Research On Multi-robot Capturing The Moving Object For Server-board Assembly Process

In the future,there will be a large number of application scenarios for robots to capture moving objects in the server-board assembly link.While smart factories will make sure highly automated production,robots that can capture moving objects will effectively increase production efficiency and enhance the versatility of production processes for different working conditions and environments,thereby effectively enhancing the intelligence of future manufacturing.Based on the server-board assembly process,in order to enable multi-robot to coordinately capture the moving server-board,the thesis studies the problem of grasping the moving object with multi-robot cooperation.In the thesis,aiming at the scenario where human carries objects close to the shared working space of multiple robots and then transfer the object to the multi-robot,an algorithm is proposed for multi-robot to cooperatively capture moving objects to finish the collaborative grasping operations.The algorithm mainly includes functions of object motion trajectory prediction,robot dynamics modeling,and design of cooperative grasping controllers.In this thesis,the LSTM(Long-Short Term Memory)recurrent neural network is used to model the motion of the object,which aims to make the trajectory prediction of the object to obtain the robot’s cooperative grasping pose and make sure that the robots can perform the pre-grasping operations;then based on the Lagrange equation The inverse dynamics model of the robot is established,and a structured dynamic neural network is used to identify the robot’s inverse dynamic model parameters based on the prior knowledge of the inverse dynamics model.Finally,based on the dynamic model of the robot,a feedback controller is designed with backstepping method in the thesis.The controller generates the expected angular velocity of the robot’s joints and the driving torque of the robot’s joints to track the grasping pose generated by the LSTM network,and then drives multiple robots to grasp the moving object,The steady-state and transient performance requirements in the grasping process are guaranteed as well.At last,a hardware platform is designed for the proposed algorithm which includes the robot’s ripper,the controller of the platform and the software architecture of the algorithm.In the thesis,the proposed algorithm is applied to a scenario where a target object is carried by a human and then passed to a dual-arm robot,then cooperative grasping operations are performed by the dual robots.Simulation analysis and experimental tests of each module of the algorithm are conducted which proves that the the proposed algorithm can capture the object effectively.Also the proposed algorithm can be used in the server-board cooperatively grasping.