Vibration Suppression Of Pine Cone Picking Device Based On QL-SI Algorithm

The pine cone picking device is prone to uncontrollable residual vibration during the movement due to the coupling movement of multiple joints and the excessively long striker.Residual vibration will lead to reduced pine cone maturity recognition,pine cone positioning accuracy,and stability of the picking device control system.The percussion pine cone picking device designed for the laboratory is an experimental platform.The traditional SI input shaping algorithm can effectively suppress vibration,but the adjustment time is too long to meet the purpose of rapid positioning.Therefore,the traditional SI input shaping algorithm is combined with reinforcement learning to suppress the residual vibration of the picking manipulator and improve the robustness of the pine cone picking device.First,the dynamic model of the pine cone picking device is introduced and the mathematical model of the flexible joint is established.The kinematic equation of the pine cone picking device is established by the DH parameter method,which is combined with the flexible joint to form the dynamic equation of the picking device with flexible joint.Use forward kinematics to analyze the movement of the picking manipulator,and inverse kinematics to solve the dynamic model of the pine cone picking device.Since the double-pendulum system is a typical multi-section control system with complex and multiple system parameters,testing with the double-pendulum system can compare the vibration suppression of each algorithm in the presence of multiple system parameters.After that,four input shapers of ZV,ZVD,El and SI were used to control the double pendulum system,and the residual vibration suppression effect of each algorithm was observed and compared and analyzed.Finally,aiming at the problems of the high position of the pine cones,the change of the length of the hitting manipulator,the coupling of multiple joints of the picking device,and the inability to establish a precise mathematical model,an SI input shaping vibration suppression strategy based on reinforcement learning is proposed.The design process is as follows:1).Under the trade-off between the amount of calculation and the damping effect,5 state decision tables are selected to express the action of SI shaping at 5 equal time intervals.In each table,the coordinates of the hit position of the robotic arm are taken as the abscissa,and the speed of the stepping motor is taken as the ordinate.The movement is selected by Boltzmann,and the calculated Q value is stored in the table;2).The vibration amplitude measured by the posture sensor of the mechanical arm,Design the reward function and the weight of the reward function;3).Compare the reward value and the reward threshold in each state to continuously optimize the behavior strategy,and obtain the QL-SI algorithm parameters through multiple trainings.The optimal Q table is obtained through the time minimization and the principle of synthetic reward optimization and optimal parameters.This paper uses a double pendulum system and a striking pine cone picking platform to verify the effectiveness and feasibility of the control system of the SI input shaper and the QL-SI algorithm.The QL-SI algorithm is added to the double pendulum model to prove its effectiveness.The stability time of the QL-SI algorithm is about 2.3ms.Compared with the SI input shaping algorithm,the stability time is reduced by 90%,and the vibration amplitude is reduced by 60%.The experimental results show that the QL-SI algorithm can suppress the vibration of complex motion when the parameter range is known.The QL-SI algorithm is transplanted to the percussion pine cone picking device to prove its feasibility.The stability time of the QL-SI algorithm is about 2.1ms.Compared with the SI input shaping algorithm,the stability time is reduced by 87%,and the vibration amplitude is reduced by 41%.The results show that the algorithm can effectively suppress the vibration of the picking manipulator under the condition of unknown parameters and has good robustness.