新型机器人转动稳定性姿态最优控制

The Attitude Optimal Control of Rotational Stability of the New Robot

在新型机器人RRRobot(Rocking and Rolling Robot)姿态控制的研究中,由于新型机器人是一个典型的欠驱动系统,且处于高重心状态下造成系统姿态控制困难。针对上述问题,采用自适应动态规划方法解决新型机器人的姿态运动最优控制问题。根据系统模型,转化为自适应动态规划问题,选取状态变量和控制变量,采用BP神经网络构建执行网络和评价网络,并分别对执行网络和评价网络进行训练,给出了系统的效用函数,保证机器人在高重心状态下姿态达到期望位置。分别在单腿和双腿机动两种条件下进行仿真。仿真结果表明针对RRRobot的姿态调整控制问题,应用自适应动态规划方法计算精度高,能够快速的完成控制,粒子群算法和遗传算法验证方法的有效性、可行性,为RRRobot的后续研究提供了重要的试验基础。

In the research of the attitude control of the new rocking and rolling robot(RRRobot),it is difficult to control the attitude of the system because the new robot is a typical underactuated system and it is in the state of high gravity.In view of the above problems,the approximate dynamic programming method is used to solve the optimal control problem of the attitude movement of the new robot.According to the system model,it is transformed into the problem of approximate dynamic programming.The state variables and control variables are selected;the BP neural network is used to construct the action network and critic network.Then,the action network and critic network are trained respectively.The utility function of the system is given to ensure that the robot posture achieves the desired position in the state of high gravity.The simulation is performed on single leg and both legs respectively.The simulation results show that the approximate dynamic programming method has high precision and can complete the control quickly for the attitude adjustment control problem of the RRRobot,the validity and feasibility of the method are verified by the particle swarm optimization and genetic algorithm,which provides an important experimental basis for the follow-up study of the RRRobot.