Multi-algorithm fusion robot motion path planning method and robot

Abstract

The invention relates to the technical field of robots, in particular to a multi-algorithm fusion robot motion path planning method and a robot. According to the method, firstly, a motion scene of a robot is modeled, then, a JPS algorithm and a GWO grey wolf optimization algorithm are improved, the JPS algorithm updates a jump point search measurement mode, and the GWO grey wolf optimization algorithm introduces a route random disturbance mechanism and reconstructs a multi-target fitness function. And finally, performing multi-algorithm fusion robot motion path optimization by using an RRT * algorithm, an improved JPS algorithm and a GWO grey wolf optimization algorithm, specifically, generating a random path by using the existing RRT * algorithm, performing first-stage optimization by using the improved GWO grey wolf optimization algorithm, performing path simplification, performing second-stage optimization by using the improved JPS algorithm, and performing multi-algorithm fusion robot motion path optimization by using the improved GWO grey wolf optimization algorithm. And finally, third-stage optimization is carried out through an improved GWO grey wolf optimization algorithm. According to the invention, the problems of low path planning efficiency and poor planning result quality in a complex scene in the existing scheme can be solved.

Description

Technical Field

[0001] The present invention relates to the technical field of robots, and particularly to a method for planning a motion path of a robot with multi-algorithm fusion and a robot. Background Art

[0002] With the rapid development of intelligent manufacturing and robot technology, intelligent mobile robots are increasingly widely used in fields such as industrial automation, warehousing logistics, medical services, and autonomous driving. And path planning, as one of the core technologies of intelligent mobile robots, directly affects the navigation efficiency, obstacle avoidance ability, and task execution effect of intelligent mobile robots. Path planning refers to planning an optimal or a path that meets specific requirements through certain intelligent methods in a given environment according to certain strategies and conditions.

[0003] Currently, commonly used algorithms in the field of path planning include the A* algorithm, Dijkstra algorithm, RRT (Rapidly-Exploring Random Tree) algorithm, etc. These algorithms have their own advantages. The A* algorithm is suitable for static environments and has high heuristic search efficiency; the Dijkstra algorithm is simple and easy to implement; the RRT algorithm can efficiently explore high-dimensional spaces and has strong scalability.

[0004] However, in the face of complex environments or dynamic changes, traditional single path planning algorithms may have problems such as being unable to find the global optimal solution, poor path quality, and low computational efficiency. This can easily result in low efficiency when an intelligent mobile robot performs path planning in a working area, or even planning an incorrect path, thus being unable to work safely and properly. Summary of the Invention

[0005] In order to solve the problem that existing path planning algorithms are difficult to efficiently generate a path that meets requirements in a complex environment, the present invention provides a method for planning a motion path of a robot with multi-algorithm fusion, and its corresponding computer program product and robot.

[0006] The technical solution provided by the present invention is as follows:

[0007] A method for planning a motion path of a robot with multi-algorithm fusion, which includes:

[0008] Obtain the position information and surrounding environment information of the robot in a specified working area; construct a grid map based on the obtained information and mark the obstacles in the robot and its surrounding environment, thereby obtaining scene data.

[0009] Design a fitness function of the GWO gray wolf optimization algorithm with path length, safety, and smoothness as evaluation indicators; encode any possible path on the grid map as a gray wolf, and introduce a perturbation mechanism in the GWO gray wolf optimization algorithm to achieve population update, thereby obtaining an improved GWO gray wolf optimization algorithm.

[0010] (i) Use the RRT* algorithm to perform global path planning based on the known scenario data, generate a random path L0 from the starting point to the target point and passing through the corresponding nodes of multiple grids on the grid map, and smooth it. (ii) Use the improved GWO grey wolf optimization algorithm to iteratively optimize L0 with L0 as the initial path to obtain a new path L1. (iii) Simplify L1, delete redundant intermediate points to obtain a simplified path L2, and further optimize L2 using the improved JPS jump point search algorithm to obtain a new path L3. (iv) Use the improved GWO grey wolf optimization algorithm to iteratively optimize the initial path again with L3 as the initial path to obtain the final optimized path L4.

[0011] As a further improvement of the present invention, the iterative optimization process of the improved GWO grey wolf optimization algorithm includes:

[0012] S1: Preset algorithm parameters, including: the number of grey wolves N wolves 、the maximum number of iterations N max2 、the path perturbation intensity σ. Initialize the grey wolf population. Each grey wolf represents a possible path. The Alpha wolf represents the current optimal path, the Beta wolf represents the sub-optimal path, and the Delta wolf represents the third-optimal path. Set the initial path L0: L0 = (L 01 , L 02 ,... L 0i ,... L nn ), where L 0i represents the i-th path node in the initial path L0.

[0013] S2: Use the perturbation mechanism to randomly adjust the current path to generate several new paths; check whether the new paths are valid. If so, add them to the wolf pack, otherwise regenerate the new paths.

[0014] S3: In each iteration, calculate the fitness value f(L k ) of each grey wolf L k through the fitness function; then update the leading wolf pack according to the fitness ranking, and use the three paths with relatively smaller fitness values as the Alpha wolf, the Beta wolf, and the Delta wolf respectively.

[0015] S4: Convert the paths into arrays, and then calculate the distances D α , D β , D δ between each grey wolf in the population and the Alpha wolf, the Beta wolf, and the Delta wolf, and update its position.

[0016] S5: Perform boundary clipping on the new positions of each updated grey wolf to ensure that each path it represents is within the passable range of the grid map.

[0017] S6: Repeat steps S2 - S5 until the maximum number of iterations N max2 or the path fitness value converges.

[0018] S7: Select the path represented by the Alpha wolf in the final optimization result, and output it after smoothing.

[0019] As a further improvement of the present invention, in step S2, the method for generating a new path using the perturbation mechanism is as follows:

[0020] Apply a random offset that follows a normal distribution N(0,σ) to the coordinates of each node in the original path for random perturbation.

[0021] Verify whether the coordinates of the new nodes after perturbation are located in the passable area through the grid map. If the verification fails, repeat the perturbation operation for the corresponding nodes until the maximum number of attempts is reached; if no valid new coordinates are found after reaching the maximum number of attempts, retain the original nodes.

[0022] Retain all the nodes after perturbation to form the required new path.

[0023] As a further improvement of the present invention, in step S2, the path existence check, path node boundary check, and path node obstacle check methods are simultaneously used to evaluate whether the newly generated path after perturbation is valid.

[0024] As a further improvement of the present invention, the expression of the fitness function in step S3 is as follows:

[0025]

[0026] where, length(L k ) represents the length of the path corresponding to the gray wolf L k , penalty(L k ) represents the safety penalty term of the path corresponding to the gray wolf L k , smoothness(L k ) represents the smoothness penalty term of the path corresponding to the gray wolf L k ; (x i , y i ) represents each point on the path, i = 1…n, n represents the number of points on the path; Grid[,] represents the function for discriminating obstacles, (dx,dy) represents the four-neighborhood offset, θ i represents the direction angle of the path segment P i-1 to P i ; |θ i - θ i-1 | represents the direction change amount of adjacent path segments.

[0027] As a further improvement of the present invention, in step S4, the position X of any gray wolf k is updated according to the following formula:

[0028]

[0029] In the above formula, X α , X β , X δ respectively represent the positions of the Alpha wolf, Beta wolf, and Delta wolf; X k represents the original position of the k-th gray wolf, and X k ' represents the updated position of the k-th gray wolf; A1, A2, A3, C1, C2, and C3 are random coefficients.

[0030] Among them, the two types of random coefficients A and C are randomly generated using the following formula:

[0031]

[0032] Among them, a represents a parameter that linearly decreases with the number of iterations, and r1, r2 represent random numbers between [0, 1].

[0033] As a further improvement of the present invention, the process of generating a random path based on the RRT* algorithm includes:

[0034] S01: Initialize the algorithm parameters, including: the maximum number of iterations N max1 , the step size λ, and the neighborhood radius R for optimizing the random tree; preset the structure of the expanding random tree, and use the starting position of the robot as the root node of the expanding random tree.

[0035] S02: Randomly determine a random sampling point q rand that can avoid obstacles within the search space, as the candidate direction for expanding the random tree.

[0036] S03: Traverse the set of child nodes in the random tree to find the node q rand nearest to the random sampling point q near , and start from the child node q near and expand by one step size λ along the direction of q near and q rand to obtain a new node q new .

[0037] S04: Check whether the path from q near to q new collides with obstacles. If the path does not collide, add the new node q new to the random tree and record its parent node as q nearOtherwise, the node is abandoned and the process returns to step S03 to determine a new q near .

[0038] S05: In q new Find all possible parent node candidate sets Q within the neighborhood radius R of near , calculated from Q near Each candidate node to q new The total path cost c(q newi ). Then check the near Each candidate node to q new Whether the path collides with obstacles, delete the candidate nodes that may collide. Finally, select the total path cost c(q newi )The smallest node is used as the new parent node q new , update the random tree structure.

[0039] S06: Optimize the rewiring of nodes within the neighborhood radius R and check whether it is possible to rewire the nodes by new As a parent node to reduce Q near The path cost of other nodes in the , if possible, update its parent node to q new , and recalculate the path cost, otherwise skip the corresponding node.

[0040] S07: Repeat S02-S06 until the maximum number of iterations N is reached max1 Or find a path from the starting point q S To the target point q G feasible path.

[0041] S08: Starting from the end point of the random tree obtained after the iteration ends, trace back to the starting point q along each parent node S , generate the corresponding random paths and smooth them.

[0042] As a further improvement of the present invention, in step S05, the parent node candidate set Q near Each candidate node to q new The total path cost c(q newi ) is calculated as follows:

[0043] c(q newi )=c(q i )+distance(q i ,q new ),

[0044] In the above formula, c(q i ) represents any candidate node q i The current path cost, distance(q i ,q nnw) represents any candidate node q i to q new the Euclidean distance between them.

[0045] As a further improvement of the present invention, the process of implementing path optimization by using the improved JPS jump point search algorithm includes:

[0046] S001: Initialize the algorithm parameters, including: the starting point q S , the target point q G and the obstacle list S p .

[0047] Generate an open list openlist and a closed list closelist, and add the starting point q S to openlist, and set its path cost c(q S ) = 0.

[0048] S002: Select the node with the minimum path cost from the open list openlist as the current node q current , and move it from the open list openlist to the closed list closelist.

[0049] S003: Determine whether the current node q current is the target point q G : If so, stop the search and end the iterative process; otherwise, perform jump point search to obtain all valid successor nodes of the current node, and form a successor node set.

[0050] S004: Calculate the temporary path cost c S from the starting point q i to all successor nodes q yenyative in the successor node set in turn according to the following formula: i ):

[0051] c tentative (q i ) = c(q current ) + cost(q current , q i ),

[0052] In the above formula, c(q current ) represents the actual path cost from the starting point q S to the current node q current , and cost(q current , q i ) represents the path cost from the current node q current to the successor node q i .

[0053] S005: Determine the successor node qi Whether it has been visited or its temporary path cost c tentat1ve (q i ) has a smaller path cost than the previous one: If so, update the path information, record the predecessor node of node q i as q curremt , and record the actual path cost from the starting point q s to node q i as c tentative (q i ); Otherwise, return to step S004 to recalculate the temporary path cost of the next successor node.

[0054] S006: Calculate the total path cost c(q i ) of each updated node q i , and move it to the open list openlist when the corresponding node q i is not in the closed list closelist; among them, the calculation formula of c(q i ) is:

[0055] c(q i ) = c tentative (q i ) + h(q i ),

[0056] In the above formula, h(q i ) represents the heuristic cost from node q i to the target point q G .

[0057] S007: Repeat steps S002 - S006 until the target point q G is found or the open list openlist is empty, and the iterative process is terminated.

[0058] S008: Starting from the end point obtained after the iteration termination, trace back along each predecessor node to the starting point q S to generate a path, and after smoothing the generated path, use it as the new path optimized by the JPS jump point search algorithm.

[0059] As a further improvement of the present invention: The method of generating a successor node set by jump point search in step S003 is:

[0060] First, generate several successor nodes corresponding to the current node through the target point priority detection method, the forced neighbor detection method, and the diagonal jump detection method respectively; then perform boundary check and obstacle detection on each successor node, and exclude the nodes belonging to the grid boundary or the nodes with the connection direction passing through obstacles to obtain the final successor node set.

[0061] The present invention further includes a robot, which includes a path planning device. The path planning device includes a memory, a processor, and a computer program stored on the memory and running on the processor. When the computer program is executed by the processor, the steps of the method for planning the motion path of a robot with multi-algorithm fusion as described above are implemented, and then the optimal path of the robot from the starting point to the target point is planned according to the known scene data.

[0062] The present invention has the following beneficial effects:

[0063] The present invention deeply integrates the RRT* algorithm, the GWO grey wolf optimization algorithm, and the JPS jump point search algorithm. Combining the advantages of global sampling of the RRT* algorithm, global search and intelligent optimization of the GWO algorithm, and efficient local search of the JPS algorithm, redundant nodes are skipped, and the number of search nodes is significantly reduced. Furthermore, it can quickly generate a globally optimal path in a complex environment, meeting the real-time requirements of the path planning result of the robot in practical applications. In addition, the optimized path generated by combining multiple algorithms in the present invention is more in line with the reality, and can avoid the problem of insufficient path quality of a single algorithm. Furthermore, it lays a foundation for the mobile robot to accurately and more efficiently pass during work operation.

[0064] The present invention introduces a path perturbation mechanism and effectiveness verification in the improved GWO grey wolf optimization algorithm to increase the diversity of search, enhance the global search ability of the GWO grey wolf optimization algorithm, and avoid falling into local optimal solutions; adopts a multi-objective fitness function, comprehensively considering factors such as path length, obstacle proximity penalty, and path smoothness, for multi-objective optimization; the improved GWO algorithm can still generate effective paths even in a complex environment, ensuring that the generated paths have high robustness in practical applications and are more in line with the actual application requirements.

[0065] The present invention significantly improves the safety and completeness of path planning by recursively detecting horizontal / vertical direction jumps and dynamically enforcing neighbor rules in the improved JPS jump point search algorithm, overcoming the defects of existing algorithms. On the one hand, the optimization result of the improved algorithm can ensure that adjacent obstacles can be detected in advance during diagonal movement, avoiding the robot path getting close to dangerous areas; on the other hand, by dynamically adapting directions and expanding the detection range, while reducing redundant calculations, key turning points can be quickly identified, and finally an optimized path that is shorter, smoother, and meets the safety distance constraint is generated, which is particularly suitable for robot path planning in a high-density obstacle environment. Brief Description of the Drawings

[0066] Figure 1 It is a flowchart of the steps of the method for planning the motion path of a robot with multi-algorithm fusion provided in Embodiment 1 of the present invention.

[0067] Figure 2It is the flowchart of the steps for the improved JPS algorithm to generate a successor node set through jump point search in Embodiment 1 of the present invention.

[0068] Figure 3 It is the flowchart of the steps for generating a random path using the RRT* algorithm in Embodiment 1 of the present invention.

[0069] Figure 4 It is the flowchart of the steps for implementing path optimization using the improved GWO gray wolf optimization algorithm in Embodiment 1 of the present invention.

[0070] Figure 5 It is the flowchart of the steps for generating a new path from the original path using the perturbation mechanism in Embodiment 1 of the present invention.

[0071] Figure 6 It is the flowchart of the steps for path optimization based on the improved JPS jump point search algorithm in Embodiment 1 of the present invention.

[0072] Figure 7 It is the scenario map designed in the simulation experiment.

[0073] Figure 8 It is the different route maps from the same starting point to the end point planned in the scenario by different schemes in the simulation experiment. Detailed implementation manners

[0074] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0075] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those of ordinary skill in the technical field to which the present invention belongs. The terms used herein in the specification of the present invention are only for the purpose of describing specific embodiments, and are not intended to limit the present invention. The term "or / and" used herein includes any and all combinations of one or more of the related listed items.

[0076] Embodiment 1

[0077] This embodiment proposes a method for planning the motion path of a robot by fusing multiple algorithms. This method simultaneously uses the RRT* algorithm, the improved JPS algorithm, and the improved GWO algorithm to plan the optimal path of the robot from the starting point to the target point according to the known scenario data. Specifically, as Figure 1 shown, the method provided in this embodiment includes the following steps:

[0078] I. Scenario modeling:

[0079] The robot in this embodiment mainly includes intelligent mobile robots widely used in fields such as industrial automation, warehousing logistics, medical services, and autonomous driving. Such robots are usually used within a designated area, which is the designated working area. In practical applications, the designated working area can be a rectangular area or a polygonal area. This embodiment will pre-acquire the map information corresponding to the designated working area.

[0080] The robot in this embodiment usually also installs a navigation module based on Beidou or other positioning systems, as well as sensors such as lidar and vision cameras. Among them, the navigation module can obtain the position information of the robot itself in real time, and sensors such as lidar and vision cameras can be used to collect the environmental information around the robot, thereby assisting in identifying obstacles around the robot.

[0081] Based on the above content, this embodiment can first rasterize the geographical image of the designated working area, divide the area into arrayed grids, and then obtain a blank grid map. Among them, the blank grid map is also marked with passable areas, road information, and various obstacle information such as buildings and fixed equipment. Then, the navigation module and various sensors installed on the robot are used to obtain the position information and surrounding environmental information of the robot in the designated working area, and then mark the robot and the obstacles in its surrounding environment on the grid map. The grid map containing geographical information and the target object information therein is used as the scene data required for this embodiment.

[0082] II. Algorithm improvement

[0083] The technical solution provided in this embodiment adopts a multi-algorithm fusion solution including the RRT* algorithm, the JPS algorithm, and the GWO grey wolf optimization algorithm to realize the motion path planning of the robot. In order to improve the effectiveness of the algorithm in the scenario of the present invention, the technical personnel improve the JPS algorithm and the GWO algorithm.

[0084] 2.1. Improve the GWO grey wolf optimization algorithm

[0085] For the GWO grey wolf optimization algorithm, on the one hand, this embodiment introduces a path perturbation mechanism and effectiveness verification to increase the diversity of search, enhance the global search ability of the GWO grey wolf optimization algorithm, and avoid falling into local optimal solutions. On the other hand, factors such as comprehensive path length, obstacle proximity penalty, and path smoothness are used as optimization objectives; the multi-objective fitness function adopted by the algorithm is reconstructed.

[0086] Specifically, in the improved GWO gray wolf optimization algorithm, each updated path is regarded as a gray wolf, and the perturbation mechanism is used to perturb the path nodes included in the three gray wolves retained in the previous round in each iteration to generate several new paths as new gray wolves, thereby forming the wolf pack required for this round of iteration.

[0087] In this embodiment, the expression of the multi-objective fitness function adopted by the improved GWO gray wolf optimization algorithm is as follows:

[0088] f(L k ) = length(L k ) + penalty(L k ) + smoothness(L k ),

[0089] In the above formula, f(L k ) represents the fitness value of the path corresponding to gray wolf L k ; length(L k ) represents the length of the path corresponding to gray wolf L k , penalty(L k ) represents the safety penalty term of the path corresponding to gray wolf L k , and smoothness(L k ) represents the smoothness penalty term of the path corresponding to gray wolf L k .

[0090] Among them, the calculation formula for the length length(L k ) of the path is:

[0091]

[0092] In the above formula, (x i , y i ) represents each point on the path, i = 1...n, and n represents the number of points on the path.

[0093] The expression of the safety penalty term penalty(L k ) is:

[0094]

[0095] In the above formula, (x, y) represents the coordinates of the path point; Grid[, ] represents the function for discriminating obstacles, Grid[i, j] = 1 indicates that the grid map is an obstacle at (i, j), and Grid[i, j] = 0 indicates that the grid map is free space at (i, j); (dx, dy) represents the four-neighborhood offset, that is: {(-1, 0), (1, 0), (0, -1), (0, 1)}.

[0096] Smoothness penalty term smoothness(L k ) is expressed as:

[0097]

[0098] In the above formula, θ i represents the direction angle from path segment P i-1 to P i ; |θ i - θ i-1 | represents the direction change amount of adjacent path segments.

[0099] 2.2. Improved JPS algorithm

[0100] The traditional JPS algorithm has problems such as imperfect diagonal movement detection and limitations of the forced neighbor detection rule. In view of these performance defects, the improved JPS algorithm provided in this embodiment significantly improves the safety and completeness of path planning through recursive horizontal / vertical direction jump detection and dynamic forced neighbor rules. On the one hand, it ensures that adjacent obstacles can be detected in advance during diagonal movement to avoid the robot path approaching dangerous areas; on the other hand, by dynamically adapting the direction and expanding the detection range, while reducing redundant calculations, it quickly identifies key turning points, and finally generates an optimized path that is shorter, smoother and meets the safety distance constraint, which is especially suitable for robot path planning in high-density obstacle environments.

[0101] Specifically, as Figure 2 shown, the way for the improved JPS algorithm provided in this embodiment to generate a successor node set through jump point search is:

[0102] First, several successor nodes corresponding to the current node are generated respectively through the target point priority detection method, the forced neighbor detection method and the diagonal jump detection method to form an initial successor point set, and then each successor node in the initial successor point set is subjected to validity review by means of boundary check and obstacle detection, excluding the nodes belonging to the grid boundary or the nodes whose connection direction passes through obstacles. Retain each remaining successor node in the initial successor node set to obtain the final successor node set.

[0103] III. Using a multi-fusion algorithm to generate an optimal path from the starting point to the target point according to the scene information

[0104] In this embodiment, the process of using the RRT* algorithm, the improved JPS algorithm and the improved GWO grey wolf optimization algorithm to generate an optimal path according to the scene information and the starting point and the end point includes the following steps:

[0105] 3.1. Use the RRT* algorithm to perform global path planning based on the known scenario data, generate a random path L0 from the starting point to the target point and passing through the corresponding nodes of multiple grids on the grid map, and smooth it.

[0106] Among them, as Figure 3 shown, the process of the RRT* algorithm generating a random path includes:

[0107] (I) Algorithm initialization

[0108] S01: First, initialize the algorithm parameters, including: the maximum number of iterations N max1 , step size λ, and neighborhood radius R for optimizing the random tree. Secondly, preset the structure of the expanding random tree, and take the starting position q s of the robot as the root node of the expanding random tree.

[0109] (II) Iterative optimization process of the algorithm

[0110] S02: In each iteration process, first randomly determine a random sampling point q rand that can avoid obstacles in the search space as the candidate direction for the expansion of the random tree.

[0111] S03: Traverse the set of child nodes in the random tree to find the node q rand closest to the random sampling point q near . Starting from the child node q near , expand by a step size λ along the direction of q near and q rand to obtain a new node q new .

[0112] S04: Check whether the path from q near to q new collides with obstacles. If the path does not collide, add the new node q new to the random tree, and record its parent node as q near , otherwise abandon the node and return to step S03 to re-determine the new q near .

[0113] S05: Within the neighborhood radius R of q new , find all possible sets of candidate parent nodes Q near , calculate the total path cost c(q near ) from each candidate node in Q new to q newi . Then check the path from each candidate node in Q near to q newWhether the path collides with obstacles, and delete the candidate nodes that may collide. Finally, select the node with the minimum total path cost c(q newi ) as the new parent node q new , and update the random tree structure.

[0114] Among them, for each candidate node in the candidate set Q near of the parent node to q new , the total path cost c(q newi ) is calculated by the formula:

[0115] c(q newi ) = c(q i ) + distance(q i , q new ),

[0116] In the above formula, c(q i ) represents the current path cost of any candidate node q i , and distance(q i , q new ) represents the Euclidean distance between any candidate node q i and q new .

[0117] S06: Perform rewiring optimization on the nodes within the neighborhood radius R, and check whether the path cost of other nodes in Q new can be reduced by taking q near as the parent node. If so, update its parent node to q new , and recalculate the path cost. Otherwise, skip the corresponding node.

[0118] S07: Repeat S02 - S06 until the maximum number of iterations N max1 is reached or a feasible path from the starting point q S to the target point q G is found.

[0119] (III) Algorithm Output

[0120] S08: Starting from the end point of the random tree obtained after the iteration termination, trace back along each parent node to the starting point q S , generate the corresponding random path and smooth it.

[0121] 3.2. Using L0 as the initial path, use the improved GWO gray wolf optimization algorithm to iteratively optimize L0 to obtain a new path L1.

[0122] Specifically, combined with the improvement of the GWO gray wolf optimization algorithm in this embodiment, such as Figure 4As shown in the figure, the process of using the improved GWO gray wolf optimization algorithm to achieve path optimization in this embodiment includes:

[0123] (1) Algorithm initialization

[0124] S1: Preset algorithm parameters, including: the number of gray wolves N wolves , the maximum number of iterations N max2 and the path perturbation intensity σ.

[0125] Set the initial path L0: L0 = (L 01 , L 02 ,... L 0i ,... L 0n ), where L 0i represents the i-th path node in the initial path L0. Encode a gray wolf of the initial path, and then initialize the gray wolf population. Each gray wolf represents a possible path, and the number of gray wolves in the initialized gray wolf population is N wolves .

[0126] At the same time, set three tags Alpha, Beta, and Delta to mark the leading wolf pack after each round of iterative optimization. Among them, the Alpha wolf represents the current optimal path, the Beta wolf represents the sub-optimal path, and the Delta wolf represents the third-best path.

[0127] (2) The iterative optimization process of the algorithm

[0128] S2: Use the perturbation mechanism to randomly adjust the current path to generate several new paths; check whether the new paths are valid. If so, add them to the wolf pack, otherwise regenerate new paths. Increase the diversity of the search space through path perturbation to avoid falling into local optimal solutions.

[0129] In this embodiment, path perturbation is performed by randomly adjusting each node in the current path, and at the same time, ensuring that the perturbed nodes are still located in the obstacle-free grids to generate new paths. For each newly generated path, it is also necessary to check whether the path is valid. If the new path is valid, it is encoded and added to the wolf pack, representing other gray wolf individuals. If the new path is invalid, new paths are regenerated.

[0130] Specifically, as Figure 5 shown, the method for using the perturbation mechanism to generate new paths based on the original path in this embodiment is:

[0131] (1) For a path containing multiple nodes, sequentially apply a random offset obeying the normal distribution N(0, σ) to the coordinates of each node in the original path for random perturbation to obtain the updated node coordinates. Among them, σ corresponds to the path perturbation intensity preset in the initialization stage.

[0132] (2) Combine the marked obstacle information and road information in the grid map, etc., and verify whether each new node coordinate after perturbation is located in the passable area. If so, retain the perturbed node coordinate; otherwise, apply a new random perturbation until the number of repeated perturbation operations for the corresponding node reaches the maximum number of attempts. If the new coordinate after reaching the maximum number of attempts is still in the passable area, retain the original node coordinate before perturbation.

[0133] (3) Repeat the above steps to perturb and update each node in the path; finally, retain all the nodes after perturbation to form the required new path.

[0134] In addition, for each generated new path, in this embodiment, the path existence check, path node boundary check, and path node obstacle check methods are simultaneously used to evaluate whether the new path generated after perturbation is valid. If the generated new path fails to pass any one of the path existence check, path node boundary check, and path node obstacle check, it means that the path is invalid and should be excluded.

[0135] S3: In each iteration, calculate the fitness value f(L k ) of each gray wolf L in the population after perturbation update through the following fitness function: k )

[0136] f(L k ) = length(L k ) + penalty(L k ) + smoothness(L k ),

[0137] Then sort the gray wolves in the population in ascending order of fitness, update the leading wolf pack, and use the three paths with relatively smaller fitness values as the Alpha wolf, Beta wolf, and Delta wolf respectively.

[0138] S4: Convert the path into an array, and then calculate the distances D α , D β , D δ between each gray wolf in the population and the Alpha wolf, Beta wolf, and Delta wolf for each path array, and update its position. Specifically, the update formula for the position X k of any gray wolf is as follows:

[0139]

[0140] In the above formula, X α , X β , X δ represent the positions of the Alpha wolf, Beta wolf, and Delta wolf respectively; Xk represents the original position of the k-th gray wolf, X k ′ represents the updated position of the k-th gray wolf; A1, A2, A3, C1, C2, C3 are random coefficients.

[0141] In this embodiment, two types of random coefficients A and C can be generated by the following random number generator:

[0142]

[0143] where a represents a parameter that linearly decreases with the number of iterations, and r1, r2 represent random numbers between [0, 1].

[0144] S5: Also perform boundary clipping on the new positions of each updated gray wolf to ensure that each path represented by it is within the passable range of the grid map.

[0145] In the above steps, the positions of each gray wolf in the population are automatically updated by combining the positions of the leading wolf pack and the preset random numbers. The updated positions may have problems not meeting the regulations. Therefore, this embodiment also performs boundary clipping on the new positions of each updated gray wolf to ensure that each path represented by it is within the passable range of the grid map.

[0146] S6: Repeat steps S2 - S5 until the maximum number of iterations N max2 or the path fitness value converges, and terminate the iterative process.

[0147] (III) Algorithm Output

[0148] S7: After the iterative optimization process ends, output the Alpha wolf in the reserved leading population as the optimization result of this stage, and perform smoothing processing on the path represented by it and then output.

[0149] 3.3. Simplify L1, delete redundant intermediate points to obtain the simplified path L2, and further optimize L2 using the improved JPS jump point search algorithm to obtain the new path L3.

[0150] In practical applications, there may still be a large number of unnecessary bends in the paths generated by the aforementioned RRT* algorithm and optimized by the improved GWO gray wolf optimization algorithm, resulting in too long routes. To address this problem, this embodiment first simplifies L1 and deletes redundant intermediate points. Specifically, the method for this embodiment to identify redundant intermediate points in path L1 is:

[0151] Denote the vector from the predecessor node to the intermediate node in the path as the in-vector of the intermediate node, and the vector from the intermediate node to the successor node in the path as the out-vector of the intermediate node. Calculate the offset angle between the in-vector and the out-vector of any intermediate node, and compare it with a preset offset angle threshold. If the offset angle of any intermediate node is less than the offset angle threshold, then identify it as a redundant intermediate node. Among them, too small an offset angle of any intermediate node indicates an unnecessary bend, increasing the path length, and it should be deleted.

[0152] In this embodiment, all the redundant intermediate nodes identified in the optimized route L1 are deleted, and the predecessor node and the successor node of the redundant intermediate node are directly connected, thereby obtaining the required simplified path L2.

[0153] For the simplified path L2, this embodiment further uses an improved JPS jump point search algorithm to perform a second-round optimization on it. Specifically, as Figure 6 shown, the path optimization process based on the improved JPS jump point search algorithm includes the following steps:

[0154] (1) Algorithm initialization

[0155] S001: Initialize the algorithm parameters, including: the starting point q S , the target point q G and the obstacle list S p .

[0156] Generate an open list openlist and a closed list closelist, and add the starting point q S to openlist, and set its path cost c(q S ) = 0.

[0157] (2) The iterative optimization process of the algorithm

[0158] S002: Select the node with the minimum path cost from the open list openlist as the current node q current , and move it from the open list openlist to the closed list closelist.

[0159] S003: Determine whether the current node q current is the target point q G : If so, stop the search and end the iterative process. Otherwise, perform a jump point search to obtain all valid successor nodes of the current node, and form a successor node set.

[0160] Among them, the method of generating the successor node set through the jump point search is:

[0161] First, generate several successor nodes corresponding to the current node through the target point priority detection method, the forced neighbor detection method, and the diagonal jump detection method respectively; then perform boundary checking and obstacle detection on each successor node, and exclude the nodes that belong to the grid boundary or whose connection direction passes through obstacles to obtain the final set of successor nodes.

[0162] In the solution of this embodiment,

[0163] Improve and enhance the traditional forced neighbor detection method. When performing straight-line motion detection, calculate its vertical direction in real time according to the current moving direction to avoid hard-coded directions, so as to achieve dynamic direction detection of straight-line motion. This adjustment can adapt to any straight-line direction without the need for separate coding for each direction, and at the same time ensure that the path does not stick closely to obstacles. When performing diagonal motion detection, not only detect the adjacent nodes on the main diagonal, but also additionally detect the horizontal and vertical adjacent nodes in the main diagonal direction, which can avoid missed detection of forced neighbors (such as corner obstacles), and at the same time the generated path is farther from obstacles, improving safety;

[0164] When performing diagonal jump detection, recursively check the horizontal sub-direction and the vertical sub-direction to ensure that possible jump points are not missed, and at the same time reduce unnecessary jump times through early termination of recursion to improve efficiency.

[0165] S004: Calculate the temporary path cost c S from the starting point q i to the successor node q tentative (q i ) in the set of successor nodes in turn as follows:

[0166] c tentative (q i ) = c(q current ) + cost(q current , q i ),

[0167] In the above formula, c(q current ) represents the actual path cost from the starting point q S to the current node q current , and cost(q current , q i ) represents the path cost from the current node q current to the successor node q i .

[0168] S005: Determine whether the successor node q i has been visited or its temporary path cost c tentative (q i ) is smaller than the previous path cost: if so, update the path information and record the node qi The predecessor node of current is q, and the starting point q S to node q i has an actual path cost of c tentative (q i ); otherwise, return to step S004 to recalculate the temporary path cost of the next successor node.

[0169] S006: Calculate the total path cost c(q i ) for each updated node q i , and move it to the open list openlist when the corresponding node q i is not in the closed list closelist; among them, the calculation formula of c(q i ) is:

[0170] c(q i ) = c tentative (q i ) + h(q i ),

[0171] In the above formula, h(q i ) represents the heuristic cost from node q i to the target point q G .

[0172] S007: Repeat steps S002 - S006 until the target point q G is found or the open list openlist is empty, and terminate the iteration process.

[0173] (III) Algorithm Output

[0174] S008: Starting from the end point obtained after the iteration termination, trace back along each predecessor node to the starting point q S to generate a path, and after smoothing the generated path, use it as the new path optimized by the JPS jump point search algorithm.

[0175] 3.4. Take L3 as the initial path, and use the improved GWO grey wolf optimization algorithm to iteratively optimize the initial path again to obtain the optimal path L4.

[0176] The iterative optimization process of the improved GWO grey wolf optimization algorithm is the same as above, and will not be elaborated here.

[0177] In summary, the robot motion path planning method of this embodiment includes four stages: In the first stage, a random path is first generated by the RRT* algorithm. In the second stage, the improved GWO gray wolf optimization algorithm is used to iteratively optimize the random path. In the third stage, redundant nodes in the path optimized in the previous stage are first deleted according to preset rules, and then the improved JPS algorithm is used to perform a second round of optimization on it. In the fourth stage, the improved GWO gray wolf optimization algorithm is used again to perform a third round of optimization on the path after the second round of optimization, so as to obtain the optimal path for the robot's work navigation.

[0178] This path optimization scheme that deeply integrates the RRT* algorithm, the GWO gray wolf optimization algorithm, and the JPS jump point search algorithm provided in this embodiment can combine the advantages of global sampling of the RRT* algorithm, global search and intelligent optimization of the GWO algorithm, and efficient local search of the JPS, skip redundant nodes, significantly reduce the number of search nodes, be able to quickly generate a globally optimal path in a complex environment, improve the efficiency of generating the optimal path, and avoid the problem of insufficient quality in the path planning of a single algorithm.

[0179] Embodiment 2

[0180] On the basis of the solution of Embodiment 1, this embodiment further provides a variety of products for implementing the corresponding methods. These include: computer program products, storage media, path planning devices, and robots that can achieve automatic path planning.

[0181] Among them, the computer program product includes a computer program. When the computer program is executed by a processor, it implements the steps of the multi-algorithm fusion robot motion path planning method as in Embodiment 1, and then plans the optimal path of the robot from the starting point to the target point according to the known scene data.

[0182] The storage medium stores a computer program. When the computer program is executed by a processor, it implements the steps of the multi-algorithm fusion robot motion path planning method as in Embodiment 1, and then plans the optimal path of the robot from the starting point to the target point according to the known scene data.

[0183] The path planning device includes a memory, a processor, and a computer program stored on the memory and running on the processor. When the computer program is executed by the processor, it implements the steps of the multi-algorithm fusion robot motion path planning method as in Embodiment 1, and then plans the optimal path of the robot from the starting point to the target point according to the known scene data.

[0184] The path automatic planning robot includes a robot body and the aforementioned path planning device. The path planning device includes a memory, a processor, and a computer program stored on the memory and running on the processor. When the computer program is executed by the processor, it realizes the steps of the method for planning the motion path of the robot with multi-algorithm fusion as in Embodiment 1, and further plans the optimal path of the robot from the starting point to the target point according to the known scene data.

[0185] In the actual application process, the path planning device provided in this embodiment is essentially a computer device. This computer device can adopt an embedded device and be deployed in various mobile robots to support data processing and path planning. It can also be used as an independent computer device to support data processing requirements in certain scenarios. Such a non-embedded computer device can be a notebook computer, a tablet computer, a desktop computer, or a medium and large-sized computer device such as a rack-mounted server, a blade server, a tower server, or a cabinet server (including an independent server or a server cluster composed of multiple servers) that can execute computer programs.

[0186] Specifically, the computer device in this embodiment at least includes, but is not limited to, a memory and a processor that can communicate with each other through a system bus. In this embodiment, the memory (i.e., the readable storage medium) includes flash memory, a hard disk, a multimedia card, a card-type memory (such as an SD or DX memory, etc.), a random access memory (RAM), a static random access memory (SRAM), a read-only memory (ROM), an electrically erasable programmable read-only memory (EEPROM), a programmable read-only memory (PROM), a magnetic memory, a magnetic disk, an optical disk, etc. In some embodiments, the memory can be an internal storage unit of the computer device, such as the hard disk or memory of the computer device. In other embodiments, the memory can also be an external storage device of the computer device, such as a plug-in hard disk equipped on the computer device, a Smart Media Card (SMC), a Secure Digital (SD) card, a Flash Card, etc. Of course, the memory can also include both the internal storage unit and the external storage device of the computer device. In this embodiment, the memory is usually used to store the operating system and various application software installed on the computer device. In addition, the memory can also be used to temporarily store various data that have been output or will be output.

[0187] In some embodiments, the processor can be a Central Processing Unit (CPU), a controller, a microcontroller, a microprocessor, or other data processing chips. The processor is usually used to control the overall operation of the computer device.

[0188] Performance Test

[0189] To verify the performance of the proposed solution of the present invention, the technical personnel formulated an experimental plan. During the experiment, a map scenario as shown in Figure 7 was first constructed. This scenario is a grid map with a size of 30*30, which contains multiple obstacles. Each obstacle is represented by two points, indicating the starting point and the ending point of the obstacle. In the map, the obstacle list representing the obstacle distribution is as follows:

[0190] [(5,5),(5,15)],

[0191] [(15,10),(25,10)],

[0192] [(10,20),(10,25)],

[0193] [(5,20),(10,20)],

[0194] [(20,5),(20,15)],

[0195] [(10,5),(10,15)],

[0196] [(25,20),(25,25)],

[0197] [(10,20),(20,20)],

[0198] [(10,15),(10,25)],

[0199] [(25,20),(27,20)],

[0200] [(20,25),(25,25)],

[0201] [(25,15),(25,20)]

[0202] In this scenario, the proposed solution of the present invention was selected as the experimental group, and the RRT*, GWO, and GWO+JPS solutions were respectively used as the control groups. The experimental group and each control group were used to perform path optimization in this scenario, where the preset path starting point and ending point were (2,2) and (28,28) respectively.

[0203] In this experiment, each solution planned the best path from the starting point to the ending point in this scenario and could effectively avoid obstacles. Among them, the paths finally optimized by each solution are as shown in Figure 8 shown.

[0204] Analysis Figure 8From the data in , it can be seen that the path length planned by the RRT* scheme in Control Group 1 is 49.44. This scheme can be used as a benchmark for the optimization results of the other schemes. The path length optimized by the GWO scheme in Control Group 2 is 47.56, and the optimization rate compared with the benchmark scheme is 3.8%. The path length optimized by the GWO+JPS scheme in Control Group 3 is 42.34, and the optimization rate compared with the benchmark scheme is 14.4%. The path length optimized by the experimental group of the present invention is 41.15, and the optimization rate compared with the benchmark scheme is 16.8%. It can be seen from this that compared with the existing schemes, the optimized path length of the present invention is shorter and the quality of the planned path is higher.

[0205] The above-described embodiments only represent one implementation manner of the present invention, and its description is relatively specific and detailed, but it cannot be construed as a limitation to the scope of the invention. It should be noted that for those of ordinary skill in the art, without departing from the concept of the present invention, several modifications and improvements can still be made, and these all belong to the protection scope of the present invention. Therefore, the protection scope of the present invention should be subject to the appended claims.

Claims

1. A method for planning the motion path of a robot by fusing multiple algorithms, characterized in that, It includes: Obtain the position information of the robot and the surrounding environment information within the specified working area, construct a grid map based on the obtained information, and mark the obstacles in the robot and its surrounding environment, thereby obtaining scene data; Design the fitness function of the GWO gray wolf optimization algorithm with the path length, safety, and smoothness as evaluation indicators; Encode any possible path on the grid map as a gray wolf, and introduce a perturbation mechanism in the GWO gray wolf optimization algorithm to achieve population update, thereby obtaining an improved GWO gray wolf optimization algorithm; (ⅰ) Use the RRT* algorithm to perform global path planning according to the known scene data, generate a random path L0 from the starting point to the target point and passing through the corresponding nodes of multiple grids on the grid map, and smooth it; (ⅱ) Use the improved GWO gray wolf optimization algorithm to iteratively optimize L0 with L0 as the initial path to obtain a new path L1; (ⅲ) Simplify L1, delete redundant intermediate points to obtain a simplified path L2, and use the improved JPS jump point search algorithm to further optimize L2 to obtain a new path L3; (ⅳ) Use L3 as the initial path, and use the improved GWO gray wolf optimization algorithm to iteratively optimize the initial path again to obtain the optimal path L4.

2. The method for planning a robot motion path with multi-algorithm fusion according to claim 1, wherein The iterative optimization process of the improved GWO gray wolf optimization algorithm includes: S1: Preset algorithm parameters, including: the number of grey wolves N wolves , the maximum number of iterations N max2 , the path perturbation intensity σ; Initialize the grey wolf population, each grey wolf represents a possible path, the Alpha wolf represents the current optimal path, the Beta wolf represents the sub-optimal path, the Delta wolf represents the third-best path, and set the initial path L0: L0 = (L 01 , L 02 ,... L 0i ..., L 0n ), where L 0i represents the i-th path node in the initial path L0; S2: Use the perturbation mechanism to randomly adjust the current path to generate several new paths; check whether the new paths are valid. If so, add them to the wolf pack, otherwise regenerate new paths; S3: In each iteration, calculate the fitness value f(L k ) of each gray wolf L k through the fitness function; then update the leading wolf pack according to the fitness ranking, and use the three paths with relatively smaller fitness values as the Alpha wolf, Beta wolf, and Delta wolf respectively; S4: Convert the paths into arrays, and then calculate the distances D α α , D β β , D δ δ between each gray wolf in the population and the Alpha wolf, Beta wolf, and Delta wolf, and update its position; S5: Perform boundary clipping on the new positions of each gray wolf after update to ensure that each path represented by it is within the passable range of the grid map; S6: Repeat steps S2 - S5 until the maximum number of iterations N is reached max2 or the path fitness value converges; S7: Select the path represented by the Alpha wolf in the final optimization result, smooth it and then output.

3. The method for planning the motion path of a robot with multi-algorithm fusion according to claim 2, wherein: In step S2, the method of generating a new path using the perturbation mechanism is: Apply a random offset that follows a normal distribution N(0,σ) to the coordinates of each node in the original path for random perturbation; Verify whether the coordinates of the new node after perturbation are located in the passable area through the grid map. If the verification fails, repeat the perturbation operation on the corresponding node until the maximum number of attempts is reached; if no valid new coordinates are found after reaching the maximum number of attempts, retain the original node; Retain all the nodes after perturbation to form the required new path; In step S2, the path existence check, path node boundary check, and path node obstacle check methods are simultaneously used to evaluate whether the new path generated after perturbation is valid.

4. The method for planning a robot motion path with multi-algorithm fusion according to claim 2, wherein, The expression of the fitness function in step S3 is as follows: Among them, length(L k ) represents the length of the path corresponding to the gray wolf L k , penalty(L k ) represents the safety penalty term of the path corresponding to the gray wolf L k , smoothness(L k ) represents the smoothness penalty term of the path corresponding to the gray wolf L k ; (x i , y i ) represents each point on the path, i = 1…n, where n represents the number of points on the path; Grid[,] represents the function for discriminating obstacles, (dx, dy) represents the four-neighborhood offset, and θ i represents the direction angle of the path segment from P i-1 to P i ; |θ i - θ i-1 | represents the direction change amount of adjacent path segments.

5. The method for planning a robot motion path with multi-algorithm fusion according to claim 2, characterized in that: In step S4, the position X of any grey wolf k is updated according to the following formula: In the above formula, X α , X β , X δ represent the positions of Alpha wolf, Beta wolf and Delta wolf respectively; X k represents the original position of the k-th gray wolf, and X k ' represents the updated position of the k-th gray wolf; A1, A2, A3, C1, C2, C3 are random coefficients; Among them, the two types of random coefficients A and C are randomly generated by the following formula: Among them, a represents a parameter that linearly decreases with the number of iterations, and r1, r2 represent random numbers between [0,1].

6. The method for planning the motion path of a robot with multi-algorithm fusion according to claim 1, wherein: The process of generating a random path based on the RRT* algorithm includes: S01: Initialize the algorithm parameters, including: the maximum number of iterations N max1 , the step size λ, and the neighborhood radius R for optimizing the random tree; preset the structure of the extended random tree, and use the starting position of the robot as the root node of the extended random tree; S02: Randomly determine a random sampling point q within the search space that can avoid obstacles rand , as a candidate direction for the extension of the random tree; S03: Traverse the set of child nodes in the random tree to find the child node q rand that is closest to q near . Starting from q near , expand by a step size of λ in the direction of q near and q rand to obtain a new node q new ; S04: Check the path from q near to q new to see if it collides with an obstacle. If there is no collision, add the new node q new to the random tree and record its parent node as q near . Otherwise, discard the node and return to step S03 to re-determine the new q near ; S05: In q new Find all possible parent node candidate sets Q within the neighborhood radius R of near , calculated from Q near Each candidate node to q new The total path cost c(q newi ); then check from Q near Each candidate node to q new Whether the path collides with obstacles; delete the candidate nodes that may collide; finally select the total path cost c(q newi )The smallest node is used as the new parent node q new , update the random tree structure; S06: Optimize the re - wiring of nodes within the neighborhood radius R, and check whether the path cost of other nodes in Q can be reduced by taking q new as the parent node. If so, update its parent node to q near , and recalculate the path cost. Otherwise, skip the corresponding node; new ​ S07: Repeat S02 - S06 until the maximum number of iterations N is reached max1 or find a feasible path S from the starting point q G to the target point q S08: Starting from the end point of the random tree obtained after iteration termination, trace back along each parent node to the starting point q S to generate the corresponding random path and smooth it 7. The method for planning the motion path of a robot with multi-algorithm fusion according to claim 4, characterized in that: In step S05, for each candidate node in the parent node candidate set Q near to q new the total path cost c(q newi ) is calculated as follows; c(q newi ) = c(q i ) + distance(q i , q new ), In the above formula, c(q i ) represents the current path cost of any candidate node q i , distance(q i , q new ) represents the Euclidean distance between any candidate node q i and q new .

8. The method for planning a robot motion path with multi-algorithm fusion according to claim 1, characterized in that, The process of implementing path optimization using the improved JPS jump point search algorithm includes: S001: Initialize the algorithm parameters, including: starting point q S , target point q G and obstacle list S p ; Generate an open list openlist and a closed list closelist, and add the starting point q S to the openlist, and set its path cost c(q S ) = 0; S002: Select the node with the minimum path cost from the open list as the current node Q current , and move it from the open list to the closed list; S003: Determine q current whether it is the target point q G : If so, stop the search and end the iterative process; otherwise, perform a jump point search to obtain all valid successor nodes of the current node and form a successor node set; S004: Calculate the starting point q successively according to the following formula S to all successor nodes q in the successor node set i of the temporary path cost c tentative (q i ): c tentative (q i ) = c(q current ) + cost(q current , q i ) In the above formula, c(q current ) represents the actual path cost from the starting point q S to the current node q current , and cost(q current , q i ) represents the path cost from the current node q current to the successor node q i ; S005: Determine the successor node q i Check if it has been visited or if its temporary path cost is less than the previous path cost. If so, update the path information, record the predecessor node of node q i as q current , and record the actual path cost from the starting point q S to node q i as c tentative (q i ). Otherwise, return to step S004 to recalculate the temporary path cost of the next successor node; S006: Calculate the total path cost c(q i ) of each updated node q i , and move it to the open list when the corresponding node q i is not in the close list; where the calculation formula of c(q i ) is: c(q i ) = c tentative (q i ) + h(q i ), In the above formula, h(q i ) represents the heuristic cost from node q i to the target point q G ; S007: Repeat steps S002 - S006 until the target point q is found g or the openlist is empty, terminating the iteration process; S008: Starting from the end point obtained after the iteration termination, trace back along each predecessor node to the starting point q S Generate a path, and after smoothing the generated path, use it as the new path optimized by the JPS jump point search algorithm.

9. The method for planning a robot motion path with multi-algorithm fusion according to claim 8, wherein: The method of generating a successor node set by jump point search in step S003 is: First, generate several successor nodes corresponding to the current node through the target point priority detection method, the forced neighbor detection method, and the diagonal jump detection method respectively; then perform boundary checks and obstacle detections on each successor node, and exclude the nodes that belong to the grid boundary or whose connection directions pass through obstacles to obtain the final set of successor nodes.

10. A robot, characterized in that: It includes a path planning device, and the path planning device includes a memory, a processor, and a computer program stored on the memory and running on the processor. When the computer program is executed by the processor, it implements the steps of the multi-algorithm fusion robot motion path planning method described in any one of claims 1-8, and further plans the optimal path of the robot from the starting point to the target point according to the known scene data.

Images