基于神经网络的轮式移动机器人非线性模型预测控制研究

Research on nonlinear model predictive control of wheeled mobile robots based on neural networks

针对轮式移动机器人受到障碍物干扰导致运动轨迹跟踪误差较大问题,提出了神经网络非线性模型预测(NN-NMP)控制方法,并对轮式移动机器人避障效果进行仿真验证。建立了轮式移动机器人运动模型,并且定义了机器人运动方程式。设计非线性模型预测控制方法,引用神经网络算法,通过训练多层神经网络对非线性模型预测控制误差进行逼近,从而使轮式移动机器人控制系统具有更好的避障能力。设置3种不同环境,利用Matlab软件对轮式移动机器人避障结果进行仿真,对比和分析轮式移动机器人采用2种控制方法的输出结果。结果显示:在无障碍物环境中,采用传统比例-积分-微分(PID)控制方法和NN-NMP控制方法,轮式移动机器人均能较好地按照期望轨迹进行移动。在有障碍物环境中,采用传统PID控制方法,轮式移动机器人虽然能够躲避障碍物,但是跟踪误差较大。采用NN-NMP控制方法,轮式移动机器人不仅能够躲避障碍物,而且跟踪误差相对较小。采用NN-NMP控制方法,能够降低轮式移动机器人跟踪误差,具有较好的避障能力。

A neural network nonlinear model prediction(NN-NMP)control method is designed to address the issue of large motion trajectory tracking errors caused by obstacles interfering with wheeled mobile robots.The obstacle avoidance effect of wheeled mobile robots is simulated and verified.A motion model for wheeled mobile robots is presented,and the robot motion equations were defined.We have designed a nonlinear model predictive control method,citing neural network algorithms,and trained multi-layer neural networks to approximate the nonlinear model predictive control error,thereby improving the obstacle avoidance ability of the wheeled mobile robot control system.Set up three different environments and use Matlab software to simulate the obstacle avoidance results of wheeled mobile robots.Compare and analyze the output results of two control methods used by wheeled mobile robots.The results show that in an obstacle free environment,both the traditional PID control method and the NN-NMP control method can effectively move the wheeled mobile robot according to the expected trajectory.In environments with obstacles,using traditional proportional-integral-derivative(PID)control methods,wheeled mobile robots can avoid obstacles,but their tracking error is relatively large.By using the NN-NMP control method,wheeled mobile robots can not only avoid obstacles,but also have relatively small tracking errors.The use of NN-NMP control method can reduce the tracking error of wheeled mobile robots and have good obstacle avoidance ability.