AGV视觉导航中Kalman滤波最优控制器的设计

Optimal Controller Design of Kalman-Filtering in Visual Navigation for Automatic Guided Vehicle

为解决电子地图辅助视觉导航AGV的控制精度问题,将AGV中基于视觉的位置信号离散化后得到AGV中心轴的偏距和偏角.作为滤波器的输入信号,用U-D分解的Kalman滤波可以有效地消除在AGV导航中存在的白噪声.其中在时域中采用递推算法为最优控制输入稳定的状态量,并保证数值的稳定性,使系统的鲁棒性好;再经过状态反馈最优控制系统得到最优的控制输出.结果表明,用此方法设计的控制器稳定性强,抗干扰能力强,AGV在该控制器的作用下,转弯半径为5m时,方向偏差在±0.5°内变化,侧向偏差控制在±4mm之内.该控制器适应于工业环境中强电气噪声的环境,对直线和曲线路径有很好的跟踪效果,用于自动导向车辆等智能运输设备,符合工业应用的要求.

In order to solve the problem of accuracy control of automatic guided vehicle（AGV） equipped with visual navigation system, the vision-based continuous state equation was turned into a discrete state equation to obtain distance deviation and angular deviation from the center axis, both of which were used as input signal of the filters. Then the U-D decomposed Kalman-filtering could effectively eliminate the white noise existing in the AGV navigation system. Recursive algorithm in time domain was adopted to provide steady state variables for the optimal controller and guarantee the computational stability to achieve better system robustness. Optimal output of the optimal control system could be obtained by status feedback. The results indicate that the robustness and noise immunity of the controller designed with the proposed method are strong. By the proposed controller, the orientation deviation of AGV ranges between ±0.5°and the lateral deviation ranges between ±4 mm, while the swerve radius is 5 m. Consequently the controller works properly under the industrial circumstances with large electric noise and can trace the linear and curve paths effectively. It is used in intelligent transportation equipment such as automatic guided vehicle and can meet the requirement of industrial application.