Research On Path Tracking And Speed Decision Of Autonomous Driving Underground Mining Truck

The deep mining is developing as the national policy with the exploration of the underground mine increasing. The deep underground is an extremely high temperature, humidity, vibration and noise environment which is harmful to the health of the mining equipment operators. Thus the autonomous driving system has caused wide public concern. This paper focuses on path tracking control and speed decision of the articulated steering vehicle, especially the 30-tonne underground mining articulated dump truck, in order to enable path tracking in the tunnel and vibration suppression caused by terrain in the autonomous driving.First of all, the electrification upgrade has been finished on the real prototype. Meanwhile a small scale model prototype and a real-time virtual model for the Hardware-In-the-Loop simulation have been built. The same data acquisition system, communication system and main controller have been installed on both real and scale prototype for manual control, remote control and autonomous driving.After modelling, the errors between reference and real path have been defined. The linear relationship between speed and vertical vibration amplitude has been proved. From the system identification, the steering control model is pure delay proportional element, and the speed control model is pure delay first order inertial element.From the reactive navigation strategy, with the PID algorithm, the adaptive PID path tracking steering controller based on reinforcement learning algorithm has been designed. The inputs are path errors and the output is steering angle. The PID parameters can adjust adaptively in reinforcement learning online. Furthermore, use the driver preview model to derive the speed control strategy. The neuro-fuzzy network speed controller based on supervised learning algorithm has been designed in which the driver data is used as the training data for learning the parameters.Based on the linear relationship between the vertical vibration amplitude and the speed, the ideal speed for the ideal vertical vibration has been deduced. The speed decision algorithm has been designed with the constraint of the speed controller output, ideal speed and longitudinal acceleration. The parameters are learned by supervised learning algorithm with driver data. Define the state vector of the laser ranger point cloud for calculating the terrain roughness score function. And classify the binary terrain using the score with the self-supervised learning algorithm. The training data comes from the filtered data of the laser ranger and the inertial measurement unit online. The terrain roughness score is used to speed prediction for improving the speed decision algorithm.Finally, simulation and real prototype test has been run to evaluate the performance of the path tracking controller and speed decision algorithm. The result of the path tracking test shows that compared with the traditional PID with constant parameters, the adaptive PID steering controller can reduce the amplitude, mean value and variance of the displacement error, orientation error, curvature error and steering angle. And the neuro-fuzzy network speed controller output is similar to the driver purpose. The result of the speed decision test shows that the improved algorithm with speed prediction can reduce the high level shock events and amplitude on the rough terrain. Meanwhile on the smooth terrain the algorithm also can speed up. The conclusion is the path tracking controllers can control the autonomous driving underground mining truck effectively and the speed decision algorithm can reduce the vertical vibration caused by rough terrain. This research can provide a reference for the path tracking and speed decision on autonomous driving underground mining truck.