Research On Decision Optimization,Planning And Control Of Modular Reconfigurable Robot Reconfiguration

Modular reconfigurable robots,as an important type of variant robots,are a combination of multiple modules with independent perception,movement,and docking functions that are assembled in a modular manner.They can change the topology of the combination according to environmental characteristics and task requirements.The reconfiguration of modular robots is a key ability to adapt to complex environments and diverse tasks.How to achieve fast,efficient,and low energy consumption is an important issue that urgently needs to be solved to leverage their application advantages.In response to the technical requirements for efficient and reliable reconfiguration of modular robots,combined with the current development status of modular reconfigurable robots and the technological achievements in the field of multi-robot collaborative planning and control,this paper presents the multi-objective reconfiguration decision method,the parallel transformation optimal trajectory planning method in a shared workspace,the path planning methods in high dynamic environments and the distributed parallel variant control methods.The main content and contribution of the research are as follows.1.In response to the energy consumption and efficiency issues in the reconfiguration process of modular reconfigurable robots,three optimization objectives were summarized,such as the minimum number of modules for transformation movement,the minimum number of disconnected docking mechanisms,and the shortest total movement distance of all modules.A multi-objective reconfiguration decision optimization method based on hierarchical sequence was proposed,which can search for Pareto optimal reconfiguration strategies.Firstly,the problem of coupling multiple optimization objectives was solved through the hierarchical sequence method.A sliding overlap testing algorithm was designed to match the maximum common sub-configuration between the initial configuration and the target configuration,minimizing the number of modules that need to be moved.Secondly,a component splitting tree was constructed with the molecular configuration,i.e.component,as the node,and a heuristic cost function was designed.The heuristic search was used in the tree to obtain the splitting scheme with the least number of disconnected connections;Then,a visual graph was constructed based on the maximum common sub-configuration vertices to quickly calculate the total distance of allosteric movement.Finally,a grouping reconfiguration optimization strategy was proposed to address the dimension explosion problem of large-scale reconfiguration decision-making.2.In response to the conflicts of multi-component modules and the assembly order of target configuration in the process of reconfiguration,a spatiotemporal parallel reconfiguration optimization trajectory planning method is proposed to minimize reconfiguration time consumption,which provides efficient and safe motion trajectories for component parallel reconfiguration.Firstly,the multi-component motion conflicts in Euclidean space were mapped to the obstacles in the path-time space,and the motion of components along the path was mapped to the time-table curve,achieving the description of the time series of component motion along the optimal path.Then,an improved particle swarm optimization algorithm was proposed to optimize the time-table curve for each component and generate a collision-free placement order.Finally,an optimal simultaneous spatiotemporal parallel reconfiguration trajectory planning method was developed to solve the conflict problem of multiple components during parallel reconfiguration motion,allowing the components to move along the optimal path to the target configuration position in the shortest time without collision.3.Aiming at the high dynamic problem of reconfigurable robots with multi-component parallel reconfiguration,focusing on the dynamic environment response ability of component individuals during reconfiguration motion,a decentralized full map random tree and loop branch path search method is proposed,which improves the navigation ability of components in high dynamic environments.Firstly,a full map random tree and cyclic connection graph with omnidirectional consistency,decentralization,and uniform distribution were constructed to form a ringed tree topology.Secondly,an iterative optimization search strategy was proposed,which continuously shortens the path of the entire map random tree planning using cyclic branches to obtain the near-optimal path.Then,a dynamic environment response mechanism was established to prune,reconnect,and regrow the random tree and cyclic branches of the entire map when obstacles move,ensuring that the tree nodes maintain connectivity,effectiveness,and universality at all times.Finally,for the problem of components being surrounded by obstacles,a gradient perception model-based bounding escape method was proposed,which achieved rapid identification of bounding states and rapid search of escape channels.The full map random tree and cyclic branch search methods can discover global nearoptimal paths while maintaining real-time planning frame rates,improving the overall conformational robustness and dynamic response ability of modular reconfigurable robots.4.In response to the difficulties in trajectory tracking and dynamic obstacle avoidance control of heterogeneous component mixed clusters in reconfigurable robots,inspired by the overall behavior generation mechanism of biological clusters,a distributed model predictive controller based on local consensus was designed,achieving efficient and reliable parallel reconfiguration of modular robots.Firstly,the mechanism of fish movement and avoidance of predators was analyzed,and trajectory tracking,collision avoidance,and motion matching rules were established for each module.Then,trajectory tracking was performed based on module trajectory deviation,and collision soft constraints were established to limit the distance between modules.Urgency and motion-matching soft constraints were defined to achieve the matching of module motion trends to the individuals with the highest urgency.Finally,a behavior-based stepper control method was proposed to achieve high-precision docking control between modules.The distributed model predictive controller based on local consensus has recursive feasibility and asymptotic stability.Through independent and autonomous control of each module,it achieves overall trajectory tracking and dynamic obstacle avoidance control of heterogeneous components,as well as collision free parallel operation control of heterogeneous component mixed clusters,improving the efficiency of modular robot reconfiguration.This article has conducted a detailed theoretical analysis,simulations,physical prototype experiments,and comparative tests to verify the effectiveness of the proposed methods,providing a technical foundation for the parallel and efficient reconfiguration of mobile modular reconfigurable robots.