无人驾驶矿用运输车辆感知及控制方法

针对当前矿区生产作业效率低下、安全事故频发等问题，提出了一种矿用运输车辆无人驾驶感知及控制方法。感知部分，设计出基于激光雷达和毫米波雷达融合的多目标识别架构，在数据关联的基础上，应用基于卡尔曼滤波的联合概率数据关联（JPDA）算法实现矿区环境下多目标识别；控制部分，采用路径预瞄-跟踪的方式，将横向控制与纵向控制进行解耦，并通过反馈机制实时进行偏差修正，实现无人驾驶矿用运输车辆精准的横向与纵向控制。此外，搭建了矿车无人驾驶系统平台，在矿区不同场景下对上述感知及控制方法进行了测试。实验结果表明，感知算法能够实现道路可行驶区域的精确检测，并可识别出多种障碍物类型，控制算法在上下坡等场景下可实现无人驾驶矿用运输车辆纵向速度和横向位置的精准控制，满足实际应用需求。 

Perception and control method of driverless mining vehicle

In order to solve the problems of low production efficiency and frequent safety accidents in mining areas, a driverless perception and control method for mining vehicles was proposed. In the part of perception, a multi-target recognition architecture based on the fusion of lidar and millimeter-wave radar was designed. On the basis of data association, the joint probabilistic data association (JPDA) algorithm based on Kalman filter was applied to realize multi-target recognition in mining environment. In the control part, the lateral control and longitudinal control were decoupled by the way of path preview tracking, and the deviation was corrected in real time through the feedback mechanism to realize the accurate lateral and longitudinal control of the driverless mining vehicle. In addition, the driverless system platform of real mine vehicle was built, and the above perception and control methods were tested in different scenarios in the mining area. The experimental results show that the perception algorithm realize the accurate detection of the drivable area of the mining road, and identify a variety of obstacle types. The control algorithm realize the accurate control of the longitudinal speed and lateral position of driverless mining vehicles in uphill and downhill scenarios, so as to meet the of practical applications.