A robot arm optimization control method

Abstract

The present application belongs to the technical field of robot arm planning and control methods, and in particular, relates to a robot arm optimization control method. It mainly includes the following steps: scene reconstruction; starting from the current robot arm position, searching in the direction of the target node to form a new node chain; counting the success rate and determining the best growth direction; judging whether the target node area has been reached, and taking corresponding measures. The robot arm optimization control method of the present application is mainly used to further improve the control effects of various types of robot arm movements. While retaining the random characteristics of the global sample, it uses local optimization and control of the growth process to improve the robot arm control process and reduce computing resources and operating costs.

Description

Technical Field

[0001] The present application belongs to the technical field of robot arm planning and control methods, and in particular, relates to a robot arm optimization control method. Background Art

[0002] The construction of an optimized control scheme for a robotic arm is often regarded as solving the optimal collision-free pathfinding problem for the robotic arm's head nodes. Common solutions include fast random exploration trees (RRT) and RRT*, which combine three-dimensional scene reconstruction with visual information collection, target recognition technology, and various optimization algorithms to determine the best control scheme. However, traditional search schemes are based on random sampling to obtain node data, so there are inevitably problems such as unstable sample quality and high uncertainty. These problems, on the one hand, further increase the required computing power and execution costs, and on the other hand, limit the robot's further accurate and smooth control effects. Summary of the invention

[0003] The purpose of this application is to provide a control scheme that is more stable and has better control effects based on actual needs, which is conducive to achieving more stable control. It can quickly optimize samples based on the random sampling characteristics of nodes, further soften the robot arm movements and improve the efficiency of the movements, and is an optimized control method for the robot arm.

[0004] To achieve the above objectives, this application adopts the following technical solutions.

[0005] A robot arm optimization control method of the present application mainly includes the following steps:

[0006] S1, scene reconstruction;

[0007] Determine the current motion space domain O t State, determine the obstacle area O around the robot obs And the feasible region O may , determine the final target domain O last ; Determine the RRT random tree step size d0, deflection rate ρ0, success rate ρ1, detection period T0, target node a last ;

[0008] S2, starting from the current robot position a0 and stepping d0, moves to the preset target node a according to probability ρ0. last Direction search forms a new node chain {a i}, i = 1, 2, ... I;

[0009] Statistics are made to see whether the success rate of generating new nodes in the fixed growth period T0 exceeds ρ1. If it exceeds ρ1, the search continues according to the current deflection rate ρ0. When the success rate of generating new nodes in the fixed growth period T0 is lower than ρ1, the search moves to the last generated node a. IThe feasible domain O may The randomly generated intermediate target node a mid Conduct a search;

[0010] S3, repeat step S2, each time searching for the randomly generated intermediate target node, according to the current node chain {a i All nodes in} and the intermediate target node a mid The distance determines the best growth direction;

[0011] S4. Perform several searches to determine whether the target node a is reached. last If we reach the target node a last Field of O j , then to the target node a last Continue searching in the direction. If the target node a is not reached, last If the area is not suitable, adjust the relevant parameters in step S1 and search again;

[0012] S5, based on the above steps, obtain the momentum position from the robot arm a0 to the target node a last The path chain L o = {b k}, k = 1, 2...K; where b1 = a0, b K =a last ;

[0013] S6, starting from the first node b1 of the current path chain, connect it with other subsequent nodes except the second node in the order of the path chain with straight paths until the straight path falls into the obstacle area O obs , determine the last one that does not fall into the obstacle domain O obs The end node b of the straight line path r , delete nodes b1 and b r r-2 nodes between them, and a new path chain L = {b1,b r ,b r+1 ....b K}.

[0014] A further improvement or preferred implementation of the above-mentioned robot arm optimization control method further includes step S7;

[0015] S7, the terminal node b of the straight path determined in step S6 r As the starting point, follow the path chain L = {b1,b r ,b r+ 1....b K}; Connect it with other subsequent nodes except the second node in sequence with a straight path until the straight path falls into the obstacle domain O obs, determine the last one that does not fall into the obstacle domain O obs The end node b of the straight line path r+m , delete node b r and node b r+m m-1 nodes between them, and a new path chain L is obtained + ={b1,b r ,b r+m ,b r+m+1 ,b r+m+2 ....b K}; Repeat the above steps until the end node of the straight line path is a last ; Determine the optimal path chain L opt ={c1,c2...c n ...c N}, where c1 = a0, c N =a last ;

[0016] A further improvement or preferred implementation of the above-mentioned robot arm optimization control method further includes step S8;

[0017] S8, based on the optimized path chain L determined in step S7 opt ={c1,c2...c n ...c N}, get the path chain L opt Internal segment path segment S0 = {s1, s2...s n ,s n+1 ....s N-1}, follow the path chain L opt The connection order in s is compared with two adjacent path segments s n =c n c n+1 and n+1 =c n+1 c n+2 The angle α between n =∠c n c n+1 c n+2 , if the angle α n ≤α min , then search for a new node c' on the midline of the angle formed between the two path segments n+1 , so that the new node c' n+1 The angle ∠c between the line connecting the endpoints of the adjacent road segment n c' n+1 c n+2 ≥α min , if there is a new node new node c' n+1 And the new node is located in the feasible region O may In the example above, we use the new node c'n+1 Replace node node c n+1 Get the new path chain L opt+ ={c1,...c n ,c' n+1 ,c n+2 ...c N}, repeat the above steps, compare all adjacent path segments in the optimized path chain in turn and update the node information to obtain the final path chain L last ;

[0018] A further improvement or preferred implementation of the above-mentioned robot arm optimization control method, wherein step 3 further comprises:

[0019] S3, given the search radius r1, for each randomly determined intermediate target node a mid , determine the intermediate target node a mid All known "adjacent" nodes within the radius r1 are considered as intermediate target nodes a. mid The parent node of the node from the initial node a0 to node a mid Several node chains; calculate the path cost of each node chain respectively, and determine the path from the initial node a0 to node a mid The node chain with the lowest path cost takes the corresponding node chain as the intermediate target node a mid Generate a chain and delete its original parent node relationship, and continue searching according to the growth order of the new node chain, and repeat the above steps each time a random search is performed.

[0020] Its beneficial effects are:

[0021] The robot arm optimization control method of the present application is mainly used to further improve the motion control effects of various types of robot arms. While retaining the random characteristics of global samples, it utilizes local optimization and growth process control to improve the robot arm control process and reduce computing resources and operating costs. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] Figure 1 It is a flow chart of the optimization control method of the mechanical arm;

[0023] Figure 2 It is a schematic diagram of the search path before optimization;

[0024] Figure 3 It is a schematic diagram of the path after the search. DETAILED DESCRIPTION

[0025] The present application is described in detail below in conjunction with specific embodiments.

[0026] This application mainly relates to an optimization scheme mainly used for robot arm control, reducing the computational cost of robot arm motion schemes and improving their execution efficiency. Based on the traditional robot arm control scheme, this application further improves the efficiency and effectiveness of sample search by optimizing and controlling random samples in the control scheme, improves the robot arm control effect, and reduces system resource consumption.

[0027] The basic process of the robot arm optimization control method of this application is as follows Figure 1 As shown, the following is a detailed description combined with specific steps:

[0028] S1. Scene reconstruction, that is, based on the robot arm itself or the supporting information acquisition device or the existing database, the scene information within a certain range of the robot arm is obtained in real time during the movement of the robot arm, so as to determine the basic control scheme of the robot arm.

[0029] For this application, its main purpose is to determine the current motion space domain O t State, determine the obstacle area O around the robot obs And the feasible region O may , determine the final target domain O last ;

[0030] At the same time, according to the preset configuration or actual needs, the RRT random tree step size d0, deflection rate ρ0, success rate ρ1, detection period T0, target node a last and other operation or control parameters; the above parameters can be extracted based on historical control schemes or obtained through simulation control, machine learning, etc.

[0031] S2, starting from the current robot position a0 and stepping d0, moves to the preset target node a according to probability ρ0. last Direction search forms a new node chain {a i}, i = 1, 2, ... I;

[0032] Statistics are made to see whether the success rate of generating new nodes in the fixed growth period T0 exceeds ρ1. If it exceeds ρ1, the search continues according to the current deflection rate ρ0. When the success rate of generating new nodes in the fixed growth period T0 is lower than ρ1, the search moves to the last generated node a. I The feasible domain O may The randomly generated intermediate target node a mid Conduct a search;

[0033] In the traditional random tree search process, this application uses the deflection probability parameter to control the deflection direction of the newly generated node, and expands it to a certain direction and an uncertain direction according to a certain probability while generating it randomly, thereby qualitatively improving the search direction, and using the success rate to determine whether the current growth direction can achieve rapid search expansion, achieving the effect of combining active guidance and random sampling, providing a basis for achieving rapid search, purposefully controlling the randomness of the system, and optimizing search results;

[0034] S3, repeat step S2, each time searching for the randomly generated intermediate target node, according to the current node chain {a i All nodes in} and the intermediate target node a mid The distance determines the best growth direction; specifically, given the search radius r1, for each randomly determined intermediate target node a mid , determine the intermediate target node a mid All known "adjacent" nodes within the radius r1 are considered as intermediate target nodes a. mid The parent node of the node from the initial node a0 to node a mid Several node chains; calculate the path cost of each node chain respectively, and determine the path from the initial node a0 to node a mid The node chain with the lowest path cost takes the corresponding node chain as the intermediate target node a mid Generate a chain, delete its original parent node relationship, and continue searching according to the growth order of the new node chain, and repeat the above steps each time a random search is performed;

[0035] The purpose of this step is to reconnect existing nodes by optimizing the path chain, delete redundant paths, and reduce the path cost to reach the node on the basis of considering the path cost, thereby improving the control effect, reducing the control cost, and achieving streamlined optimization of path search and cost reduction and efficiency improvement.

[0036] The process is as follows Figure 2 , Figure 3 As shown; the dotted line is the path before and after the new node, the direction of the arrow is the direction of node generation, and the number on the line is the path cost.

[0037] S4. Perform several searches to determine whether the target node a is reached. last If we reach the target node a last Field of O j , then to the target node a last Continue searching in the direction. If the target node a is not reached, last If the area is not suitable, adjust the relevant parameters in step S1 and search again;

[0038] S5, based on the above steps, obtain the momentum position from the robot arm a0 to the target node a last The path chain L o = {b k}, k = 1, 2...K; where b1 = a0, b K =a last ;

[0039] S6, starting from the first node b1 of the current path chain, connect it with other subsequent nodes except the second node in the order of the path chain with straight paths until the straight path falls into the obstacle area O obs , determine the last one that does not fall into the obstacle domain O obs The final node b of the straight line path r , delete nodes b1 and b r r-2 nodes between them, and a new path chain L = {b1,b r ,b r+1 ....b K};

[0040] S7, the terminal node b of the straight path determined in the above steps r As the starting point, follow the path chain L = {b1,b r ,b r+1 ....b K}; Connect it with other subsequent nodes except the second node in sequence with a straight path until the straight path falls into the obstacle domain O obs , determine the last one that does not fall into the obstacle domain O obs The end node b of the straight line path r+m , delete node b r and node b r+m m-1 nodes between them, and a new path chain L is obtained + ={b1,b r ,b r+m ,b r+m+1 ,b r+m+ 2....b K}; Repeat the above steps until the end node of the straight line path is a last ; Determine the optimal path chain L opt ={c1,c2...c n ...c N}, where c1 = a0, c N =a last ;

[0041] S8, based on the determined optimized path chain L opt ={c1,c2...c n ...c N}, get the path chain Lopt Internal segment path segment S0 = {s1, s2...s n ,s n+1 ....s N-1}, follow the path chain L opt The connection order in s is compared with two adjacent path segments s n =c n c n+1 and n+1 =c n+1 c n+2 The angle α between n =∠c n c n+1 c n+2 , if the angle α n ≤α min , then search for a new node c' on the midline of the angle formed between the two path segments n+1 , so that the new node c' n+1 The angle ∠c between the line connecting the endpoints of the adjacent road segment n c' n+1 c n+2 ≥α min , if there is a new node new node c' n+1 And the new node is located in the feasible region O may In the example above, we use the new node c' n+1 Replace node node c n+1 Get the new path chain L opt+ ={c1,...c n ,c' n+1 ,c n+2 ...c N}, repeat the above steps, compare all adjacent path segments in the optimized path chain in turn and update the node information to obtain the final path chain L last .

[0042] Based on the aforementioned steps S6 to S8, the path chain determined based on the random tree search results is further optimized by optimizing the paths between nodes and optimizing the acute angles of the path segments, thereby shortening the path chain, reducing the number of nodes, and making the path chain smoother, thereby effectively improving the efficiency of the robot arm operation.

[0043] Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present application, rather than to limit the scope of protection of the present application. Although the present application has been described in detail with reference to the preferred embodiments, ordinary technicians in this field should understand that the technical solution of the present application can be modified or replaced by equivalents without departing from the essence and scope of the technical solution of the present application.

Claims

1. A robot arm optimization control method, characterized in that: The steps include: S1, a step for scene reconstruction and determining basic parameters; specifically comprising: determining the current motion space domain O t State, determine the obstacle area O around the robot obs And the feasible domain O may , determine the final target domain O last ; Determine the RRT random tree step size d0, deflection rate ρ0, success rate ρ1, detection period T0, target node a last ; S2, the step of generating nodes based on probabilistic deflection, specifically, taking the current robot position a0 as the starting point and d0 as the step length, the robot moves toward the preset target node a according to probability ρ0. last Direction search forms a new node chain {a i }, i = 1, 2, ... I; Statistics are made to see whether the success rate of generating new nodes in the fixed growth period T0 exceeds ρ1. If it exceeds ρ1, the search continues according to the current deflection rate ρ0. When the success rate of generating new nodes in the fixed growth period T0 is lower than ρ1, the search moves to the last generated node a. I The feasible domain O may The randomly generated intermediate target node a mid Conduct a search; S3, repeat step S2, each time searching for the randomly generated intermediate target node, according to the current node chain {a i All nodes in} and the intermediate target node a mid The distance determines the best growth direction; S4. Perform several searches to determine whether the target node a is reached. last If we reach the target node a last Field of O j , then to the target node a last Continue searching in the direction. If the target node a is not reached, last If the area is not suitable, adjust the relevant parameters in step S1 and search again; S5: Based on the above steps, obtain the distance from the robot's starting position a0 to the target node a. last The path chain L o = {b k }, k = 1, 2...K; where b1 = a0, b K =a last ; S6, starting from the first node b1 of the current path chain, connect it with other subsequent nodes except the second node in the order of the path chain with straight paths until the straight path falls into the obstacle area O obs , determine the last one that does not fall into the obstacle domain O obs The end node b of the straight line path r , delete nodes b1 and b r r-2 nodes between them, and a new path chain L = {b1,b r ,b r+1 ....b K }.

2. The robot arm optimization control method according to claim 1, characterized in that: Also includes step S7; S7, the terminal node b of the straight path determined in step S6 r As the starting point, follow the path chain L = {b1,b r ,b r+1 ....b K }; Connect it with other subsequent nodes except the second node in sequence with a straight path until the straight path falls into the obstacle domain O obs , determine the last one that does not fall into the obstacle domain O obs The end node b of the straight line path r+m , delete node b r and node b r+m m-1 nodes between them, and a new path chain L is obtained + ={b1,b r ,b r+m ,b r+m+1 ,b r+m+2 ....b K }; Repeat the above steps until the end node of the straight line path is a last ; Determine the optimal path chain L opt ={c1,c2...c n ...c N }, where c1 = a0, c N =a last .

3. The robot arm optimization control method according to claim 2, characterized in that: Also includes step S8; S8, based on the optimized path chain L determined in step S7 opt ={c1,c2...c n ...c N }, get the path chain L opt Internal segment path segment S0 = {s1, s2...s n ,s n+1 ....s N-1 }, follow the path chain L opt The connection order in s is compared with two adjacent path segments s n =c n c n+1 and n+1 =c n+1 c n+2 The angle α between n =∠c n c n+1 c n+2 , if the angle α n ≤α min , then search for a new node c' on the midline of the angle formed between the two path segments n+1 , so that the new node c' n+1 The angle ∠c between the line connecting the endpoints of the adjacent road segment n c' n+1 c n+2 ≥α min , if there is a new node new node c' n+1 And the new node is located in the feasible region O may In the example above, we use the new node c' n+1 Replace node node c n+1 Get the new path chain L opt+ ={c1,...c n ,c' n+1 ,c n+2 ...c N }, repeat the above steps, compare all adjacent path segments in the optimized path chain in turn and update the node information to obtain the final path chain L last .

4. The robot arm optimization control method according to claim 1, characterized in that: The step 3 also includes: S3, given the search radius r1, for each randomly determined intermediate target node a mid , determine the intermediate target node a mid All known "adjacent" nodes within the radius r1 are considered as intermediate target nodes a. mid The parent node of the node from the initial node a0 to node a mid Several node chains; calculate the path cost of each node chain respectively, and determine the path from the initial node a0 to node a mid The node chain with the lowest path cost takes the corresponding node chain as the intermediate target node a mid Generate a chain and delete its original parent node relationship, and continue searching according to the growth order of the new node chain, and repeat the above steps each time a random search is performed.

Images