安防巡检机器人变预瞄距离LQR横向控制

LQR lateral control of security inspection robots with variable preview distance

摘要

针对安防巡检机器人动力学参数难以测量、运动控制精度低的问题,提出一种基于动力学模型的自适应变预瞄横向控制算法。首先,建立阿克曼转向运动学及二自由度侧向动力学模型。针对模型中未知或难以测量的动力学参数,设计基于M序列的系统辨识实验,以前轮转角为输入、横摆角速度为输出,拟合得到机器人动力学模型传递函数,拟合度达95.87%,解决了参数难以准确获取的难题。其次,基于动力学模型提出变预瞄距离线性二次型调节器(linear quadratic regulator,LQR)控制方法,引入前馈控制消除道路曲率引起的稳态误差;为进一步提升自适应能力,采用模糊控制方法,根据当前车速与道路曲率动态调节预瞄距离。结果表明,自适应预瞄LQR控制算法最大横向跟踪误差控制在±5 cm以内,相较于无预瞄方法跟踪精度提高51%,相较于固定预瞄方法提高26%,有效提升了安防巡检机器人在复杂道路条件下的轨迹跟踪性能。

Abstract

To address the challenges of difficult-to-measure dynamic parameters and low motion control accuracy in security inspection robots, an adaptive preview lateral control algorithm based on a dynamic model is proposed. First, the Ackermann steering kinematic model and a two-degree-of-freedom lateral dynamic model are established. To handle unknown or difficult-to-measure parameters, an M-sequence-based system identification experiment is designed. Using the front wheel steering angle as input and the yaw rate as output, the transfer function of the robot′s dynamic model is obtained through fitting, achieving a goodness of fit of 95.87%. Second, a variable preview distance linear quadratic regulator (LQR) control method is proposed, with feedforward control to eliminate steady-state errors caused by road curvature. Additionally, a fuzzy control method is employed to dynamically adjust the preview distance according to vehicle speed and road curvature. The results show that the proposed algorithm maintains a maximum lateral tracking error within ±5 cm. Compared to the non-preview method, tracking accuracy is improved by 51%, and compared to the fixed preview method, by 26%, effectively enhancing trajectory tracking performance under complex road conditions.

关键词

安防巡检机器人 / 系统辨识 / 运动控制 / LQR / 变预瞄距离

Key words

security inspection robot / system identification / motion control / LQR / variable preview distance