analysis of the experimental results, it can be known that the initial population generation strategy based on the Bresenham line algorithm proposed by the present invention can further reduce the generation of path discontinuities while ensuring the diversity of the generated feasible solutions. Through the horizontal comparative test based on the public dataset, it can be known that the path generated by the EWB-GWO algorithm performs excellently in the tests of path length and path smoothness indicators.

Claims

1. A path planning method for an unmanned delivery robot based on the EWB-GWO algorithm, characterized in that, It includes the following steps: S1. Regarding the problem of generating the initial wolf pack based on the GWO algorithm, a two-dimensional Tent-Sine coupled chaotic map is introduced; the formula of the two-dimensional Tent-Sine coupled chaotic map is as follows: where x n and y n represent the sequence numbers currently generated by the coupled chaotic maps, x n+1 and y n+1 represent the next two sequence numbers generated by the chaotic maps, a represents the system parameter of the Tent chaotic map, λ represents the system weight value, and it satisfies λ ∈ (0, 1); S2. The GWO algorithm is improved by combining an adaptive non-linear inertia weight strategy, an opposition-based learning method based on an elite strategy, and a random walk rule integrating the butterfly optimization algorithm; S3. The improved GWO algorithm is combined with the path planning of the unmanned delivery robot, and the rational generation of the initial population of the GWO algorithm is strengthened through a line-of-sight detection technique based on the Bresenham line algorithm.

2. The path planning method of the unmanned delivery robot based on the EWB-GWO algorithm according to claim 1, characterized in that: In step S1, the specific expression formula of the Tent chaotic map is as follows: where x k represents the sequence number currently generated by the Tent chaotic map, and x k+1 represents the next sequence number generated by the Tent chaotic map, and a represents the system parameter of the Tent chaotic map; The specific expression formula of the Sine chaotic map is as follows: x z+1 = bsin(πx z ) b ∈ (0, 1] where x z represents the sequence number currently generated by the Sine chaotic map, and x z+1 represents the next sequence number generated by the Sine chaotic map, and b represents the system parameter of the Sine chaotic map.

3. The path planning method of the unmanned delivery robot based on the EWB-GWO algorithm according to claim 1, characterized in that: Step S2 specifically includes: S21. An adaptive non-linear inertia weight is introduced in the global search stage of the GWO algorithm; S22. A random walk rule based on the butterfly optimization algorithm is introduced during the iteration process of the GWO algorithm; S23. An opposition-based learning method based on an elite strategy is introduced during the iteration process of the GWO algorithm.

4. The path planning method for an unmanned delivery robot based on the EWB-GWO algorithm according to claim 3, characterized in that: The adaptive non-linear inertia weight is as follows: Among them, W GWO represents the inertia weight of the current grey wolf pack, Fit t,i represents the fitness value of the i-th grey wolf in the t-th iteration process, Fit t,g represents the fitness value of the optimal grey wolf in the t-th iteration process, Max Iter represents the maximum number of iterations of the GWO algorithm.

5. The path planning method for an unmanned delivery robot based on the EWB-GWO algorithm according to claim 3, characterized in that: The expression formula of the random walk rule based on the butterfly optimization algorithm is as follows: Among them represents the \(i\)-th butterfly individual in the \(t\)-th iteration process, \(r\) represents a random number satisfying \(r\in[0,1]\), and represent the \(z\)-th and \(k\)-th butterfly individuals in the solution space, \(f i represents the intensity of the pheromone of the fragrance emitted by the \(i\)-th butterfly, where \(f i The expression of is as follows: f i = cI α where f i represents the perceived intensity of the pheromone of the i-th butterfly, I represents the fitness of the butterfly, that is, the stimulation intensity of the fragrance produced by the current butterfly, c represents the perception modality, and α represents the power exponent dependent on the pheromone.

6. The path planning method for the unmanned delivery robot based on the EWB-GWO algorithm according to claim 3, wherein: The reverse learning method based on the elite strategy is specifically as follows: By selecting elites, the reverse wolf pack is merged with the current wolf pack, and excellent individuals are selected to enter the next iteration cycle for searching. Let x i (t) and be the current individual and the reverse individual respectively in the t-th iteration process. Fit i,j (t) and are the fitness values on the j-th dimension corresponding to x i (t) and . The m (2 ≤ m ≤ N) elite individuals are represented as follows: Among them, {e1(t), e2(t),..., e m (t)} represents the set composed of elite individuals, N represents the number of grey wolf individuals, and the reverse solution is expressed as the following formula: where η is a random number and satisfies η ∈ (0, 1), the min function represents returning the minimum value of the input sequence, and the max function represents returning the maximum value of the input sequence.

7. The path planning method for the unmanned delivery robot based on the EWB-GWO algorithm according to any one of claims 1-6, characterized in that: The calculation steps of the improved GWO algorithm in step S2 specifically include: (1) Initialize the number of grey wolf individuals N, the number of algorithm iterations Max Iter , the dimension of the solution space Dim, the conversion probability p between global search and local wandering, and the boundary conditions (ub, lb). (2) In each case, an initial generation of gray wolf populations is generated through a Tent-Sine coupled chaotic map in the solution space. The fitness value of each individual is calculated according to the fitness function, and the Alpha wolf, Beta wolf, and Delta wolf are initialized; (3) The perception intensity generated by the scent pheromone of the gray wolf individuals is calculated according to the scent formula, and the search behavior of the current algorithm is determined by the size of the random number and the transition probability; (4) When rand < p, the distance intervals between the Alpha wolf, Beta wolf, and Delta wolf and the wolf pack are calculated respectively according to the gray wolf migration formula, and the migration direction of the overall wolf pack is obtained by multiplying with the inertia weight; when rand ≥ p, the current algorithm executes the local walk rule, where rand represents a random value between 0 and 1; (5) Based on the migrated wolf pack individuals, the Alpha wolf, Beta wolf, and Delta wolf are recalculated; (6) If the number of algorithm iterations reaches the termination condition, the loop is ended and the optimal individual is output; otherwise, a new generation of populations is further generated through an opposition-based learning method based on an elite strategy, and it returns to step (3).

8. The path planning method of the unmanned delivery robot based on the EWB-GWO algorithm according to claim 1, characterized in that: S3 specifically includes: S31. The working space of the mobile robot is discretely segmented and modeled through equal-area grids, and the grid volume is set to 1.2 to 1.5 times the volume of the mobile robot; S32. Feasible path nodes are generated through a Tent-Sine two-dimensional chaotic map, and the line-of-sight reachability between all generated nodes and their parent nodes and the target node is detected through the line-of-sight algorithm. S33. If the current node is LOS reachable from both its parent node and the target point simultaneously, output this individual and generate the next gray wolf individual; if the current node is only LOS reachable from its parent node, retain the current node as the parent node and continue to generate the next node; if there is no reachability between the current node and its parent node, directly discard the current node. S34. Generate the initial population in an iterative manner until all the initial gray wolf individuals satisfy the feasible solution, and then the algorithm ends.

9. A computer-readable storage medium, characterized in that: A computer program is stored on the computer-readable storage medium. When the computer program is executed by a processor, it implements the unmanned delivery robot path planning method based on the EWB-GWO algorithm according to any one of claims 1 to 8.

Images