基于神经网络的移动机器人轨迹预测控制

Trajectory predictive control of mobile robot based on neural network

针对移动机器人在未知离散障碍物环境下感知、导航控制的问题,设计了一种基于神经网络与模型预测的控制系统。在感知阶段,采用传统视觉算法提取十二维度离散障碍物特征,并将其传入神经网络以控制机器人避障感知;在规划阶段,基于处理后的感知地图设计JPS全局规划器和MPC局部规划器,实现机器人自主导航。对该系统进行Simulink/CarSim联合仿真,并搭建移动机器人在室外环境测试。结果表明,仿真及测试过程中机器人感知信息准确可靠,导航轨迹效果稳定。

Aiming at the problem of perception and navigation control of mobile robot in unknown discrete obstacle environment, a control system based on neural network and model predictive control was designed. In the perception stage, the traditional vision algorithm was used to extract the features of the 12-d discrete obstacle and the feature was transferred to the neural network to control the obstacle avoidance perception of the robot;In the planning stage, JPS global planner and MPC local planner were designed based on the processed perceptual map to realize the autonomous navigation of the robot. The system was simulated by Simulink/CarSim, and the mobile robot was built to test in the real outdoor environment. The results show that the sensing information is accurate and reliable, and the navigation trajectory is stable during the simulation and test.