视觉与GPS/IMU融合的智轨电车循迹控制研究

Automatic Tracking Control of Autonomous-rail Rapid Tram Based on Camera Vision, Inertial Measurement Unit and GPS Data

视觉识别技术易受地面标识污损、强光反射及视场角小等因素影响,而全球定位系统(GPS)则易受遮挡与电磁干扰等因素影响,使用任意单一定位方式均会出现循迹可用性不高的问题。对此,文章提出一种融合视觉与GPS组合惯导的容错控制方法。其首先基于图像识别方法对车道线进行识别,得到车辆与道路的几何关系参数后,设计PID控制器进行转向控制;接着,采用实时动态全球定位系统(RTK-GPS)构建分层电子地图,以输出平滑参考轨迹,进而设计Stanely控制器进行转向控制;最后,对视觉识别技术与GPS循迹控制进行融合。实车试验结果表明,采用所提出的融合控制策略,有效提高了智轨电车小半径转弯通过率,弯道最高通过速度提高了66.7%(可达25 km/h),单程接管次数减少了83.8%,运营时间缩短了24.4%,系统可用性得到有效提升。

Visual recognition is easily affected by broken ground identification, reflection of strong light and narrow field of view, while GPS(global positioning system) location is susceptible to occlusion and electromagnetic interference. Using any single positioning method will cause the problem of low tracking availability. Therefore, a fault-tolerant control method of combined vision and GPS combined inertial navigation is proposed in this paper. Firstly, lane line recognition is carried out based on image recognition method, geometric relationship parameters between vehicle and road are obtained, and PID controller is designed for steering control. Secondly, layered electronic map is constructed by RTK-GPS(real-time kinematic GPS), smooth reference track is output, and then Stanley controller is designed for steering control. Finally, vision and GPS tracking control are fused. Through real vehicle verification, the results show that the proposed fusion control strategy effectively improves the small radius turning passing rate of autonomous-rail rapid tram, the maximum passing speed can reach 25 km/h, which is increased by 66.7%, the number of oneway takeovers is reduced by 83.8%, the operation time is reduced by 24.4%, and the system availability is effectively improved.