Design And Implementation Of Multi Degree-of-Freedom Robot Motion Planning System

With the advancement of artificial intelligence,a number of industrial robots are integrated into the production process.Mankind is liberated from heavy and repetitive labor by the robots.As the number of robots increases and the demands for production flexibility increase,robots are facing greater collision risks and more calculation during motion planning.In this thesis,a logistics sorting-and-stacking assembly line is taken as a scene.Two six-degree-of-freedom robotic arms are used as the research object.A multi-degree-of-freedom robot real-time motion planning system is designed based on a sampling-based algorithm.For obstacles formed by palletizing,a static-palletizing-collision-detection accelerator is designed.The resolution of collision detection is improved by decomposing the space and using the occupied grids to encode the paths.Parallel detection is used to improve the execution speed.For dual-arm collision avoidance and collaboration,a dynamic-path-collision-detection accelerator is designed.Collision detection is implemented by model calculation based on the bounding box method.Parallel computing is used to improve detection speed.The static-palletizing-collision-detection accelerator is optimized by compressing the encoding area,selecting the Cartesian coordinate to encode,and performing asymmetric encoding.The random access memory is used to save the encoding table.In this way,the accelerator can be deployed and implemented.The applicable configuration and the bus delay of the functional subunits in the dynamic-path-collision-detection accelerator are analyzed.The constraints are proposed according to the hardware platform.A hardware-resource / execution-time multi-objective optimization problem is constructed.The configurations that meet the requirements are obtained by solving the multi-objective optimization problem.The motion control and motion planning of the two robotic arms are implemented by the robot controller.By comparing the two collision-detection accelerator solutions,it is proved that the designed static-palletizing-collision-detection accelerator can save hardware resources and ensure control performance;the designed dynamic-path-collision-detection accelerator is optimal in hardware-resource / execution-time.Functional simulation and physical system tests on the proposed scheme are carried out.The simulation and test results show that the two robotic arms can complete the dynamic motion planning and collision detection smoothly.The feasibility and effectiveness of the multi-degree-of-freedom robot motion planning system are proved.