A hierarchical planning method and system for reconfiguration of a heterogeneous modular robot

Abstract

The application relates to a hierarchical planning method and system for heterogeneous modular robot reconfiguration, wherein initial and target configurations of a heterogeneous modular robot system are given; in a high-level planner, a heuristic search algorithm is adopted to search in a discrete configuration space of the heterogeneous modular robot system, so that a macro operation sequence of module repositioning from the initial configuration to the target configuration is obtained; in a low-level planner, for each macro operation in the macro operation sequence, an optimal execution trajectory composed of a series of atomic actions is planned for an assembly robot, and the cost of the atomic action execution trajectory is fed back to the high-level planner; and the macro operation sequence and the atomic action execution trajectory are integrated to generate a robot reconfiguration plan with the optimal global cost. Compared with the prior art, the application is significantly superior to the prior art in planning success rate, efficiency and scheme quality, and provides a specific and feasible engineering approach for realizing a large-scale and functional self-reconfiguration robot system.

Description

Technical Field

[0001] This invention relates to the field of robotics and automated construction, and in particular to an automated reconfiguration planning method and system for heterogeneous modular robot systems. Background Technology

[0002] Automated construction is a long-term goal in the field of robotics, aiming to achieve unmanned assembly and maintenance of structures in extreme environments. However, the ultimate vision of this field goes far beyond building passive houses or bridges; it is to realize large-scale functional systems capable of sensing, responding, and acting—that is, to achieve the automated construction of robotic systems themselves. Modular robotics technology offers a promising approach to realizing this vision, allowing for the creation of morphologically adaptive and functionally diverse robotic systems by combining different basic modules.

[0003] Currently, methods for assembling modular robots can be mainly divided into two categories: self-reconfiguration methods and external assembly methods. Self-reconfiguration methods embed drive and intelligent units into each independent module, completing reconfiguration through autonomous collaboration between modules. However, such methods typically result in excessively high complexity, cost, and weight for individual modules. Their inherent high complexity makes them difficult to apply to large-scale systems requiring numerous modules, rendering them economically and physically infeasible.

[0004] External assembly methods employ specialized assembly robots to manipulate relatively simple modules to construct large structures. Existing research has demonstrated the potential of this method in building large passive structures, such as building skeletons. However, most of these systems overlook a more critical subsequent challenge: how to effectively integrate heterogeneous modules with "muscle" and "nerve" functions, such as joints and wheels, into the structure to build a truly functional robot. In highly heterogeneous scenarios containing various types of modules, this integration process presents significant combinatorial challenges, a gap that current technologies have not effectively addressed.

[0005] The key to solving the aforementioned construction challenges lies in the restructuring planning algorithm itself. The planning space for robot restructuring tasks is a high-dimensional hybrid space spanning discrete module layouts and continuous robot motions. Directly searching this space leads to combinatorial explosion, resulting in low planning efficiency or even an inability to solve the problem. Existing planning algorithms suffer from severe limitations in scalability and robustness when faced with large-scale (numerous modules) and highly heterogeneous (diverse module types) tasks.

[0006] Therefore, how to efficiently solve the reconfiguration planning problem of large-scale, highly heterogeneous modular robot systems in order to achieve automated and low-cost conversion from the initial configuration to the target configuration remains a major technical challenge in this field. Summary of the Invention

[0007] The purpose of this invention is to solve the technical problems of combinatorial explosion, low planning efficiency and poor scalability in existing modular robot reconfiguration planning methods when facing large-scale, highly heterogeneous tasks, and to provide a hierarchical planning method and system for heterogeneous modular robot reconfiguration.

[0008] The objective of this invention can be achieved through the following technical solutions: As a first aspect of the present invention, a hierarchical planning method for reconstructing heterogeneous modular robots is provided, comprising the following steps: Given the initial and target configurations of a heterogeneous modular robot system, a hierarchical planning framework is adopted, in which a high-level planner performs discrete task planning and a low-level planner performs continuous motion planning. In the high-level planner, a heuristic search algorithm is used to search in the discrete configuration space of the heterogeneous modular robot system to obtain the module relocation macro-operation sequence that transforms the initial configuration into the target configuration. In the lower-level planner, for each macro operation in the macro operation sequence, a cost-optimal execution trajectory consisting of a series of atomic actions is planned for the assembly robot, and the cost of the atomic action execution trajectory is fed back to the higher-level planner. The high-level planner uses the trajectory cost returned by the low-level planner as the actual execution cost of the macro operation and updates the search queue. By integrating macro operation sequences with atomic action execution trajectories, a robot reconfiguration plan that is globally cost-optimal and physically executable is generated.

[0009] As a preferred technical solution, the high-level planner adopts a bidirectional heuristic search algorithm, which searches simultaneously from the initial configuration and the target configuration until the intersection point is found.

[0010] As a preferred technical solution, the heuristic search algorithm uses a heuristic function calculated based on the Hungarian algorithm or a greedy algorithm to estimate the minimum number of relocations required for the current configuration to reach the target configuration.

[0011] As a preferred technical solution, the heuristic function used by the heuristic search algorithm includes a type penalty term for heterogeneous modules, which is used to assign unequal weights to modules of different types during the search process.

[0012] As a preferred technical solution, when generating a macro operation sequence, the high-level planner performs a validity check on each macro operation. The check includes: ensuring that the picked module is a leaf node in the current configuration diagram, ensuring that the picking operation does not disrupt the overall connectivity of the robot structure, and ensuring that the placement position of the module is within the kinematic reach of the assembly robot.

[0013] As a preferred technical solution, the underlying planner uses a search algorithm to search in the state space of the assembly robot in order to plan the execution trajectory of atomic actions; The state space is defined by the position of the assembly robot's supporting legs and whether the assembly robot holds a module; The atomic actions include: movement of the assembly robot on the surface of the assembled structure, rotation around the support legs of the assembly robot, and picking up and placing of modules.

[0014] As a preferred technical solution, the lower-level planner assigns a preset cost value to each atomic action, and the sum of the preset costs of each atomic action constituting the execution trajectory is used as the cost of the atomic action execution trajectory; the cost of the execution trajectory is used as the actual execution cost of the macro operation sequence and fed back to the higher-level planner to guide the higher-level planner to perform the global cost-optimal path search.

[0015] As a second aspect of the present invention, a hierarchical planning system for the reconfiguration of heterogeneous modular robots is provided, the system executing the hierarchical planning method for the reconfiguration of heterogeneous modular robots as described above, including: The high-level planning module receives the initial configuration and the target configuration, and uses a bidirectional heuristic search algorithm with a type penalty term to generate a series of module relocation macro operations; The lower-level planning module, connected to the higher-level planning module, is used to plan the most cost-effective atomic motion execution trajectory for the robot using a search algorithm for the macro operation, and feeds back the cost of the trajectory to the higher-level planning module to assist it in searching for the globally optimal path.

[0016] As a preferred technical solution, the high-level planning module in the high-level planning module is used to receive the initial configuration and the target configuration, and to generate a series of module relocation macro operations by using a bidirectional heuristic search algorithm that includes a type penalty term. The bottom-level planning module, connected to the high-level planning module, is used to plan the most cost-effective atomic motion execution trajectory for the robot using a search algorithm for the macro operation, and feeds back the cost of the trajectory to the high-level planning module to assist it in searching for the globally optimal path. The high-level planning module receives the initial configuration and the target configuration, and uses a bidirectional heuristic search algorithm with a type penalty term to generate a series of module relocation macro operations; The lower-level planning module, connected to the higher-level planning module, is used to plan the most cost-effective atomic motion execution trajectory for the robot using a search algorithm for the macro operation, and feeds back the cost of the trajectory to the higher-level planning module to assist it in searching for the globally optimal path.

[0017] As a preferred technical solution, the underlying planning module assigns a preset cost value to each atomic action, and the sum of the preset costs of each atomic action constituting the execution trajectory is used as the cost of the atomic action execution trajectory; the cost of the execution trajectory is used as the actual execution cost of the macro operation sequence and fed back to the high-level planner to guide the high-level planner to perform the global cost-optimal path search.

[0018] Compared with the prior art, the present invention has the following beneficial effects: 1) This invention decouples the high-dimensional hybrid programming problem into two low-dimensional subproblems: high-level discrete task planning and low-level continuous motion planning. It also employs an efficient bidirectional heuristic search, which greatly reduces computational complexity and enables it to quickly solve large-scale reconstruction tasks involving dozens of modules.

[0019] 2) This invention introduces a type penalty term in the heuristic function, which effectively solves the symmetry problem caused by the large number of module types, enabling the planner to robustly handle complex refactoring tasks containing multiple different functional modules, with a success rate far higher than methods without this penalty term.

[0020] 3) Economic efficiency of the planning scheme: This invention calculates the accurate physical execution cost through the bottom-level planner and feeds it back to the high-level planner to ensure that the final generated scheme is not only feasible, but also more economical and optimized in terms of physical execution time, path length or energy consumption.

[0021] This invention significantly outperforms existing technologies in terms of planning success rate, efficiency, and solution quality, providing a concrete and feasible engineering approach for realizing large-scale, functional self-reconfigurable robot systems. Attached Figure Description

[0022] Figure 1 This is a flowchart of a hierarchical planning method for the reconfiguration of a heterogeneous modular robot according to the present invention.

[0023] Figure 2 This is a schematic diagram illustrating the implementation process of an embodiment of the present invention.

[0024] Figure 3 The following are schematic diagrams of the assembly robot of the present invention: a) a three-dimensional view of the assembly robot, b) a schematic diagram of the degrees of freedom of the assembly robot, and c) a schematic diagram of the handling module of the assembly robot.

[0025] Figure 4 These are the planning results under different reconstruction tasks of this invention. Detailed Implementation

[0026] The present invention will now be described in detail with reference to the accompanying drawings and specific embodiments. These embodiments are based on the technical solution of the present invention and provide detailed implementation methods and specific operating procedures. However, the scope of protection of the present invention is not limited to the following embodiments.

[0027] Example 1 This invention utilizes a hierarchical planning framework to decompose complex reconfiguration tasks into two sub-problems: high-level discrete task planning and low-level continuous motion planning. By introducing a type penalty term for heterogeneous modules in the high-level planning and calculating accurate physical execution costs in the low-level planning, it achieves efficient, robust, and economical reconfiguration path planning.

[0028] like Figure 1 As shown, the hierarchical planning method for reconfiguration of heterogeneous modular robots according to an embodiment of the present invention includes the following steps: Step 1: Given the initial configuration of the heterogeneous modular robot system and target configuration The configuration is represented by a graph structure, where nodes represent modules and edges represent connections between modules. Each module has position, orientation, and type attributes (e.g., basic structure module, joint module, wheel module, etc.). A hierarchical planning framework is adopted to decouple high-level discrete task planning from low-level continuous motion planning.

[0029] Step 2: In the high-level planner, a bidirectional heuristic search algorithm is used to search in the discrete configuration space of the heterogeneous modular robot system. This algorithm simultaneously searches from the initial configuration... and target configuration The starting point is to find a sequence of module relocation macro operations that connects the two.

[0030] Step 3: For each candidate macrooperation in the macrooperation sequence generated by the high-level planner during the search process, the low-level planner is invoked. The low-level planner uses the A* search algorithm to calculate the optimal execution trajectory, consisting of a series of atomic actions, required for the assembly robot to execute the macrooperation, and returns the cost of that trajectory. .

[0031] Step 4: The high-level planner calculates the trajectory cost returned by the low-level planner. This serves as the actual execution cost of the macro operation, and is used to update its search queue.

[0032] Step 5: Repeat steps 2 to 4 until the forward and backward searches of the high-level planner find an intersection. Finally, integrate the macro-operation sequence found by the high-level planner and the atomic motion trajectories calculated by the low-level planner for each macro-operation to output a complete, globally cost-optimal, and physically executable robot reconfiguration plan.

[0033] In step 2, the high-level planner employs a bidirectional heuristic search, the core of which lies in an efficient heuristic function. Used to evaluate the current configuration With target configuration The estimated distance between them, the heuristic function in this embodiment example The calculation process is as follows:

[0034] in, Representing different module types; For weighting coefficients, functional modules Typically assigned a higher level than the basic module Higher weighting is given to ensure that core functions are in place first; Represents the current configuration set of modules of the same type. With the target configuration set The sum of the minimum Manhattan distances between them; Indicates type of penalty.

[0035] First, by comparing the current configuration and target configuration It identifies all misaligned modules that are not in the correct position.

[0036] Then, the Hungarian algorithm or a greedy algorithm is used to estimate the minimum number of relocations required for the current configuration to reach the target configuration, and the minimum allocation cost required to move these misaligned modules to their target positions is used as a base heuristic.

[0037] To address the symmetry issues arising from heterogeneous modules, different weights are assigned to different types of modules when calculating allocation costs. For example, the misalignment cost weight for functional modules (such as joints and wheels) is higher than that for ordinary structural modules. This allows the planner to prioritize critical functional modules, effectively breaking the symmetry in the search process and preventing the planner from making ineffective explorations among multiple similar options.

[0038] To address the deadlock problem where the position is correct but the type is incorrect, this invention introduces a type penalty term to penalize situations where a critical target position is occupied but the type does not match.

[0039] The specific calculation logic is as follows: traverse all key positions in the target configuration. If a certain position in the current configuration... If a module is already in use, but the type of that module is not equal to the type required by the target configuration at that location, an additional penalty will be applied.

[0040] Quantification value: In a specific embodiment, the penalty weight is set to a constant of 2, that is:

[0041] in, This indicates a location shared by the current configuration and the target configuration, meaning the current module already occupies a certain position in the target. Indicates the position in the current configuration Module type at the location; Indicates the position in the target configuration Required module type. This setting allows the boot planner to prioritize removing modules of incorrect types from critical positions.

[0042] The high-level planner generates macro operation sequences based on heuristics as follows: Action generation: The high-level planner traverses the current configuration, identifies all legal leaf node modules (which can be picked up) and their reachable free positions (which can be placed), and generates a series of candidate macro operations. .

[0043] Heuristic evaluation: for performing any macro operation The new configuration that emerged later Calculate its overall score:

[0044] in, The estimated execution cost (usually set to a default value, such as 1, initially); To calculate the remaining cost of the distance to the target using the aforementioned heterogeneous heuristic function.

[0045] Sequence Enqueue: All candidate macro operations are enqueued based on a comprehensive score. The value is stored in the search queue, and macro-operation branches with smaller heuristic values ​​(closer to the target) are expanded first. When generating the macro-operation sequence, the high-level planner performs a validity check on each macro-operation, which includes: (1) Leaf node picking verification: The graph traversal algorithm is used to ensure that the picked module is a leaf node in the current configuration connection graph to prevent damage to the main structure during the picking process.

[0046] (2) Connectivity maintenance check: Ensure that the picking operation does not cause the robot’s main structure to split into two or more disconnected parts.

[0047] (3) Reachability placement verification: Ensure that the target placement position of the module is within the kinematic reach of the assembly robot.

[0048] An action is considered a valid edge only if the moved module is a leaf node in the current configuration graph (removing it does not break the overall connectivity) and its placement satisfies physical reachability. This improved verification significantly reduces the search space by pruning a large number of physically infeasible states.

[0049] In step 3, the underlying planner employs the A* search algorithm to search the state space of the assembly robot to plan the execution trajectory of atomic actions. Considering the robot's characteristics of containing different functional modules such as the base, joints, and wheels, this invention decouples the single distance evaluation in the traditional A* algorithm into type-based matching, calculating the misalignment cost for each type of module separately, and introducing a type penalty term to avoid the algorithm getting trapped in local optima (e.g., moving a joint module to the target position of the wheel). This embodiment combines... Figure 2 It is a small six-legged assembly robot that plans its movement trajectory.

[0050] like Figure 3 As shown, the state space is defined by the position of the assembly robot's supporting legs, its body posture, and whether its end effector holds a module. The atomic actions of the assembly robot include: (a) movement: moving one step forward, backward, left, or right on the assembled surface; (b) rotation: rotating in place around any supporting leg; and (c) pick-up and place: extending or retracting the end effector to pick up or place a module.

[0051] Each atomic action is assigned a preset cost value; for example, movement cost is 1, rotation cost is 1, and pick / place cost is 2. The total cost of a single atomic action trajectory is... This is the sum of the costs of all atomic actions that constitute this trajectory. This cost... It accurately reflects the physical cost of executing a sequence of macro operations and is fed back to the higher-level planner to guide it in searching for the globally cost-optimal path, ensuring that the final refactoring plan is the most efficient at the physical execution level.

[0052] The search queue is updated based on underlying feedback, as follows: Triggering lower-level planning: When the higher-level planner is ready to expand a certain optimal node in the search queue (macro operation) When ), the underlying planner is invoked; Calculating the true cost: The underlying planner uses the A* algorithm to plan a specific sequence of actions (movement, rotation, picking, placement) in a continuous space and calculates the precise physical cost of the trajectory. (For example: number of steps moved × 1 + number of rotations × 1 + number of pickup actions × 2).

[0053] Update queue: High-level planner receives physical cost This cost is then used as the actual execution cost of the macro operation, replacing the original estimated execution cost. The total cost of updating this execution trajectory. .

[0054] Reordering: Utilizing the total cost of the corrected execution trajectory Recalculate the overall score The algorithm updates the queue. If a macro operation is logically only one step but has a high physical execution cost (e.g., requires detours), its priority is reduced, and the planner will explore other physically more economical paths. This invention changes the traditional assumption that the number of steps equals the cost in high-level search. The high-level planner no longer uses only the logical number of steps as the execution cost. G Instead of calculating the physical cost, the algorithm updates the path cost using the actual physical execution cost (including movement distance, number of rotations, and other physical energy consumption) fed back by the underlying planner. This changes the algorithm's optimality criterion from minimizing the number of transports to minimizing the physical execution cost.

[0055] Example 2 As a second aspect of the present invention, this embodiment provides a hierarchical planning system for reconstructing heterogeneous modular robots that performs the above method embodiments, the hierarchical planning system comprising: The high-level planning module receives the initial configuration and the target configuration, and uses a bidirectional heuristic search algorithm with a type penalty term to generate a series of module relocation macro operations; The lower-level planning module, connected to the higher-level planning module, is used to plan the most cost-effective atomic motion execution trajectory for the robot using a search algorithm for the macro operation, and feeds back the cost of the trajectory to the higher-level planning module to assist it in searching for the globally optimal path.

[0056] Figure 4 The embodiments of the present invention demonstrate the planning results of the hierarchical planning system under four different reconstruction tasks.

[0057] In this embodiment, the types of modules are distinguished by color: red squares represent basic structure modules, blue squares represent joint modules, and green squares represent special function modules such as wheels. The small gray robot in the figure is an assembly robot that performs picking, moving, and placing operations.

[0058] Each row of this figure illustrates a separate reconstruction task, with the initial configuration on the left and the target configuration planned by the method of this invention on the right. The first row shows a reconstruction task from a ring structure to a compact structure.

[0059] The second line illustrates how a simple discrete structure can be reconstructed into a functional mobile platform with four wheel modules (green). This demonstrates the invention's ability to handle highly heterogeneous modules (structural and functional modules).

[0060] The third line shows how an initial "I"-shaped configuration was reconstructed into a more complex symmetrical airfoil structure.

[0061] The fourth line shows a more complex reconstruction task, from an asymmetrical structure to a humanoid structure with functional modules.

[0062] These embodiments collectively demonstrate that the hierarchical planning method and system proposed in this invention can efficiently and robustly solve the problem of large-scale, highly heterogeneous modular robot reconfiguration, successfully planning a physically executable path from the initial configuration to the functional target configuration, and has significant advantages in planning success rate and solution quality.

[0063] The above description represents the preferred embodiments of this application. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of this application, and these improvements and modifications are also considered within the scope of protection of this application. For example, the high-level planner can be replaced with other forward search algorithms, and the low-level planner can be adapted to different types of assembly robots.

[0064] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0065] The preferred embodiments of the present invention have been described in detail above. It should be understood that those skilled in the art can make numerous modifications and variations based on the concept of the present invention without creative effort. Therefore, all technical solutions that can be obtained by those skilled in the art based on the concept of the present invention through logical analysis, reasoning, or limited experimentation on the basis of existing technology should be within the scope of protection defined by the claims.

Claims

1. A hierarchical planning method for the reconfiguration of heterogeneous modular robots, characterized in that, The steps include: Given the initial and target configurations of a heterogeneous modular robot system, a hierarchical planning framework is adopted, in which a high-level planner performs discrete task planning and a low-level planner performs continuous motion planning. In the high-level planner, a heuristic search algorithm is used to search in the discrete configuration space of the heterogeneous modular robot system to obtain the module relocation macro-operation sequence that transforms the initial configuration into the target configuration. In the lower-level planner, for each macro operation in the macro operation sequence, a cost-optimal execution trajectory consisting of a series of atomic actions is planned for the assembly robot, and the cost of the atomic action execution trajectory is fed back to the higher-level planner. The high-level planner uses the trajectory cost returned by the low-level planner as the actual execution cost of the macro operation and updates the search queue. By integrating macro operation sequences with atomic action execution trajectories, a robot reconfiguration plan that is globally cost-optimal and physically executable is generated.

2. The hierarchical planning method for reconfiguration of heterogeneous modular robots according to claim 1, characterized in that, The high-level planner employs a bidirectional heuristic search algorithm, simultaneously searching from the initial configuration and the target configuration until the intersection point is found; the heuristic function of the bidirectional heuristic search algorithm includes basic allocation cost and type penalty term; The basic allocation cost is the sum of the minimum Manhattan distances between the current configuration and the target configuration; The type penalty term is used to penalize cases where a critical target position is occupied but the type does not match; iterate through all critical positions in the target configuration, and if a critical position in the current configuration is occupied by a module, but the type of that module is not equal to the type required by the target configuration at that critical position, then apply an additional penalty.

3. The hierarchical planning method for reconfiguration of heterogeneous modular robots according to claim 2, characterized in that, When calculating the allocation cost, the heuristic function assigns different weights to different types of modules, with functional modules being given a higher weight than basic modules.

4. The hierarchical planning method for reconfiguration of heterogeneous modular robots according to claim 2, characterized in that, The high-level planner generates a sequence of macro operations based on heuristic values ​​as follows: The high-level planner traverses the current configuration, identifies all pickable leaf node modules and their reachable free locations, and generates a series of candidate macro operations; For the new configuration generated after performing macro operations The remaining allocation cost to the target configuration after executing macro operations is calculated using heuristic functions and added to the estimated execution cost to obtain a comprehensive score. All candidate macro operations are stored in the search queue according to their comprehensive score, and the macro operation branches with smaller heuristic function values ​​are expanded first.

5. The hierarchical planning method for reconfiguration of heterogeneous modular robots according to claim 4, characterized in that, When generating a sequence of macro operations, the high-level planner performs a validity check on each macro operation. The check includes: ensuring that the picked module is a leaf node in the current configuration diagram, ensuring that the picking operation does not disrupt the overall connectivity of the robot structure, and ensuring that the placement of the module is within the kinematic reach of the assembly robot.

6. The hierarchical planning method for reconfiguration of heterogeneous modular robots according to claim 4, characterized in that, When the higher-level planner is ready to expand a certain optimal node in the search queue, it calls the lower-level planner. The lower-level planner uses a search algorithm to plan a sequence of specific actions in a continuous space. The lower-level planner assigns a preset cost value to each atomic action, and the sum of the preset costs of the atomic actions that constitute the execution trajectory is used as the cost of the atomic action execution trajectory, and is fed back to the higher-level planner. The high-level planner receives the cost of the atomic action execution trajectory as the actual execution cost of the macro operation sequence, replaces the estimated execution cost, and updates the total cost of the execution trajectory; The overall score of the configuration is recalculated using the corrected and updated total cost of the execution trajectory, the search queue is updated, and a global optimal path is searched.

7. The hierarchical planning method for reconfiguration of heterogeneous modular robots according to claim 1, characterized in that, The underlying planner uses a search algorithm to search in the state space of the assembly robot to plan the execution trajectory of atomic actions; the state space is defined by the position of the support foot of the assembly robot and whether the assembly robot holds a module; the atomic actions include: the movement of the assembly robot on the surface of the assembled structure, the rotation around the support foot of the assembly robot, and the picking and placing of modules.

8. A hierarchical planning system for the reconfiguration of heterogeneous modular robots, characterized in that, The system executes the hierarchical planning method for heterogeneous modular robot reconfiguration as described in any one of claims 1-7, including: The high-level planning module receives the initial configuration and the target configuration, and uses a bidirectional heuristic search algorithm with a type penalty term to generate a series of module relocation macro operations; The lower-level planning module, connected to the higher-level planning module, is used to plan the most cost-effective atomic motion execution trajectory for the robot using a search algorithm for the macro operation, and feeds back the cost of the trajectory to the higher-level planning module to assist it in searching for the globally optimal path.

9. A hierarchical planning system for reconfiguration of heterogeneous modular robots according to claim 8, characterized in that, The high-level planning module is used to receive the initial configuration and the target configuration, and to generate a series of module relocation macro operations using a bidirectional heuristic search algorithm that includes a type penalty term. The bottom-level planning module, connected to the high-level planning module, is used to plan the most cost-effective atomic motion execution trajectory for the robot using a search algorithm for the macro operation, and feeds back the cost of the trajectory to the high-level planning module to assist it in searching for the globally optimal path. The high-level planning module receives the initial configuration and the target configuration, and uses a bidirectional heuristic search algorithm with a type penalty term to generate a series of module relocation macro operations; The lower-level planning module, connected to the higher-level planning module, is used to plan the most cost-effective atomic motion execution trajectory for the robot using a search algorithm for the macro operation, and feeds back the cost of the trajectory to the higher-level planning module to assist it in searching for the globally optimal path.

10. A hierarchical planning system for heterogeneous modular robot reconfiguration according to claim 8, characterized in that, The underlying planning module assigns a preset cost value to each atomic action, and the sum of the preset costs of each atomic action that constitutes the execution trajectory is used as the cost of the atomic action execution trajectory. The cost of the execution trajectory is used as the actual execution cost of the macro operation sequence and fed back to the high-level planner to guide the high-level planner in searching for the globally cost-optimal path.

Images