To ensure that a robot can maintain a stable contact force when it contacts the environment,an adaptive PID control for robot constant force tracking was designed based on a dueling deep Q-network.The contact process between a robot and the environment was analyzed,and a robot force controller based on a PID algorithm was constructed.The adaptive PID algorithm based on a dueling deep Q-network was proposed to adapt to changes in the external environment.In this algorithm,the dueling deep Q network was used to learn and find optimal control parameters.Finally,robot constant force tracking experiments were expanded on Coopeliasim and MAT-LAB software platforms.The simulation results show that the adaptive PID algorithm based on the dueling deep Q-network can achieve good force-tracking effects,verifying the algorithm's feasibility;compared with a deep Q network algorithm,the average absolute value of the force error is reduced by 51.6％,and the convergence speed is improved,allowing the robot to track the external environment better.
关键词
机器人/恒力控制/自适应PID控制/对决深度Q网络
Key words
robot/contact force control/adaptive PID control/dueling deep Q-network
维普
万方数据