Modeling, Optimation And Control Of Reconfigurable Modular Robot

To meet different task requirements, a Reconfigurable Modular Robot (RMR) isdesigned, which can be assembled into different configurations by some modules with thesame connection interfaces. Compared with traditional robots, RMR is more flexible andadaptable for different environments. It also can simplify the design, manufacturing andmaintenance, shorten the development cycle and reduce development costs, etc. It’s worthyand practical to research on the key technologies of RMR, such as kinematics, dynamics, typesynthesis, dimensional synthesis, error and control, etc. The works of this paper areintroduced as follows:(1) Mathematical modelling: A method to generate the kinematics and dynamics modularautomatically is proposed. The configuration of RMR is described as a tree structure or anassembly incidence matrix in assembly-level. Kinematics and dynamics models can begenerated automatically according to different configuration. A task-oriented method, whichcombined with genetic algorithm and iterative method, is proposed to solve inversekinematics problem for different task and configurations of RMR.(2) Configuration and dimension design: For solving the problem of finding suitablerobot configurations in application of RMR, a task-oriented multi-objective configurationoptimization method is proposed. This method is applicable to the configuration design ofRMR, which take into account performance optimization of degrees-of-freedom, reachability,energy consumption, and so on. Dimension design is to establish an optimal model by resizethe link size based on a fixed configuration. Using the normalization method, multipleperformance metrics are combined into a single dimensionless objective function value in themodel, then, the optimal solution is given by the global optimization algorithm. The modelcan the modular robot design, and has a strong versatility for the dimension optimization ofgeneral serial robot.(3) Joint surface modeling: The structure continuity and overall performance of robotlink is destroyed by the fixed joint surface between the adjacent modules. To analyze theinfluence on the accuracy from the joint surface between the modules of RMR, a relationalmodel between force and deformation error of module joint surface is established based onfractal theory. Unlike the contact model of general mechanical joint surface under the uniformexternal load, this model is about non-uniform load. The static and dynamic performance ofRMR can be analysed by joint surface error model. (4) Control system: a RMR control system with open and distributed structure isproposed. Via CAN bus communication, the efficiency of data transfer between the controlmodules is improved. Using the modular design method, the proposed control system is open,short in development and reconfigurable. The control system not only meets the traditionalgeneral control requirements of teach/playback, but also meets the intelligence requirementsof autonomous/semi-autonomous. A practical spatial trajectory planning algorithm ispresented in the section of motion planning, which can achieve a good transition between thebasic trajectory segments and smooth motion. This simple and practical algorithm can ensurethe continuity of the trajectory acceleration and effectively reduce the impact. Orientation isinterpolated using unit quaternion to avoid the gimbal lock. Finally, experiments are done onthe manipulator, pole-climbing and wall-climbing robot, verify the reasonableness of thecontrol system and the correctness of the key algorithms.