A fast path planning method and system for a robotic arm based on dynamic candidate pool guidance and collision node inspiration

Abstract

The present invention provides a method and system for rapid path planning of a robot arm based on dynamic candidate pool guidance and collision node inspiration, belonging to the field of robot arm path planning. In order to solve the problems of repeated expansion failure of random trees and high time cost of path planning caused by the probabilistic target biased expansion method of existing robot arm path planning algorithms. On the basis of the existing RRTs algorithm, the present invention proposes a method for rapid path planning of a robot arm based on dynamic candidate pool guidance and collision node inspiration, which guides the optimal selection of the nearest node and adjusts the expansion direction of the collision node, improves the efficiency of path planning of the robot arm and shortens the time cost of path planning, realizes the rapid expansion of random trees toward the target node to generate collision-free nodes, thereby greatly improving the efficiency of path planning of the robot arm and reducing the time cost of path planning.

Description

Technical Field

[0001] The present invention relates to the technical field of robot arm path planning, and in particular to a robot arm fast path planning method and system based on dynamic candidate pool guidance and collision node inspiration. Background Art

[0002] Robotic arms have the advantages of high precision and high efficiency, and are widely used in production activities such as equipment processing and manufacturing, large-scale facility maintenance, and automatic fruit picking. Robotic arm path planning refers to planning and generating a collision-free path consisting of a series of discrete nodes (each node corresponds to a set of robot arm joint configurations) between the initial configuration and the target configuration of the robot arm, so that the robot arm can move from the initial state to the target state without collision along the path to complete the task. Robotic arm path planning is usually performed in a high-dimensional joint space, which is convenient for handling joint angle state constraints and avoiding the existence and multi-solution problems of the robot arm inverse kinematics solution. However, due to the multi-dimensional coupling and strong nonlinear mapping relationship between Cartesian space and high-dimensional joint space, it is impossible to explicitly describe the obstacle-free space of the robot arm movement in the high-dimensional joint space, which makes it difficult for planning algorithms such as Dijkstra and A* that rely on explicit grid maps to efficiently adapt to the robot arm system. In addition, with the increase of the dimension of the robot arm joint space and the complexity of the task, the computational complexity and time cost of path planning increase exponentially, which makes fast path planning of the robot arm still a challenging problem.

[0003] In order to solve the problem of high-dimensional joint space path planning for robotic arms, researchers at home and abroad have proposed a rapidly exploring random tree (RRT) algorithm based on random sampling. This algorithm plans a collision-free path for the robotic arm to move from the initial configuration to the target configuration by randomly sampling and expanding new nodes in the joint space of the robotic arm and incrementally growing random trees, thus avoiding the explicit expression of the obstacle-free joint space area of ​​the robotic arm. Therefore, it has good applicability to the motion planning problem of high-dimensional robotic arms. However, due to the strong randomness and blindness of the joint space sampling exploration process, the RRT algorithm has low path node expansion efficiency and slow random tree growth, which in turn leads to low efficiency and long time consumption of robotic arm path planning.

[0004] In recent years, in response to the actual needs of manufacturing automation for fast path planning of robotic arms, domestic and foreign scholars have successively proposed variant algorithms (expressed as RRTs) such as RRT-connect and GBi-RRT, which are improved from the RRT algorithm, and verified their effectiveness in the path planning task of the robotic arm. However, in order to speed up the growth of the random tree to the target node, the RRTs algorithm probabilistically sets the sampling point to coincide with the target node. By minimizing the cost determined by the Euclidean distance, the node closest to the target node is selected from the existing nodes in the random tree, and then a new node is generated and collision detection is performed. The expansion process will be iteratively executed until a new collision-free node is obtained. However, it turns out that the nearest neighboring node that fails to expand a new node toward the target node will also fail to expand toward the target node again due to collision. Therefore, the random tree will repeatedly generate invalid new nodes toward the target node, resulting in additional time consumption. This seriously limits the practical application of the robotic arm in tasks that are extremely sensitive to time consumption. In addition, in existing studies, if the new node encounters an obstacle, the new node extended from the nearest neighboring node to the sampling point will be discarded. Then, a new node is regenerated in the sampling space toward the new sampling point and collision detection is performed. The above process will be iteratively executed until a new collision-free node is successfully obtained or the iteration limit is reached. However, due to the randomness of the RRTs algorithm, the new sampling point may be located anywhere in the sampling space, and the new node extending from the node near the obstacle to the new sampling point still has a high collision probability, resulting in inefficient obstacle avoidance exploration and excessive time consumption, which limits the application of the RRTs algorithm in multi-obstacle environments. Summary of the invention

[0005] The technical problems to be solved by the present invention are:

[0006] In order to solve the problems of repeated random tree expansion failures and high path planning time cost caused by the probabilistic target biased expansion method of the existing robot arm path planning algorithm.

[0007] The present invention adopts the following technical solutions to solve the above technical problems:

[0008] The present invention provides a method for fast path planning of a robotic arm based on dynamic candidate pool guidance and collision node inspiration, comprising the following steps:

[0009] S100, establish the ground inertial coordinate system, the robot arm base coordinate system and the robot arm joint coordinate system, and determine the initial configuration Θ of the robot arm according to the robot arm operation task requirements start and the target configuration Θ goal ; Construct the collision bounding box model of environmental obstacles and determine the three-dimensional size and position coordinates of the collision bounding box models of all obstacles;

[0010] S200, create two mutually attractive random trees in the robot joint space at the same time E tree1 and random tree E tree2 , random tree E tree1 With the initial configuration of the robot arm Θ start As the starting node, with the target configuration Θ goal is the target node; random tree E tree2 The target configuration of the robot arm Θ goal As the starting node, with the initial configuration Θ start is the target node; random tree E tree1 and random tree E tree2 Alternately expand new nodes;

[0011] S300, given target bias baseline probability β goal ∈[0,1], generate random number β with uniform probability rand =Rand(1,1), if the random number β rand Less than the baseline probability β goal , then randomly generate Θ in the joint space rand As sampling point θ sample , and based on minimizing the weighted Euclidean distance from the random tree E tree1 Find the sample The nearest node Θ nearest ; If the random number β rand Greater than or equal to the baseline probability β goal , then select the target node Θ goal As sampling point θ sample , and based on minimizing the weighted Euclidean distance cost, from the candidate pool E unexplored Select the closest sampling point Θ from the existing nodes sample The nearest node Θ nearest ;

[0012] S400, obtaining sampling point θ in step S300 sample The nearest node Θ nearest After that, from Θ nearest Departure along Θ nearest and θ sample The connection direction Θ dir According to the predetermined step length θ step Expand to generate new node Θ new , judge Θ nearest With Θ new Whether the line of the node θ collides with the obstacle, if no collision occurs, the new node Θ new Add to Random Tree E tree1 In; if a collision occurs, based on the original expansion direction Θ dir Turn to get the new expansion direction Θ ver_unit, and then let the nearest node Θ nearest Along the new extension direction Θ ver_unit Expand to get a new node Θ new , and again judge Θ nearest With Θ new Whether the line of collides with the obstacle, if it collides again, discard Θ new Then jump to step S300. If no collision occurs, the successfully generated new node Θ new Add to Random Tree E tree1 middle;

[0013] S500, when random tree E tree1 Successfully expanded to obtain a new node Θ new Then, the new node Θ new With random tree E tree2 All nodes on the network are paired one by one, and shortcut connections are made from near to far according to the distance, and the given step length θ is calculated on the connection line. step_collision Perform interpolation point collision detection. If no paired nodes are detected where all interpolation points on the shortcut connection line do not collide, swap the two random trees and jump to step S300. If a paired node is found where there is no collision, it indicates that the two random trees E tree1 and E tree2 Successfully connected, then follow the growth step θ step Insert a new node Θ between paired nodes add , and add to the random tree E tree1 middle;

[0014] S600, in E tree1 and E tree2 From the connection point Θ new and θ other Start from each node and trace back to their respective starting nodes, and get a path Π which is connected in order by a series of discrete nodes path , thus obtaining the robot arm from the initial configuration Θ start Move to the target configuration Θ goal collision-free path.

[0015] Further, in step S300, when the random number β rand Less than the baseline probability β goal When the sampling point Θ sample The nearest node Θ nearest for,

[0016] Θ sample Θ rand , β rand <β goal (1)

[0017]

[0018] Among them, ω i represents the weight of the i-th joint angle of the robot, Θ k is a random tree E tree1 The kth node in .

[0019] Further, in step S300, when the random number β rand Greater than or equal to the baseline probability β goal When the sampling point Θ sample The nearest node Θ nearest for,

[0020] Θ sample Θ goal , β rand ≥β goal (3)

[0021]

[0022] E unexplored =Exclude(E unexplored , Θ nearest ) (5)

[0023] Among them, E unexplored =Exclude(E unexplored , Θ nearest ) indicates that the candidate pool E unexplored is selected as the target node Θ goal The nearest node's Θ nearest Remove candidate pool E unexplored .

[0024] Further, in step S400, from θ nearest Set out along Θ nearest and θ sample The connection direction Θ dir According to the predetermined step length θ step Expand to generate new node Θ new ,Right now,

[0025] Θ sn =Θ sample -Θ nearest (6)

[0026]

[0027] Θ new =Θ nearest +θ step Θ dir (8)

[0028] Among them, Θ sn Indicates the node expansion direction, Θdir A unit vector representing the direction along which the node extends.

[0029] Further, in step S400, the new expansion direction θ ver_unit The calculation method is,

[0030] Θ ver =Rand(1,n) (9)

[0031]

[0032]

[0033]

[0034] Among them, Θ dir is the original expansion direction of the node, Θ ver is the new expansion direction after turning, Θ ver_unit is the unit vector of the new expansion direction, i max is Θ dir The column number corresponding to the maximum absolute value of all elements of β flag It is a flag used to indicate whether there is a collision between the shortcut connection lines of two nodes.

[0035] Further, in step S500, the interpolation point collision detection is:

[0036]

[0037]

[0038] Among them, the function For E tree2 The elements in are rearranged in ascending order according to rule f.

[0039] Further, in step S500, after finding a collision-free paired node, the new node θ add Add to Random Tree E tree1 middle,

[0040] Θ add =Θ new +k(Θ near -Θ new ) / n add , k=1,2,…,n add -1 (15)

[0041]

[0042] A robot arm fast path planning system based on dynamic candidate pool guidance and collision node inspiration, the system has a program module corresponding to the above steps, and executes the steps in the above robot arm fast path planning method based on dynamic candidate pool guidance and collision node inspiration during operation.

[0043] A computer-readable storage medium stores a computer program, wherein the computer program is configured to implement the steps of a fast path planning method for a robotic arm based on dynamic candidate pool guidance and collision node inspiration when called by a processor.

[0044] Compared with the prior art, the present invention has the following beneficial effects:

[0045] ① In view of the problems of high time cost and poor expansion quality of existing path planning methods, the present invention proposes a node selection strategy guided by a dynamic candidate pool to guide the optimal selection of the nearest node. The node memory is given by the candidate pool to avoid the same node being selected as the nearest node of the target node multiple times, and the problem of repeated expansion failure of the random tree caused by the probabilistic target biased expansion method of the existing algorithm is solved. The planning process can jump out of the local optimum in time, and the random tree can be quickly expanded toward the target node to generate collision-free nodes, thereby greatly reducing the time cost of path planning.

[0046] ② In view of the high collision probability of existing path algorithms when sampling and expanding new nodes near obstacles, and the difficulty in quickly generating feasible new nodes, an obstacle avoidance area exploration strategy inspired by collision nodes is proposed to adjust the expansion direction of new nodes near obstacles, so that the random tree can quickly expand collision-free new nodes with less blindness to bypass the obstacle area, avoiding excessive repeated exploration and invalid expansion of the random tree near obstacles, thereby improving node quality and shortening path planning time.

[0047] In summary, the present invention takes the robot arm as the application object, improves and optimizes the traditional path planning algorithm, and constructs a fast path planning method for the robot arm based on dynamic candidate pool guidance and collision node inspiration. While ensuring the obstacle avoidance effect of the robot arm, it can respond to the task requirements of the robot arm more quickly, plan a collision-free path in a short time, and then perform more work tasks in the same time, which can better meet the fast and low time consumption requirements of the robot arm and has a high engineering application value. It can also be proved through comparative experiments with existing path planning algorithms that the present invention has improvements in time cost, expansion quality, memory cost and exploration quality, especially in implementation cost, the average time is only about 1 / 7 of the existing algorithm, which proves that this algorithm can effectively improve the path planning speed. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] Figure 1It is a flow chart of a method for fast path planning of a robotic arm based on dynamic candidate pool guidance and collision node inspiration in an embodiment of the present invention;

[0049] Figure 2 Schematic diagram of the operation task scenario and obstacle distribution of the robot arm in an embodiment of the present invention;

[0050] Figure 3 Schematic diagram of the coordinate system of the base of the robot arm and the coordinate systems of each joint of the robot arm in an embodiment of the present invention;

[0051] Figure 4 A schematic diagram of a process of guiding two random trees to establish a connection through a pairing node shortcut in an embodiment of the present invention;

[0052] Figure 5 This is a schematic diagram of a collision-free path quickly generated by a robotic arm using this method in an embodiment of the present invention. DETAILED DESCRIPTION

[0053] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, specific embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0054] Specific implementation plan 1: Combine Figures 1 to 5 As shown, the present invention provides a method for fast path planning of a robotic arm based on dynamic candidate pool guidance and collision node inspiration, comprising the following steps:

[0055] S100, establish the ground inertial coordinate system, the robot arm base coordinate system and the robot arm joint coordinate system, and determine the initial configuration Θ of the robot arm according to the robot arm operation task requirements start and the target configuration Θ goal ; Use a cuboid with adjustable 3D direction and 3D size to build a collision bounding box model of environmental obstacles, and determine the 3D size and pose coordinates of all obstacle bounding boxes;

[0056] S200, create two mutually attractive random trees in the robot joint space at the same time E tree1 and random tree E tree2 , random tree E tree1 With the initial configuration of the robot arm Θ start As the starting node, with the target configuration Θ goal is the target node; and the random tree E tree2 With the target configuration v of the robot arm goal As the starting node, with the initial configuration Θ start is the target node; random tree E tree1 and random tree E tree2 Alternately expand new nodes; in order to improve the adequacy and flexibility of random sampling, limit Θ according to the joint angle limit and its root node (Θstart or Θ goal ) to set the configuration sampling space of each random tree respectively;

[0057] S300, given target bias baseline probability β goal ∈[0,1], generate random number β with uniform probability rand =Rand(1,1), if the random number β rand Less than the baseline probability β goal , then randomly generate Θ in the joint space rand As sampling point θ sample , and based on minimizing the weighted Euclidean distance from the random tree E tree1 Find the sample The nearest node Θ nearest ,Right now,

[0058] Θ sample =Θ rand ,β rand <β goal (1)

[0059]

[0060] Among them, ω i represents the weight of the i-th joint angle of the robot, Θ k is a random tree E tree1 The kth node in ;

[0061] If the random number β rand Greater than or equal to the baseline probability β goal , then directly select the target node Θ goal As sampling point θ sample , and based on minimizing the weighted Euclidean distance cost, from the candidate pool E unexplored Select the closest sampling point Θ from the existing nodes sample The nearest node Θ nearest ,Right now,

[0062] Θ sample =Θ goal ,β rand ≥β goal (3)

[0063]

[0064] E unexplored =Exclude(E unexplored ,Θ nearest ) (5)

[0065] Among them, E unexplored =Exclude(E unexplored,Θ nearest ) indicates that the candidate pool E unexplored is selected as the target node Θ goal The nearest node's Θ nearest Remove candidate pool E unexplored ;

[0066] S400, from Θ nearest Set out along Θ nearest and θ sample The connection direction Θ dir According to the predetermined step length θ step Expand to generate new node Θ new ,Right now,

[0067] Θ sn =Θ sample -Θ nearest (6)

[0068]

[0069] Θ new =Θ nearest +θ step Θ dir (8)

[0070] Among them, Θ sn Indicates the node expansion direction, Θ dir Represents the unit vector along the node extension direction;

[0071] Judgment Θ nearest With Θ new Whether the line collides with the obstacle, if no collision occurs, the new node Θ new Add to Random Tree E tree1 If a collision occurs, based on the original expansion direction Θ dir Turn to get the new expansion direction Θ ver_unit , and then let the nearest node Θ nearest Along the new expansion direction Θ ver_unit Expand to get a new node Θ new , and again judge Θ nearest With Θ new Whether the line of θ collides with the obstacle, if it collides again, it will be discarded. new And jump to step S300, if no collision occurs, the successfully generated new node Θ new Add to Random Tree E tree1 middle;

[0072] Among them, the new expansion direction Θ ver_unit The calculation method is,

[0073] Θ ver=Rand(1,n) (9)

[0074]

[0075]

[0076]

[0077] Among them, Θ dir is the original expansion direction of the node, Θ ver is the new expansion direction after turning, Θ ver_unit is the unit vector of the new expansion direction, i max is Θ dir The column number corresponding to the maximum (non-zero value) of the absolute values ​​of all elements;

[0078] S500, when random tree E tree1 Successfully expanded to obtain a new node Θ new After that, Θ new With random tree E tree2 Pair all nodes on the network one by one, try to connect them by shortcut according to the distance from near to far, and follow the given step length θ on the connection line step_collision Perform interpolation point collision detection, that is,

[0079]

[0080]

[0081] Among them, the function For E tree2 The elements in β are rearranged in ascending order according to rule f, flag It is a sign used to indicate whether there is a collision between the shortcut connection lines of two nodes;

[0082] If no paired node is detected where all interpolation points on the shortcut connection do not collide, the two random trees are swapped and the process jumps to step S300. If a paired node is found where all interpolation points on the shortcut connection do not collide, it indicates that the two random trees E tree1 and E tree2 Successfully connected, such as Figure 4 As shown, then according to the growth step θ step Insert a new node Θ between paired nodes add , and add to the random tree E tree1 In, that is,

[0083] Θ add =Θ new +k(Θ near -Θ new ) / n add,k=1,2,…,n add -1 (15)

[0084] n add = Ceil(||Θ near -Θ new || / θ step ) (16)

[0085] S600, in E tree1 and E tree2 From the connection point Θ new and θ other Start from each node and trace back to their respective starting nodes, and get a path Π which is connected in order by a series of discrete nodes path , thus obtaining the robot arm from the initial configuration Θ start Move to the target configuration Θ goal collision-free path.

[0086] Specific embodiment 2: The present invention provides a robotic arm fast path planning system based on dynamic candidate pool guidance and collision node inspiration. The system has a program module corresponding to the above steps, and executes the steps in the above-mentioned robotic arm fast path planning method based on dynamic candidate pool guidance and collision node inspiration during operation.

[0087] The other combinations and connection relationships of this embodiment are the same as those of the first embodiment.

[0088] Specific embodiment three: The present invention provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and the computer program is configured to implement the steps of a robot arm fast path planning method based on dynamic candidate pool guidance and collision node inspiration when called by a processor.

[0089] The other combinations and connection relationships of this embodiment are the same as those of the first embodiment.

[0090] Comparative experiment

[0091] Considering that the most typical operation that a robot arm needs to perform in industrial manufacturing automation and intelligent tasks is to grasp an object and transfer it from the starting position to the target position, a typical robot arm grasping and transferring object task scenario with multiple obstacles is designed. Figure 2 As shown in , the path planning algorithm is required to plan a collision-free path for the robot arm, so that the robot arm can transfer the object held at the end to the target position along this path, and the robot arm does not collide with obstacles in the environment during the entire movement. Set the position, posture and three-dimensional size of each obstacle in the task scene as shown in Table 1 below, and establish the robot arm base coordinate system and the robot arm joint coordinate system as shown in Table 1 below. Figure 3 As shown, determine the initial configuration Θ of the robot arm start=[-0.21,-2.78,-1.22,-2.28,2.93,1.57]rad and target configuration Θ goal =[2.64,-1.80,-2.44,-2.04,4.21,1.57]rad.

[0092] Table 1 The position, posture and 3D size of each obstacle in the task scene

[0093]

[0094] The method of the present invention and the existing planning algorithm based on bidirectional exploration random tree are compared and applied to the above-mentioned robot path planning task. The path planning experiment is repeated 200 times respectively, and the average value and standard deviation of the 200 experimental data are calculated and listed in Table 2. As can be seen from Table 2, the robot fast path planning method based on dynamic candidate pool guidance and collision node inspiration proposed in the present invention has a lower time cost than the existing path planning algorithm, and the mean time used for path planning is about 1 / 7 of the mean time of the existing algorithm, and the node expansion quality is higher, which is about 2 times that of the existing algorithm. The method of the present invention can effectively complete the robot obstacle avoidance path planning task, and can plan a collision-free path in a short time, and then perform more work tasks in the same time, which can better meet the robot's fast and low time consumption requirements, and has a high engineering application value. And the memory occupied by this method is about 1 / 3 of the existing algorithm, while improving the exploration quality and expansion quality, greatly reducing the storage space.

[0095] Table 2 compares the performance index data of the existing algorithm and the method of the present invention in the robot path planning task

[0096]

[0097] Although the present invention is disclosed as above, the protection scope of the present invention is not limited thereto. Those skilled in the art may make various changes and modifications without departing from the spirit and scope of the present invention, and these changes and modifications will fall within the protection scope of the present invention.

Claims

1. A fast path planning method for a robotic arm based on dynamic candidate pool guidance and collision node inspiration, characterized in that: The following steps are involved: S100, establish the ground inertial coordinate system, the robot arm base coordinate system and the robot arm joint coordinate system, and determine the initial configuration Θ of the robot arm according to the robot arm operation task requirements start and the target configuration Θ goal ; Construct the collision bounding box model of environmental obstacles and determine the three-dimensional size and position coordinates of the collision bounding box models of all obstacles; S200, create two mutually attractive random trees in the robot joint space at the same time E tree1 and random tree E tree2 , random tree E tree1 With the initial configuration of the robot arm Θ start As the starting node, with the target configuration Θ goal is the target node; random tree E tree2 The target configuration of the robot arm Θ goal As the starting node, with the initial configuration Θ start is the target node; random tree E tree1 and random tree E tree2 Alternately expand new nodes; S300, given target bias baseline probability β goal ∈[0,1], generate random number β with uniform probability rand =Rand(1,1), if the random number β rand Less than the baseline probability β goal , then randomly generate Θ in the joint space rand As sampling point θ sample , and based on minimizing the weighted Euclidean distance from the random tree E tree1 Find the sample The nearest node Θ nearest ; If the random number β rand Greater than or equal to the baseline probability β goal , then select the target node Θ goal As sampling point θ sample , and based on minimizing the weighted Euclidean distance cost, from the candidate pool E unexplored Select the closest sampling point Θ from the existing nodes sample The nearest node Θ nearest ; S400, obtaining sampling point θ in step S300 sample The nearest node Θ nearest After that, from Θ nearest Departure along Θ nearest and θ sample The connection direction Θ dir According to the predetermined step length θ step Expand to generate new node Θ new , judge Θ nearest With Θ new Whether the line of the node θ collides with the obstacle, if no collision occurs, the new node Θ new Add to Random Tree E tree1 In; if a collision occurs, based on the original expansion direction Θ dir Turn to get the new expansion direction Θ ver_unit , and then let the nearest node Θ nearest Along the new extension direction Θ ver_unit Expand to get a new node Θ new , and again judge Θ nearest With Θ new Whether the line of collides with the obstacle, if it collides again, discard Θ new Then jump to step S300. If no collision occurs, the successfully generated new node Θ new Add to Random Tree E tree1 middle; S500, when random tree E tree1 Successfully expanded to obtain a new node Θ new Then, the new node Θ new With random tree E tree2 All nodes on the network are paired one by one, and shortcut connections are made from near to far according to the distance, and the given step length θ is calculated on the connection line. step_collision Perform interpolation point collision detection. If no paired nodes are detected where all interpolation points on the shortcut connection line do not collide, swap the two random trees and jump to step S300. If a paired node is found where there is no collision, it indicates that the two random trees E tree1 and E tree2 Successfully connected, then follow the growth step θ step Insert a new node Θ between paired nodes add , and add to the random tree E tree1 middle; S600, in E tree1 and E tree2 From the connection point Θ new and θ other Start from each node and trace back to their respective starting nodes, and get a path Π which is connected in order by a series of discrete nodes path , thus obtaining the robot arm from the initial configuration Θ start Move to the target configuration Θ goal collision-free path.

2. According to claim 1, a method for fast path planning of a robotic arm based on dynamic candidate pool guidance and collision node inspiration is characterized in that: In step S300, when the random number β rand Less than the baseline probability β goal When the sampling point Θ sample The nearest node Θ nearest for, I sample =Θ rand ,b rand <b goal (1) Among them, ω i represents the weight of the i-th joint angle of the robot, Θ k Represents a random tree E tree1 The kth node in .

3. According to claim 2, a method for fast path planning of a robotic arm based on dynamic candidate pool guidance and collision node inspiration is characterized in that: In step S300, when the random number β rand Greater than or equal to the baseline probability β goal When the sampling point Θ sample The nearest node Θ nearest for, I sample =Θ goal ,b rand ≥β goal (3) AND unexplored =Excludes(E unexplored ,Θ nearest ) (5) Among them, E unexplored =Exclude(E unexplored ,Θ nearest ) indicates that the candidate pool E unexplored is selected as the target node Θ goal The nearest node's Θ nearest Remove candidate pool E unexplored .

4. According to claim 3, a method for fast path planning of a robotic arm based on dynamic candidate pool guidance and collision node inspiration is characterized in that: In step S400, from θ nearest Set out along Θ nearest and θ sample The connection direction Θ dir According to the predetermined step length θ step Expand to generate new node Θ new ,Right now, I sn =Θ sample -I nearest (6) I new =Θ nearest +θ step I dir (8) Among them, Θ sn Indicates the node expansion direction, Θ dir A unit vector representing the direction along which the node extends.

5. The method for fast path planning of a robotic arm based on dynamic candidate pool guidance and collision node inspiration according to claim 4, characterized in that: In step S400, the new expansion direction θ ver_unit The calculation method is, I ver =Rand(1,n) (9) Among them, Θ dir is the original expansion direction of the node, Θ ver is the new expansion direction after turning, Θ ver_unit is the unit vector of the new expansion direction, i max is Θ dir The column number corresponding to the maximum absolute value of all elements in .

6. The method for fast path planning of a robotic arm based on dynamic candidate pool guidance and collision node inspiration according to claim 5, characterized in that: In step S500, the interpolation point collision detection is: Among them, the function For E tree2 The elements in β are rearranged in ascending order according to rule f, flag It is a flag used to indicate whether there is a collision between the shortcut connection lines of two nodes.

7. The method for fast path planning of a robotic arm based on dynamic candidate pool guidance and collision node inspiration according to claim 6, characterized in that: In step S500, after finding a collision-free paired node, the new node θ add Add to Random Tree E tree1 middle, I add =Θ new +k(Θ near -I new ) / n add ,k=1,2,…,n add -1 (15) n add =Ceil(||Θ near -I new || / θ step ) (16).

8. A fast path planning system for a robotic arm based on dynamic candidate pool guidance and collision node inspiration, characterized in that: The system has a program module corresponding to the steps of any one of claims 1 to 7, and executes the steps in the above-mentioned robot arm fast path planning method based on dynamic candidate pool guidance and collision node inspiration during operation.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and the computer program is configured to implement the steps of the robot arm fast path planning method based on dynamic candidate pool guidance and collision node inspiration according to any one of claims 1 to 7 when called by a processor.

Images