Overtaking Stationary and Moving Obstacles for Autonomous Ground Vehicles

Abstract

In this paper, a method for overtaking stationary and moving obstacles will be introduced. The method consists of designing a desired trajectory for lateral motion of the vehicle and then using a lateral motion controller for tracking this desired trajectory. The desired trajectory is a sigmoid exponential function of relative distance between the vehicle and the obstacle and guarantees overtaking the obstacle, if tracked exactly, despite of lateral and longitudinal motions of the obstacle. Lateral acceleration of the vehicle should not exceed safety limits during tracking desired trajectory. This matter has been used as a decision criterion for determining feasible and unfeasible desired trajectories. A neural network has been trained for predicting maximum lateral acceleration (MLA) during overtaking maneuver. The lateral motion controller is a sliding mode controller which has been designed to be robust to uncertainties existing in lateral dynamic model of the vehicle. A virtual test environment has been developed as a design tool for developing new control strategies for autonomous vehicles. The lateral controller has been tested extensively using this virtual test environment and has shown satisfactory performance in controlling the vehicle, even in existence of noises and disturbances.