Robot path planning method and device, robot equipment and medium

Abstract

The invention relates to a robot path planning method and device, robot equipment and a medium, and belongs to the technical field of robot motion control path planning, and the method comprises the steps: 1, exploring a motion path of a robot in a constructed target map space based on a preset rapid search random tree, obtaining an offset sampling point in the target map space based on a fixed direction offset strategy; 2, taking the direction vector of the bias sampling point as a reference direction, and adopting a four-direction discretization strategy to determine an exploration direction; 3, decomposing the exploration direction by adopting a layered exploration strategy, and expanding the random tree based on the decomposed exploration direction; and 4, when the random tree is not expanded to the connectable range of the target point of the robot, repeating the steps 1 to 3 until the random tree is expanded to the connectable range of the target point, and obtaining a path planning solution of the robot, thereby improving the path planning efficiency in the narrow scene.

Description

Technical Field

[0001] The present invention relates to the technical field of robot motion control path planning, and in particular to a robot path planning method, device, robot equipment and medium. Background Art

[0002] With the rapid development of autonomous driving technology, how to achieve efficient, safe, and optimized path planning in space has become one of the research hotspots. In order to efficiently complete the planning task, a variety of path planning algorithms have been proposed, including graph-based search, heuristic-based intelligent search, and random sampling-based search. Among them, the rapid search random tree (RRT) algorithm is based on random sampling of the unknown space, extracting spatial information, and exploring the spatial state through incremental growth of the tree. After continuously iterating random samples, the planner gradually establishes a set of trajectories. As the number of sampling points increases, the random tree can describe the structure of the unknown space with higher resolution. Therefore, RRT can easily integrate kinematic constraints, dynamic constraints, and environmental constraints. At the same time, the randomness of RRT enables it to jump out of the local area and ensure global convergence.

[0003] Although the RRT algorithm has probabilistic completeness in spatial exploration, its expansion mode has a significant efficiency bottleneck, and its planning efficiency is still insufficient in complex environments, especially in narrow channels with multiple geometric constraints. In addition, the artificial potential field method is combined with RRT to accelerate the expansion of random trees through gravitational field guidance. However, the repulsive field generated by the potential field at the entrance of the narrow channel will cause the expansion point to deviate from the optimal path direction, causing the algorithm to fall into a local extreme value and be unable to penetrate the channel.

[0004] Therefore, the existing technology has the problem of low path planning efficiency when facing narrow scenes. Summary of the Invention

[0005] In view of this, it is necessary to provide a robot path planning method, system and robot equipment to solve the technical problem of low path planning efficiency in narrow scenarios.

[0006] In order to solve the above problems, in a first aspect, the present invention provides a robot path planning method, comprising: Step 1: Based on the preset fast search random tree, the robot's motion path is explored in the constructed target map space, and bias sampling points are obtained in the target map space based on the fixed direction bias strategy; Step 2: Using the direction vector of the biased sampling point as a reference direction, a four-way discretization strategy is used to determine the exploration direction; Step 3: Decompose the exploration direction using a hierarchical exploration strategy, and expand the random tree based on the decomposed exploration direction; Step 4: When the random tree does not extend to the target point of the robot, repeat steps 1 to 3 until the random tree extends to the target point to obtain a path planning solution for the robot.

[0007] In one possible implementation, the fast search random tree uses the robot's starting point as the root node and a preset step size as the exploration step size to explore toward the robot's target point; the preset fast search random tree is used to explore the robot's motion path in the constructed target map space, and a biased sampling point is obtained in the target map space based on a fixed direction bias strategy, including: Constructing a direction sampling table, and assigning a sampling weight to each direction in the direction sampling table; Obtaining the expansion failure record when the fast search random tree explores the motion path in the target map space, and obtaining the expansion failure direction and the adjacent direction of the expansion failure direction in the direction sampling table according to the expansion failure record; Updating the sampling weights of the expansion failure direction and the directions adjacent to the expansion failure direction; Normalizing the updated sampling weights using the normalized weight probability distribution, and selecting the main direction based on the normalized probability distribution of the sampling weights; A random point on the main direction is determined, and a normal distribution bias distance and Gaussian perturbation are superimposed on the random point to determine a bias sampling point.

[0008] In a possible implementation, the sampling weight of the update extension failure direction is: , in, is the sampling weight for the direction of failure expansion, is the sampling weight of the updated expansion failure direction, For the current moment, For the next moment, is the bias weight; The sampling weights of the adjacent directions of the update extension failure direction are: , in, is the sampling weight of the adjacent direction of the updated expansion failure direction, is the sampling weight of the adjacent direction of the failed extension direction.

[0009] In a possible implementation, the calculation formula for the bias sampling point is: , in, is the bias sampling point, is a random point, is the offset distance, The main direction, For small disturbances.

[0010] In a possible implementation, the direction vector of the biased sampling point is used as a reference direction, and a four-way discretization strategy is adopted to determine the exploration direction, including: Calculating the Euclidean distance between the biased sampling point and the tree node, and determining the nearest tree node based on the Euclidean distance; Taking the nearest tree node as the origin, constructing a local homogeneous coordinate system, and dividing the plane of the target map into a plurality of mutually exclusive direction domains based on the local homogeneous coordinate system; Determine the unit direction vector of each direction domain and the direction vector of the nearest tree node and the offset sampling point; The cosine angle between the direction vector and the unit direction vector of each direction domain is calculated, and the exploration direction is determined based on the cosine angle.

[0011] In a possible implementation, decomposing the exploration direction using a hierarchical exploration strategy and expanding the random tree based on the decomposed exploration direction includes: Step 1: Decomposing the exploration direction into a first exploration direction and a second exploration direction based on the constructed three-way extension base; Step 2: Take the nearest tree node as the first starting state and perform single-step expansion along the first exploration direction to obtain the first new node; Step 3: When no obstacle is detected during single-step expansion along the first exploration direction, single-step expansion is performed in sequence along the second exploration direction with the first new node as the second starting state to obtain a second new node, and a backtracking optimization strategy is used to perform node optimization on the second new node to determine whether the second new node meets the backtracking optimization criteria; Step 4: When the second new node does not meet the retrospective selection criteria, repeat step 3 until the second new node meets the retrospective selection criteria; Step 5: When an obstacle is detected during single-step expansion along the second exploration direction, repeat steps 2 to 4 until the second new node meets the backtracking optimization criteria or an obstacle is detected during single-step expansion along the first exploration direction.

[0012] In a possible implementation, performing node selection on the second new node by adopting a backtracking optimization strategy to determine whether the second new node meets a backtracking optimization criterion includes: Taking the second new node as the center, constructing a first detection point matrix based on the orthogonal direction vector of the second new node, and performing bidirectional expansion detection along the orthogonal direction of the second new node; When no obstacle is detected during the expansion of the first detection point, a backtracking point is generated and a second detection point matrix is ​​constructed; When an obstacle is detected during the expansion of the second detection point, it is determined that the second new node meets the backtracking optimization criterion, the second new node is used as the leap node, and the random tree is updated.

[0013] In a second aspect, the present invention further provides a robot path planning device, comprising: The sampling point acquisition module is used to explore the robot's motion path in the constructed target map space based on a preset fast search random tree, and obtain biased sampling points in the target map space based on a fixed direction bias strategy; An exploration direction determination module, configured to determine the exploration direction using a four-way discretization strategy with the direction vector of the biased sampling point as a reference direction; A random tree expansion module, configured to decompose the exploration direction using a hierarchical exploration strategy, and expand the random tree based on the decomposed exploration direction; The path planning solution obtaining module is used to obtain the path planning solution of the robot when the rapid search random tree explores the target point of the robot.

[0014] In a third aspect, the present invention further provides a robotic device comprising: a processor and a memory; The memory stores a computer-readable program executable by the processor; When the processor executes the computer-readable program, the steps in the robot path planning method described above are implemented.

[0015] In a fourth aspect, the present invention also provides a storage medium for storing computer-readable programs or instructions, which, when executed by a processor, can implement the steps in the robot path planning method described in any one of the above-mentioned method items.

[0016] The beneficial effects of the present invention are as follows: in the process of exploring the robot's motion path by using a fast search random tree, bias sampling points are obtained in the target map space based on a fixed direction bias strategy, the exploration direction is determined according to four-way discretization, and then the random tree is expanded based on hierarchical exploration in the exploration direction to find the path solution, thereby completing the path planning of the mobile robot. By optimizing the expansion process of the fast search random tree, combining four-way discretization and hierarchical exploration, the topological structure of the random tree is effectively optimized. The fixed direction bias strategy multiplies the weight of the failed direction, collaboratively strengthens the adjacent directions, and periodically attenuates, so as to achieve a dynamic balance between exploration direction reinforcement and suppression, which significantly improves the global convergence efficiency in narrow scenarios. The four-way discretization strategy is used to determine the exploration direction, and a quadrant direction domain partitioning method is proposed. The normalized exploration vector is generated by the direction consistency maximization criterion, which effectively reduces the computational complexity. The hierarchical exploration strategy is used to expand the random tree, and the orthogonal basis decomposition is used to achieve rapid exploration of the local area, further shortening the planning time and improving the efficiency of path planning in narrow scenarios. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative work.

[0018] Figure 1 A flow chart of an embodiment of the robot path planning method provided by the present invention; Figure 2 A schematic diagram of a target map of the robot path planning method provided by the present invention; Figure 3 A schematic diagram of the hierarchical exploration strategy of the robot path planning method provided by the present invention; Figure 4 This is a flow chart of an embodiment of S103 of the robot path planning method provided by the present invention; Figure 5 A schematic diagram of the planning results of the robot path planning method provided by the present invention; Figure 6 A schematic structural diagram of an embodiment of a robot path planning device provided by the present invention; Figure 7 This is a schematic structural diagram of an embodiment of the robot device provided by the present invention. DETAILED DESCRIPTION

[0019] The preferred embodiments of the present invention will be described in detail below in conjunction with the accompanying drawings, wherein the accompanying drawings constitute a part of this application and are used together with the embodiments of the present invention to illustrate the principles of the present invention, and are not used to limit the scope of the present invention.

[0020] References herein to "embodiments" mean that a particular feature, structure, or characteristic described in connection with the embodiments may be included in at least one embodiment of the present invention. The appearance of this phrase in various places in the specification does not necessarily refer to the same embodiment, nor does it constitute a separate or alternative embodiment that is mutually exclusive of other embodiments. It is understood, both explicitly and implicitly, by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0021] Before presenting the embodiments, the following terms are explained.

[0022] Rapid Random Tree (RRT): is a path planning algorithm based on random sampling. It efficiently explores feasible paths in high-dimensional or non-convex spaces through incremental expansion of tree data structures. Its core goal is to quickly find a feasible path from the starting point to the target under complex constraints (such as obstacles and non-holonomic systems).

[0023] The present invention discloses a robot path planning method, apparatus, robot device, and medium, which can be used in a computer. The method, apparatus, or computer-readable storage medium involved in the present invention can be integrated with the above-mentioned apparatus or can be relatively independent.

[0024] A specific embodiment of the present invention discloses a robot path planning method, which can be executed by a computer, specifically by one or more processors of the computer. Figure 1 As shown, the robot path planning method includes: S101, step 1, exploring the robot's motion path in the constructed target map space based on a preset fast search random tree, and obtaining bias sampling points in the target map space based on a fixed direction bias strategy; It should be noted that the fixed direction bias strategy introduces a probabilistic bias in a specific direction during the random sampling process, guiding the tree structure to preferentially expand in a preset direction (such as the direction of the target point or a known feasible area), thereby reducing invalid searches and accelerating path convergence.

[0025] S102, step 2, using the direction vector of the bias sampling point as the reference direction, and using a four-way discretization strategy to determine the exploration direction; It should be noted that the four-way discretization strategy simplifies the exploration direction and effectively reduces the computational complexity.

[0026] S103, step 3, using a hierarchical exploration strategy to decompose the exploration direction, and expanding the random tree based on the decomposed exploration direction; It should be noted that the hierarchical exploration strategy enables rapid exploration of local areas, further shortening the planning time.

[0027] S104, step 4, when the random tree has not been expanded to the target point of the robot, repeat steps 1 to 3 until the random tree is expanded to the target point, and obtain the path planning solution of the robot.

[0028] In some embodiments, in step S101, the motion path of the robot is explored in the constructed target map space based on a preset fast search random tree. The fast search random tree takes the starting point of the robot as the root node and the preset step size as the exploration step size to explore the target point of the robot, and the fast search random tree outputs the real-time progress of the exploration in the form of T=(V,E), where T is the feasible path of the robot explored by the fast search random tree, V is the set of tree nodes, and E is the set of link relationships between tree nodes; the motion path of the robot is explored in the target map by using the fast search random tree. First, the map space is initialized, that is, the position information, size parameters and other relevant data of the obstacle are input into the map matrix to establish an initial target map model, and then the obstacle area is expanded based on the predetermined safety distance parameter to expand the obstacle boundary range to form an expanded obstacle area containing a safety margin, and the expansion obstacle area is formed by using For a schematic diagram of the target map in a simulated multi-channel narrow environment, see Figure 2 ,like Figure 2 As shown, the right direction is the positive direction of the X axis and the upward direction is the positive direction of the Y axis; secondly, set the planning parameters, whose specifications include the starting point and target point of the robot, that is, the starting position of the robot and target location ; Exploration step of random tree , the exploration step size is used to control the accuracy of the expansion process; the bias weight , the bias weight is used to increase the sampling probability of the direction of exploration failure and the direction adjacent to the exploration failure direction; the expected exploration distance , the expected exploration distance makes the sampling points more stably distributed within a certain range of the exploration direction; the disturbance parameter , the perturbation parameter is used to introduce a small random offset when generating sampling points; the attenuation weight , the decay weight is used for periodic weight decay to prevent local optimality, the number of sampling is set to 2000 times; the exploration step of the random tree is set to 5 meters; the bias weight ;Expected exploration distance Meters, 20% of the diagonal distance of the map; disturbance parameter Meters, 10% of the robot radius; attenuation weight .

[0029] Based on the fixed direction bias strategy, bias sampling points are obtained in the target map space. First, a direction sampling table is constructed and sampling weights are assigned to each direction in the direction sampling table. That is, the direction sampling table and the initial sampling weights of each direction are defined. Eight fixed directions are defined and the direction sampling table is constructed. , each direction is associated with a sampling weight , set the initial value of the sampling weight ; Obtain the expansion failure record when the fast search random tree explores the motion path in the target map space, obtain the expansion failure direction and the adjacent direction of the expansion failure direction in the direction sampling table according to the expansion failure record, update the sampling weights of the expansion failure direction and the adjacent direction of the expansion failure direction, multiply the sampling weight of the expansion failure direction, and perform collaborative compensation on the sampling weights of the adjacent direction of the expansion failure direction, that is, according to the expansion failure record, perform the multiplication of the sampling weight of the failure direction and the collaborative compensation of the failure adjacent direction; When the fast search random tree expands in the target map space, when the node Direction When the expansion fails, the sampling weight of the direction of the expansion failure is updated as follows: , in, is the sampling weight for the direction of failure expansion, is the sampling weight of the updated expansion failure direction, For the current moment, For the next moment, is the bias weight; At the same time, the sampling weights of the adjacent directions of the failed extension direction are updated as follows: , , in, is the sampling weight of the adjacent direction of the updated expansion failure direction, is the sampling weight of the adjacent direction of the failed extension direction; Normalize the updated sampling weights using the normalized weight probability distribution, and select the main direction based on the normalized probability distribution of the sampling weights. , that is, the sampling direction is selected based on the normalized weight probability distribution , the probability distribution of the normalized sampling weight is calculated as: , in, is the probability of the normalized sampling weight; Determine the random points in the main direction, superimpose the normal distribution bias distance and Gaussian perturbation on the random points in the main direction to determine the bias sampling points. The calculation formula of the bias sampling points is: , in, is the bias sampling point, is a random point, is the offset distance, , To explore the distance, The main direction, For a small disturbance, ; Rapid search random tree expands in the target map space. After each successful expansion, the sampling weights of all directions will be attenuated. The attenuated sampling weights are: ,in, .

[0030] Based on the successful expansion record, periodic attenuation of sampling weights in all directions is performed to suppress the cumulative effect of historical weights.

[0031] In some embodiments, in step S102, the direction vector of the bias sampling point is used as the reference direction, and a four-way discretization strategy is used to determine the exploration direction. A tree node is obtained by using a fast search random tree to explore the target point of the robot, and the Euclidean distance between the bias sampling point and the tree node is calculated. The nearest tree node is determined based on the Euclidean distance. , taking the nearest tree node as the origin, construct a local homogeneous coordinate system, and its local homogeneous coordinate system is: ,in, 、 Aligned with the global coordinate system, the plane of the target map is divided into multiple mutually exclusive direction domains based on the local homogeneous coordinate system, that is, the plane is divided into four mutually exclusive direction domains , whose direction domain satisfies: , in, is a symbol pair, which defines each direction domain through the symbol pair; determines the unit direction vector of each direction domain, and Define the unit direction vector , the unit direction vector satisfies: , Determine the direction vector between the nearest tree node and the bias sampling point, calculate the cosine angle between the direction vector and the unit direction vector of each direction domain, and determine the exploration direction based on the cosine angle , the calculation formula of its exploration direction is: , in, , is the angle between the direction vector and the unit direction vector of each direction domain; The direction with the smallest angle is preferentially selected as the exploration direction. By maximizing the direction consistency, the expansion vector is simplified and the exploration direction is obtained, which effectively reduces the computational complexity.

[0032] In some embodiments, in step S103, a hierarchical exploration strategy is used to decompose the exploration direction, and the random tree is expanded based on the decomposed exploration direction. For a schematic diagram of the hierarchical exploration strategy, please refer to Figure 3 , a flowchart of expanding tree nodes using a hierarchical exploration strategy is available. Figure 4 ,include: S1031, step 1, decomposing the exploration direction into a first exploration direction and a second exploration direction based on the constructed three-way extension base; Based on the local homogeneous coordinate system of four-way discretization, the direction of exploration will be Decomposed into the first exploration direction And the second exploration direction 、 ,in, is the unit vector of the original exploration direction, the second exploration direction 、 To define two unit vectors in orthogonal directions in the orthogonal complement space, the orthogonal complement space is: , in, is an orthogonal complementary space.

[0033] S1032, step 2, taking the nearest tree node as the first starting state, single-step expansion is performed along the first exploration direction to obtain a first new node; Select the nearest tree node As the first starting state of growth, single-step expansion is performed along the first exploration direction: ,in, is the first new node obtained, For the step size scalar, return the first detection flag .

[0034] S1033, step 3, when no obstacle is detected during single-step expansion along the first exploration direction, single-step expansion is performed in sequence along the second exploration direction with the first new node as the second starting state to obtain a second new node, and a backtracking optimization strategy is used to perform node optimization on the second new node to determine whether the second new node meets the backtracking optimization criteria; When a single step is expanded along the first exploration direction, no obstacles are detected, that is, , indicating that the random tree encounters no obstacles while expanding along the first exploration direction and selects the first new node As the second state of growth, it follows the second exploration direction 、 Perform single-step exploration to obtain the second new node 、 , and its calculation formula is: , After obtaining the second new node, the backtracking optimization strategy is used to select the second new node. Specifically, with the second new node as the center, the first detection point matrix is ​​constructed based on the orthogonal direction vector of the second new node, and bidirectional expansion detection is performed along the orthogonal direction of the second new node. When no obstacle is detected during the expansion of the first detection point, a backtracking point is generated and the second detection point matrix is ​​constructed; when an obstacle is detected during the expansion of the second detection point, the second new node is determined to meet the backtracking optimization criteria, the second new node is used as the leap node, and the random tree is updated. Specifically, given the unit vector of the current node expansion direction, an orthogonal basis matrix is ​​constructed. Its orthogonal basis matrix is: , The first column is the orthogonal direction , , the second column is the original direction , , based on the orthogonal basis matrix at the second new node Perform a single-step expansion along the orthogonal direction: , in, is the first detection point matrix, , Return the second detection flag according to the expansion result ,when When the first detection point is expanded and no obstacle is detected, it returns to the single-step backtracking point. , , at the backtracking point Single-step expansion is performed along the orthogonal direction. , in, is the second detection point matrix, , Return the third detection flag according to the expansion result ,when When an obstacle is detected during the expansion of the second detection point, the second new node is determined to meet the backtracking selection criteria and the second new node is used as the leap node. , update the random tree .

[0035] S1034, step 4, when the second new node does not meet the retrospective selection criteria, repeat step 3 until the second new node meets the retrospective selection criteria; After the second new node is selected by the backtracking optimization strategy, if the second new node does not meet the backtracking evaluation criteria, continue to explore in the second direction 、 Perform single-step exploration and repeat until the second new node meets the backtracking selection criteria, that is, until or , that is, obtain the leap node Or obstacles are detected in the second exploration direction.

[0036] S1035, step 5, when an obstacle is detected during single-step expansion along the second exploration direction, repeat steps 2 to 4 until the second new node meets the backtracking optimization criterion or an obstacle is detected during single-step expansion along the first exploration direction; When an obstacle is detected during single-step expansion along the second exploration direction, return to S1032 and continue along the first exploration direction. Perform single-step exploration and repeat until or , that is, obtain the leap node Or an obstacle is detected in the first exploration direction.

[0037] In some embodiments, in step S104, when the random tree has not been expanded to the target point of the robot, steps one to three are repeated until the random tree is expanded to the target point, and a path planning solution for the robot is obtained. After the random tree is expanded, if the random tree has not been expanded to the connectable range of the target point, the process returns to S101 until the random tree is expanded to the connectable range of the target point. That is, when the new node generated during the expansion of the random tree enters the distance threshold (exploration step) range of the target point, it means that the tree is close enough to the target point. At this time, the new node will be directly connected to the target point, thereby completing the path search and obtaining the path planning solution.

[0038] After improving the fast search random tree RRT through a fixed direction bias strategy, a four-way discretization strategy, and a hierarchical exploration strategy, it was compared with Bi-RRT under the same target map and planning parameters. The efficiency of robot path planning was compared using four evaluation indicators: Average time is the average planning time, Average cost is the average planning cost, Mean nodes is the average number of nodes used in planning, and Success rate is the planning success rate. The improved RRT algorithm and Bi-RRT algorithm were statistically analyzed in 50 independent runs under four planning scenarios. For specific data, please refer to Table 1. For a schematic diagram of the planning results, please refer to Figure 5 ; Table 1

[0039] Table 1 and Figure 5 It can be seen that the planning success rate of the improved RRT algorithm is higher than that of the Bi-RRT algorithm, the average planning time is lower than that of the Bi-RRT algorithm, the average planning cost is lower than that of the Bi-RRT algorithm, and the average number of nodes used in planning is much lower than that of the Bi-RRT algorithm. In narrow scenarios, the efficiency of the improved RRT algorithm for robot path planning is improved.

[0040] In summary, the robot path planning method provided by the present invention includes the following steps: step 1, exploring the robot's motion path in the constructed target map space based on a preset fast search random tree, and obtaining bias sampling points in the target map space based on a fixed direction bias strategy; step 2, taking the direction vector of the bias sampling point as the reference direction, and adopting a four-way discretization strategy to determine the exploration direction; step 3, adopting a hierarchical exploration strategy to decompose the exploration direction, and expanding the random tree based on the decomposed exploration direction; step 4, when the random tree is not extended to the target point of the robot, repeating steps 1 to 3 until the random tree is extended to the target point, and obtaining the path planning solution of the robot, thereby improving the efficiency of path planning in narrow scenes.

[0041] In order to better implement the robot path planning method in the embodiment of the present invention, based on the robot path planning method, correspondingly, Figure 6 As shown, an embodiment of the present invention further provides a robot path planning device, and the robot path planning device 600 includes: The sampling point acquisition module 601 is used to explore the robot's motion path in the constructed target map space based on a preset fast search random tree, and obtain biased sampling points in the target map space based on a fixed direction bias strategy; An exploration direction determination module 602 is configured to determine an exploration direction using a four-way discretization strategy with the direction vector of the bias sampling point as a reference direction; A random tree expansion module 603 is configured to decompose the exploration direction using a hierarchical exploration strategy and expand the random tree based on the decomposed exploration direction; The path planning solution obtaining module 604 is used to obtain the path planning solution of the robot when the rapid search random tree explores the target point of the robot.

[0042] like Figure 7 As shown, the present invention also provides a robot device 700 , which can be a computing device such as a mobile terminal, a desktop computer, a notebook, a palmtop computer, or a server. The robot device 700 includes a processor 701 , a memory 702 , and a display 703 . Figure 7Only some of the components of the robotic device 700 are shown, but it should be understood that implementing all of the shown components is not a requirement, and more or fewer components may alternatively be implemented.

[0043] In some embodiments, the memory 702 may be an internal storage unit of the robot device 700, such as the hard drive or memory of the robot device 700. In other embodiments, the memory 702 may also be an external storage device of the robot device 700, such as a plug-in hard drive, a Smart Media Card (SMC), a Secure Digital (SD) card, a flash memory card, etc. equipped on the robot device 700. Furthermore, the memory 702 may include both an internal storage unit of the robot device 700 and an external storage device. The memory 702 is used to store application software installed in the robot device 700 and various data, such as program code installed in the robot device 700. The memory 702 may also be used to temporarily store data that has been output or is about to be output. In one embodiment, the memory 702 stores a robot path planning program, which can be executed by the processor 701 to implement the robot path planning method of various embodiments of the present invention.

[0044] In some embodiments, the processor 701 may be a central processing unit (CPU), a microprocessor, or other data processing chip, configured to execute program codes or process data stored in the memory 702 , such as a robot path planning method.

[0045] In some embodiments, display 703 can be an LED display, a liquid crystal display, a touch-sensitive liquid crystal display, or an OLED (Organic Light-Emitting Diode) touchscreen. Display 703 is used to display identification information from the robot's path planning program and to display a visual user interface. Components 701-703 of robotic device 700 communicate with each other via a system bus.

[0046] In some embodiments, when the processor 701 executes the robot path planning program in the memory 702, the various steps in the robot path planning method described in the above embodiments are implemented. Since the robot path planning method has been described in detail above, it will not be repeated here.

[0047] Accordingly, the present invention also provides a computer-readable storage medium, which is used to store computer-readable programs or instructions. When the program or instructions are executed by a processor, it can implement the steps or functions in the robot path planning method provided by the above-mentioned method embodiments.

[0048] Those skilled in the art will appreciate that all or part of the process steps of the above-described embodiments can be implemented by instructing related hardware through a computer program, and the program can be stored in a computer-readable storage medium, such as a magnetic disk, an optical disk, a read-only memory, or a random access memory.

[0049] The above description is only a preferred specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any changes or substitutions that can be easily thought of by any technician familiar with this technical field within the technical scope disclosed by the present invention should be covered by the scope of protection of the present invention.

Claims

1. A robot path planning method, characterized in that: include: Step 1: Based on the preset fast search random tree, the robot's motion path is explored in the constructed target map space, and bias sampling points are obtained in the target map space based on the fixed direction bias strategy; Step 2: Using the direction vector of the biased sampling point as a reference direction, a four-way discretization strategy is used to determine the exploration direction; Step 3: Decompose the exploration direction using a hierarchical exploration strategy, and expand the random tree based on the decomposed exploration direction; Step 4: When the random tree does not extend to the target point of the robot, repeat steps 1 to 3 until the random tree extends to the target point to obtain a path planning solution for the robot.

2. The robot path planning method according to claim 1, characterized in that: The fast search random tree uses the robot's starting point as the root node and a preset step size as the exploration step size to explore the robot's target point; the preset fast search random tree is used to explore the robot's motion path in the constructed target map space, and a biased sampling point is obtained in the target map space based on a fixed direction bias strategy, including: Constructing a direction sampling table, and assigning a sampling weight to each direction in the direction sampling table; Obtaining the expansion failure record when the fast search random tree explores the motion path in the target map space, and obtaining the expansion failure direction and the adjacent direction of the expansion failure direction in the direction sampling table according to the expansion failure record; Updating the sampling weights of the expansion failure direction and the directions adjacent to the expansion failure direction; Normalizing the updated sampling weights using the normalized weight probability distribution, and selecting the main direction based on the normalized probability distribution of the sampling weights; A random point on the main direction is determined, and a normal distribution bias distance and Gaussian perturbation are superimposed on the random point to determine a bias sampling point.

3. The robot path planning method according to claim 2, characterized in that: The sampling weight of the update extension failure direction is: ， in, is the sampling weight for the direction of failure expansion, is the sampling weight of the updated expansion failure direction, For the current moment, For the next moment, is the bias weight; The sampling weights of the adjacent directions of the update extension failure direction are: ， in, is the sampling weight of the adjacent direction of the updated expansion failure direction, is the sampling weight of the adjacent direction of the failed extension direction.

4. The robot path planning method according to claim 2, characterized in that: The calculation formula of the bias sampling point is: ， in, is the bias sampling point, is a random point, is the offset distance, The main direction, For small disturbances.

5. The robot path planning method according to claim 2, characterized in that: The direction vector of the biased sampling point is used as a reference direction, and a four-way discretization strategy is adopted to determine the exploration direction, including: Calculating the Euclidean distance between the biased sampling point and the tree node, and determining the nearest tree node based on the Euclidean distance; Taking the nearest tree node as the origin, constructing a local homogeneous coordinate system, and dividing the plane of the target map into a plurality of mutually exclusive direction domains based on the local homogeneous coordinate system; Determine the unit direction vector of each direction domain and the direction vector of the nearest tree node and the offset sampling point; The cosine angle between the direction vector and the unit direction vector of each direction domain is calculated, and the exploration direction is determined based on the cosine angle.

6. The robot path planning method according to claim 4, characterized in that: Decomposing the exploration direction by adopting a hierarchical exploration strategy, and expanding the random tree based on the decomposed exploration direction, includes: Step 1: Decomposing the exploration direction into a first exploration direction and a second exploration direction based on the constructed three-way extension base; Step 2: Take the nearest tree node as the first starting state and perform single-step expansion along the first exploration direction to obtain the first new node; Step 3: When no obstacle is detected during single-step expansion along the first exploration direction, single-step expansion is performed in sequence along the second exploration direction with the first new node as the second starting state to obtain a second new node, and a backtracking optimization strategy is used to perform node optimization on the second new node to determine whether the second new node meets the backtracking optimization criteria; Step 4: When the second new node does not meet the retrospective selection criteria, repeat step 3 until the second new node meets the retrospective selection criteria; Step 5: When an obstacle is detected during single-step expansion along the second exploration direction, repeat steps 2 to 4 until the second new node meets the backtracking optimization criteria or an obstacle is detected during single-step expansion along the first exploration direction.

7. The robot path planning method according to claim 6, characterized in that: The adopting the backtracking optimization strategy to perform node optimization on the second new node to determine whether the second new node meets the backtracking optimization standard includes: Taking the second new node as the center, constructing a first detection point matrix based on the orthogonal direction vector of the second new node, and performing bidirectional expansion detection along the orthogonal direction of the second new node; When no obstacle is detected during the expansion of the first detection point, a backtracking point is generated and a second detection point matrix is ​​constructed; When an obstacle is detected during the expansion of the second detection point, it is determined that the second new node meets the backtracking optimization criterion, the second new node is used as the leap node, and the random tree is updated.

8. A robot path planning device, characterized in that: include: The sampling point acquisition module is used to explore the robot's motion path in the constructed target map space based on a preset fast search random tree, and obtain biased sampling points in the target map space based on a fixed direction bias strategy; An exploration direction determination module, configured to determine the exploration direction using a four-way discretization strategy with the direction vector of the biased sampling point as a reference direction; A random tree expansion module, configured to decompose the exploration direction using a hierarchical exploration strategy, and expand the random tree based on the decomposed exploration direction; The path planning solution obtaining module is used to obtain the path planning solution of the robot when the rapid search random tree explores the target point of the robot.

9. A robotic device, characterized in that: including memory and processor; The memory stores a computer-readable program executable by the processor; When the processor executes the computer-readable program, the steps in the robot path planning method according to any one of claims 1 to 7 are implemented.

10. A storage medium, characterized in that: Used to store computer-readable programs or instructions, which, when executed by a processor, can implement the steps of the robot path planning method described in any one of claims 1 to 7.

Images