Modular self-reconfigurable robot fault-tolerant control method

Abstract

The modular self-reconfigurable robot fault-tolerant control method comprises the following steps: dividing a substructure, after the modular self-reconfigurable robot detects a fault position and a fault angle, searching for the number of reachable poses of the robot when the fault module is at different positions by a Monte Carlo method, obtaining a target position of the fault module according to the robot structure when the maximum number of reachable poses is obtained, and combining the initial position of the fault module to decompose all the modules into three substructures. Inter-substructure reconfiguration, before the different substructures are separated, the modular robot needs to connect the end module to the magnetic connection surface to form a closed chain, establish the kinematics model corresponding to the combined substructure, determine the node configuration and the motion path, and complete the fault-tolerant control of the modular self-reconfigurable robot according to the inter-substructure reconfiguration process. The modular self-reconfigurable robot fault-tolerant control method can generate a reconfiguration plan, and improves the fault tolerance and operation ability of the robot.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of robots, and relates to a fault-tolerant control method for a modular self-reconfigurable robot when a fault module exists. BACKGROUND

[0002] With the continuous development of robot technology, its application field is increasingly wide. In extreme harsh environments such as space exploration, nuclear power operation and maintenance, deep sea exploration, it is often not possible to have artificial maintenance or the cost of artificial maintenance is extremely high, so the ability of robots to cope with faults is particularly important. Robots of fixed configuration generally achieve fault tolerance through kinematic redundancy, but the fixed configuration limits the further improvement of fault tolerance capability. Modular self-reconfigurable robots have unique advantages in handling faults and improving their robustness due to their self-reconfigurable ability and homogeneity. However, the current self-repairing of modular self-reconfigurable robots requires backup modules to replace the faulty modules, and in extreme environments such as space, it is often impossible to provide backup modules to complete self-repairing. SUMMARY

[0003] The application provides a fault-tolerant control method for a modular self-reconfigurable robot to overcome the prior art. The fault-tolerant control method based on the PSO algorithm and the RRT algorithm can generate a reconfiguration plan and improve the fault tolerance and operation ability of the robot when a joint of a chain-type modular self-reconfigurable robot fails.

[0004] The fault-tolerant control method for a modular self-reconfigurable robot comprises the following steps:

[0005] I. Substructure division

[0006] After the modular self-reconfigurable robot detects the fault position and the fault angle, the Monte Carlo method is used to retrieve the number of reachable poses of the robot when the fault module is in different positions, the target position of the fault module is obtained according to the robot structure when the maximum number of reachable poses is obtained, and then all the modules are divided into three substructures S1, S2 and S3. Before the different substructures are separated, the modular robot needs to connect the end module to the magnetic connection surface to form a closed chain.

[0007] The total number of robot modules is N, the initial position of the fault module is P i , and the target position is P t . If P i is greater than P t , the sequence of the substructure S1 is [1:C t -1-sign(N+C i )], the sequence of the substructure S2 is [C t -sign(N+C i ):C i-1-sign(N+C i )] and the sequence of the substructure S3 is [C i -sign(N+C i ):(N-C i )+C t , the sequence of the substructure S1 is [1:C i -1-sign(N+C i )], the sequence of the substructure S2 is [C i -sign(N+C i ):(N-C t )+C i ] and the sequence of the substructure S3 is [(N-C t )+C i +1:N], and the expression of the function sign is:

[0008]

[0009] II. Reconstruction between substructures

[0010] First, the substructure S3 is connected with the base connection point B1, and then disconnected with the substructure S2;

[0011] Second, the substructure S2 is connected with the base connection point B2, and then disconnected with the substructure S1;

[0012] Third, the substructure S1 is connected with S3, and the substructure S3 is disconnected with the base connection point B1;

[0013] Fourth, the substructure S3 is connected with S2, and the substructure S2 is disconnected with the base connection point B2;

[0014] III. Determination of node configuration and motion path

[0015] According to the total number of modules, the population in the PSO algorithm and the nodes in the RRT algorithm are initialized, and the algorithm elements corresponding to the fault modules are always set to the fault state and do not participate in the search process;

[0016] According to the reconstruction process between substructures, the kinematic model corresponding to the combined substructure is established, and the pose error between the robot end and the target connection point is taken as the reference to determine the value of the fitness:

[0017]

[0018] wherein p, R are the position vector and attitude matrix of the robot end relative to the base under the combined substructure S, are the position vector and attitude matrix of the target point, and λ is a regulation factor for balancing the position error and the attitude error;

[0019] If a collision occurs or the fitness requirement is not met even after reaching the maximum number of iterations, the search fails. Within the specified maximum number of iterations, if the particle's fitness value meets the requirement and the modular robot does not collide, the configuration q of the combined substructure is determined. goal The initial configuration q of the modular robot start Defined as the initial node of the RRT algorithm, the configuration q obtained by PSO search is... goal Defined as the target node of the RRT algorithm, q rand q represents a random sampling point in joint space. near Represents distance q rand The most recent node, then in q rand and q near New nodes q can be extended from the connection lines. new If in q near Move to the new node q new If no collision occurs in the path, then q new Add it to the expanded tree; otherwise, q needs to be selected again. rand Continue iterative calculation until q new Reaching the target point completes the process, thus obtaining the motion path between the two key configurations, which is the first step in the reconstruction process. Then, q... goal q as the next iteration start PSO search yields q for the next iteration. goal Repeat the above process until the faulty module is reconstructed to the target location.

[0020] Furthermore, the position and orientation of the robot's end effector relative to the base are calculated using the DH method:

[0021]

[0022] in, These are the homogeneous transformation matrices of module 1's coordinate system relative to the base coordinate system and the robot's end effector relative to module m, respectively. Let be the homogeneous transformation matrix between adjacent modules, R be the pose matrix of the modular robot's end effector, and p be the position vector of the end effector. The joint angles of the modules are discrete values, i.e., q. i ∈[-90,0,90].

[0023] The advantages of this invention compared to the prior art are:

[0024] This invention proposes a fault-tolerant reconfiguration method that maximizes the robot's operational capabilities (i.e., the number of achievable poses) in response to joint lock-up failures in chain-modular self-reconfigurable robots.

[0025] According to the operation ability of the robot under the fault condition, the optimal configuration is determined, the reconfiguration problem is divided into a substructure sequence problem, and the PSO algorithm is further used to search the sub-configuration of the key node, and the RRT algorithm is used to generate a collision-free reconfiguration path. By changing the distribution position of the fault module, the operation ability of the modular self-reconfigurable robot is improved.

[0026] When a module occurs 0-degree lock failure, the remaining rate of reachable poses of the robot is low, and the operation ability sharply decreases. Through the fault-tolerant control method, the fault module can be transferred to another position, so that the number of reachable poses of the robot is restored to 100%.

[0027] The technical solutions of the present application will be further described below in combination with the drawings and embodiments: BRIEF DESCRIPTION OF DRAWINGS

[0028] Figure 1 The modular self-reconfigurable robot fault-tolerant control flowchart of the present application is shown in the figure.

[0029] Figure 2 The module diagram and configuration description schematic diagram in the embodiment are shown in the figure.

[0030] Figure 3 The motion path diagram of the fault module 8 from position 8 to position 6 in the embodiment is shown in the figure. DETAILED DESCRIPTION

[0031] The embodiments of the technical solutions of the present application will be described in detail below in combination with the drawings. Unless otherwise specified, the technical terms or scientific terms used in the present application should be the usual meanings understood by the skilled in the art to which the present application belongs.

[0032] In view of the deficiencies of the prior art, the modular self-reconfigurable robot fault-tolerant control method of the present embodiment is proposed. The method is based on the PSO and RRT algorithms.

[0033] The module adopted in the present embodiment is abstracted as a cubic shape, has four connecting surfaces, and has one rotational degree of freedom around the z-axis (see Figure 2 ), and different modules can be connected to each other. The module joint angle takes discrete values: -90 degrees, 0 degrees, and 90 degrees. The number of feasible directions of the modular robot end to reach the same grid position is used to represent the dexterity of the robot at the grid. The global number of reachable poses, i.e., the sum of the feasible directions of all reachable grids, is used to represent the operation ability of the robot in this configuration, and is also the optimization index of the algorithm.

[0034] Figure 1 The modular self-reconfigurable robot fault-tolerant control method of the present embodiment is shown, which includes the following steps:

[0035] I. Substructure division

[0036] After the modular self-reconfigurable robot detects the fault position and the fault angle, the number of reachable poses of the robot is retrieved when the fault module is in different positions by the Monte Carlo method, the target position of the fault module is obtained according to the robot structure when the maximum number of reachable poses is obtained, and all modules are divided into three sub-structures S1, S2 and S3. Before the different sub-structures are separated, the modular robot needs to connect the end module to the magnetic connection surface to form a closed chain.

[0037] The total number of robot modules is N, the initial position of the fault module is P i , and the target position is P t . If P i is greater than P t , the sequence of sub-structure S1 is [1:C t -1-sign(N+C i )], the sequence of sub-structure S2 is [C t -sign(N+C i ):C i -1-sign(N+C i )], and the sequence of sub-structure S3 is [C i -sign(N+C i ):N]; if P i is less than P t , the sequence of sub-structure S1 is [1:C i -1-sign(N+C i )], the sequence of sub-structure S2 is [C i -sign(N+C i ):(N-C t )+C i ], and the sequence of sub-structure S3 is [(N-C t )+C i +1:N], and the expression of the function sign is:

[0038]

[0039] II. Reconfiguration between sub-structures

[0040] First, sub-structure S3 is connected to base connection point B1, and then disconnected from sub-structure S2.

[0041] Second, sub-structure S2 is connected to base connection point B2, and then disconnected from sub-structure S1.

[0042] Third, sub-structure S1 is connected to S3, and sub-structure S3 is disconnected from base connection point B1.

[0043] Fourth step: Substructure S3 is connected to S2, and substructure S2 is disconnected from the base connection point B2;

[0044] At this point, the configuration of the modular robot is reconstructed from the initial configuration S1-S2-S3 to S1-S3-S2, realizing the position transfer of the faulty module and reducing the impact of the fault on the operational capability;

[0045] III. Determine the node configuration and motion path

[0046] Based on the determination of the substructure, the closed-loop configuration formed by the robot and the base is further determined, as well as the motion path between different configurations. In this step, the discrete particle swarm algorithm is used to determine the configuration connected to the connection surface during the reconstruction process.

[0047] Initialize the population in the PSO algorithm and the nodes in the RRT algorithm according to the total number of modules, and always set the algorithm elements corresponding to the faulty modules to the faulty state, so that they do not participate in the search process.

[0048] Following the reconstruction process between substructures, a kinematic model corresponding to the combined substructure is established, and the pose error between the robot's end effector and the target connection point is used as a reference to determine the fitness value:

[0049]

[0050] Where p,R are the position vector and attitude matrix of the robot end effector relative to the base under the combined substructure S. Let be the position vector and attitude matrix of the target point, and λ be the adjustment factor that balances the position error and attitude error;

[0051] If a collision occurs or the fitness requirement is not met even after reaching the maximum number of iterations, the search fails. Within the specified maximum number of iterations, if the particle's fitness value meets the requirement and the modular robot does not collide, the configuration q of the combined substructure is determined. goal The initial configuration q of the modular robot start Defined as the initial node of the RRT algorithm, the configuration q obtained by PSO search is... goal Defined as the target node of the RRT algorithm, q rand q represents a random sampling point in joint space. near Represents distance q rand The most recent node, then in q rand and q near New nodes q can be extended from the connection lines. new If in q near Move to the new node q new If no collision occurs in the path, then q new Add it to the expanded tree; otherwise, q needs to be selected again.rand , continue iteration until q new reaches the target point, thus obtaining the motion path between the two key configurations, i.e. completing the first step of the reconstruction process, and then q goal is taken as q start for the next iteration. goal The above process is repeated until the faulty module is reconstructed to the target position.

[0052] The method is aimed at the chain modular self-reconfigurable robot after encountering joint lock failure and without backup module supplement. First, according to the operation performance of the robot in different failure states, the target configuration is determined and the substructure is divided. Then, the configuration of the key node is searched based on the PSO algorithm, and the collision-free motion path between the intermediate configurations is generated based on the RRT algorithm. By optimizing the configuration of the remaining modules, the failure influence is reduced, the failure damage is eliminated or weakened, and the operation ability of the robot is ensured.

[0053] Specifically, the position vector and attitude matrix of the robot end relative to the base are calculated by the following method,

[0054] According to the D-H method, the position and attitude of the robot end relative to the base are calculated:

[0055]

[0056] wherein, and are the homogeneous transformation matrices of the coordinate system of module 1 relative to the base coordinate system and the robot end relative to module m, respectively, is the homogeneous transformation matrix between adjacent modules, R is the attitude matrix of the modular robot end, and p is the position vector of the end. The joint angle of the module takes discrete values, i.e. q i ∈[-90, 0, 90].

[0057] Specifically, the process of each step in the reconstruction process between substructures is described as follows:

[0058] First, the reconstruction process of connecting substructure S3 with base connection point B1 and then disconnecting with substructure S2 is as follows:

[0059] According to the kinematic model established by combining substructures S1-S2-S3, the position vector p 123 and the attitude matrix R 123 of the end are obtained. In addition, according to the position vector p b1 and the attitude matrix R b1 of the base connection point B1, the PSO algorithm is initialized according to the number of modules, the population size, the inertia weight w, the learning factors c1 and c2, and the maximum iteration step number, and the position of each particle qi For a configuration of the modular robot, the velocity v of the particle along the following formula i Update the particle position information:

[0060]

[0061]

[0062] Wherein, r1, r2 is a random number in [0,1], gbest represents the current optimal particle position of the population, pbest represents the current optimal particle position of the individual, t represents the number of generations of evolution; PSO search S1-S2-S3 corresponding to the configuration q goal .

[0063] The second step, the substructure S2 and the base B2 point connection, and then with the substructure S1 disconnected reconstruction process is:

[0064] The configuration q goal in the first step is converted to the initial configuration q start in this step, according to the combined substructure S1-S2 to establish the kinematic model, PSO algorithm searches with base connection point B2 as the target point, search for the next configuration q goal , using RRT algorithm search to get the second motion path, after the reconstruction of S1 and S2 is disconnected.

[0065] The third step, the substructure S1 and S3 connection, substructure S3 and base B1 point disconnected reconstruction process is:

[0066] The configuration q goal in the second step is converted to the initial configuration q start in this step, according to the combined substructure S1-S3 to establish the kinematic model, PSO search with base connection point B1 as the target point, search for the next configuration q goal , using RRT algorithm search to get the third motion path, after the reconstruction of S1 and S3 is connected.

[0067] The fourth step, the substructure S3 and S2 connection, S2 and base B2 point disconnected reconstruction process is:

[0068] The configuration q goal in the third step is converted to the initial configuration q start in this step, according to the combined substructure S1-S3-S2 to establish the kinematic model, PSO search with base connection point B2 as the target point, search for the next configuration q goal , using RRT algorithm search to get the fourth motion path, after the reconstruction of S3 and S2 is connected.

[0069] Embodiment

[0070] Based on the above embodiments,

[0071] In combination Figures 1-3 It is explained that a modular self-reconfigurable robot fault-tolerant control method based on PSO and RRT algorithm includes the following steps:

[0072] A chain-shaped modular mechanical arm composed of m modules, one face of module 1 is connected to the base, and the other face is connected to module 2. According to the D-H method, the position and attitude of the robot end relative to the base can be calculated:

[0073]

[0074] Wherein, The homogeneous transformation matrix of the coordinate system of module 1 relative to the base coordinate system, the homogeneous transformation matrix of the robot end relative to module m, The homogeneous transformation matrix between adjacent modules, R is the attitude matrix of the modular robot end, and p is the position vector of the end. The module joint angle takes discrete values, that is, q i ∈[-90, 0, 90].

[0075] Step 1: Substructure division

[0076] When the module fails, the operation ability of the modular robot decreases, and the distribution of different fault modules will also affect the operation ability. Therefore, the fault module is reconfigured from the initial position P i to the more optimal target position P t . All modules are divided into three substructures: S1, S2 and S3. Before separating different substructures, the modular robot needs to connect the end module to the magnetic connection surface to form a closed chain. The total number of robot modules is N, the initial position of the fault module is P i , and the target position is P t . If P i is greater than P t , the sequence of substructure S1 is [1:C t -1-sign(N+C i )], the sequence of substructure S2 is [C t -sign(N+C i ):C i -1-sign(N+C i )], and the sequence of substructure S3 is [C i -sign(N+C i ):N]; if P i is less than P t , the sequence of substructure S1 is [1:C i -1-sign(N+C i)], the sequence of substructure S2 is [C i -sign(N+C i ):(N-C t )+C i ], the sequence of substructure S3 is [(N-C t )+C i +1:N].

[0077] Step two: substructure S3 is connected with base connection point B1, and then disconnected with substructure S2.

[0078] According to the combined substructure S1-S2-S3, a kinematics model is established, and the position vector p 123 and the attitude matrix R 123 of the end are obtained. In addition, the position vector p b1 and the attitude matrix R b1 of the base connection point B1 are obtained.

[0079] According to the number of modules, the PSO algorithm is initialized, and the population number, inertia weight w, learning factors c1 and c2, and the maximum iteration step number of the PSO algorithm are set. Each particle position q i corresponds to a configuration of the modular robot, and the particle position information is updated along the particle velocity v i according to the following formula:

[0080]

[0081]

[0082] wherein r1 and r2 are random numbers in [0, 1], gbest represents the current optimal particle position of the population, pbest represents the current optimal particle position of the individual, t represents the evolution number, and the fitness is represented as:

[0083]

[0084] The configuration q goal corresponding to S1-S2-S3 is obtained by PSO search, and is defined as the target point of the RRT algorithm.

[0085] q rand represents a random sampling point in the joint space, q near represents the nearest node to q rand , and then a new node q rand is extended on the line connecting q near and q new . If no collision occurs in the path from q near to the new node q new , q newJoin to the extended tree, otherwise reselect q rand Continue iteration until q new reaches the target point, thus obtaining the initial configuration q start and the first segment of the motion path between q goal , after the reconstruction is completed, S3 is disconnected from S2.

[0086] Step three: substructure S2 is connected to base connection point B2, and then disconnected from S1

[0087] The configuration q goal in step two is converted into the initial configuration q start in this step, a kinematic model is established according to the combined substructure S1-S2, the PSO algorithm takes the base connection point B2 as the target point, and the next key configuration q goal is searched and obtained, and then the RRT algorithm is used to search the second segment of the motion path, and after the reconstruction is completed, S1 is disconnected from S2.

[0088] Step four: substructure S1 is connected to S3, and S3 is disconnected from base connection point B1

[0089] The configuration q goal in step three is converted into the initial configuration q start in this step, a kinematic model is established according to the combined substructure S1-S3, the PSO algorithm takes the base connection point B1 as the target point, and the next key configuration q goal is searched and obtained, and then the RRT algorithm is used to search the third segment of the motion path, and after the reconstruction is completed, the substructure S1 is connected to S3.

[0090] Step five: substructure S3 is connected to S2, and S2 is disconnected from base connection point B2.

[0091] The configuration q goal in step four is converted into the initial configuration q start in this step, a kinematic model is established according to the combined substructure S1-S3-S2, the PSO algorithm takes the base connection point B2 as the target point, and the next key configuration q goal is searched and obtained, and then the RRT algorithm is used to search the fourth segment of the motion path, and after the reconstruction is completed, S3 is connected to S2.

[0092] Taking a robot with 8 modules as an example, when module 8 has a 0-degree lock failure, the remaining rate of reachable poses of the robot (the ratio of the number of reachable poses after the failure to the number of reachable poses without failure) is only 69.23%, and the operation ability decreases sharply. The fault-tolerant control method of the modular self-reconfigurable robot proposed in the present application can move the faulty module to position 6, and the motion path of self-reconfiguration is shown in Figure 2 , so that the number of reachable poses of the robot is restored to 100%.

[0093] Although the present application has been described in connection with the preferred embodiment thereof with reference to the drawings, it will be apparent to those skilled in the art that various changes and modifications can be made thereto without departing from the scope of the present application.

Claims

1. A modular self-reconfigurable robot fault-tolerant control method, characterized in that: Includes the following steps: I. Divide into substructures After the modular self-reconfigurable robot detects the fault location and fault angle, it retrieves the number of robot poses that the fault module can reach at different positions using the Monte Carlo method. Based on the robot structure when the maximum number of reachable poses is obtained, the target position of the fault module is obtained. Then, combined with the initial position of the fault module, all modules are decomposed into three sub-structures: S1, S2, and S3. Before the different sub-structures are separated, the modular robot needs to connect the end module to the magnetic connection surface to form a closed chain. The total number of robot modules is N, and the initial position of the faulty module is P. i And the target location is P t If P i Greater than P t Then the sequence of substructure S1 is [1:C t -1-sign(N+C i The sequence of substructure S2 is [C] t -sign(N+C i ):C i -1-sign(N+C i The sequence of substructure S3 is [C] i -sign(N+C i ):N];If P i Less than P t The sequence of substructure S1 is [1:C i -1-sign(N+C i The sequence of substructure S2 is [C] i -sign(N+C i ):(NC t )+C i The sequence of substructure S3 is [(NC t )+C i +1:N], the expression for the function sign is: II. Restructuring between substructures First, substructure S3 is connected to base connection point B1, and then disconnected from substructure S2; The second step is to connect substructure S2 to base connection point B2, and then disconnect it from substructure S1. The third step is to connect substructure S1 and S3, and disconnect substructure S3 from the base connection point B1. Fourth step: Substructure S3 is connected to S2, and substructure S2 is disconnected from the base connection point B2; III. Determine the node configuration and motion path Initialize the population in the PSO algorithm and the nodes in the RRT algorithm according to the total number of modules, and always set the algorithm elements corresponding to the faulty modules to the faulty state, so that they do not participate in the search process. Following the reconstruction process between substructures, a kinematic model corresponding to the combined substructure is established, and the pose error between the robot's end effector and the target connection point is used as a reference to determine the fitness value: Where p and R are the position vector and attitude matrix of the robot end effector relative to the base under the combined substructure S. , Let be the position vector and attitude matrix of the target point, and λ be the adjustment factor that balances the position error and attitude error; If a collision occurs or the fitness requirement is not met even after reaching the maximum number of iterations, the search fails. Within the specified maximum number of iterations, if the particle's fitness value meets the requirement and the modular robot does not collide, the configuration q of the combined substructure is determined. goal The initial configuration q of the modular robot start Defined as the initial node of the RRT algorithm, the configuration q obtained by PSO search is... goal Defined as the target node of the RRT algorithm, q rand q represents a random sampling point in joint space. near Represents distance q rand The most recent node, then in q rand and q near New nodes q can be extended from the connection lines. new If in q near Move to the new node q new If no collision occurs in the path, then q new Add it to the expanded tree; otherwise, q needs to be selected again. rand Continue iterative calculation until q new Reaching the target point completes the process, thus obtaining the motion path between the two key configurations, which is the first step in the reconstruction process. Then, q... goal q as the next iteration start PSO search yields q for the next iteration. goal Repeat the above process until the faulty module is reconstructed to the target location.

2. The modular self-reconfigurable robot fault-tolerant control method according to claim 1, characterized in that: The position and orientation of the robot's end effector relative to the base are calculated using the DH method: in, , These are the homogeneous transformation matrices of module 1's coordinate system relative to the base coordinate system and the robot's end effector relative to module m, respectively. Let R be the homogeneous transformation matrix between adjacent modules, R be the pose matrix of the modular robot's end effector, p be the position vector of the end effector, and the module joint angles be discrete values. .

3. The modular self-reconfigurable robot fault-tolerant control method according to claim 1, characterized in that: The reconstruction process of substructure S3 being connected to base B1 and then disconnected from substructure S2 is as follows: A kinematic model is established based on the combined substructures S1-S2-S3 to obtain the end-effector position vector p. 123 Attitude matrix R 123 Additionally, based on the position vector p of the target point B1 of the base... b1 Attitude matrix R b1 ; Initialize the PSO algorithm based on the number of modules, setting the population size, inertia weight w, learning factors c1 and c2, maximum number of iterations, and position q for each particle. i For one configuration of a modular robot, according to the following formula, along the particle's velocity v i Update particle position information: Where r1 and r2 are random numbers in [0,1], gbest represents the current best particle position of the population, pbest represents the current best particle position of the individual, and t represents the generation number of the evolution; PSO search obtains the configuration q corresponding to S1-S2-S3. goal .

4. The modular self-reconfigurable robot fault-tolerant control method according to claim 3, characterized in that: Substructure S2 and The reconstruction process, which connects to base B2 and then disconnects from substructure S1, is as follows: The configuration q in the first step goal Convert to the initial configuration q in this step start A kinematic model is established based on the combined substructures S1-S2. The PSO search uses the base connection point B2 as the target point to obtain the next configuration q. goal The second motion path is obtained by using the RRT algorithm. After reconstruction, S1 and S2 are disconnected.

5. The modular self-reconfigurable robot fault-tolerant control method according to claim 4, characterized in that: Substructure S1 and The reconstruction process for connecting S3 and disconnecting substructure S3 from base B1 is as follows: The configuration q in step two goal Convert to the initial configuration q in this step start Motion is established based on the combined substructures S1-S3 The learning model uses the base connection point B1 as the target point to search for the next configuration q. goal The RRT algorithm is used to search for the third motion path, and after reconstruction, S1 and S3 are connected.

6. The modular self-reconfigurable robot fault-tolerant control method according to claim 5, characterized in that: The reconstruction process for substructure S3 connected to S2 and S2 disconnected from base B2 is as follows: The configuration q in step three goal Convert to the initial configuration q in this step start A kinematic model is established based on the combined substructures S1-S3-S2. The PSO search uses the base connection point B2 as the target point to obtain the next configuration q. goal The fourth motion path is obtained by using the RRT algorithm. After reconstruction, S3 and S2 are connected.

Images