Obstacle avoidance for a non-holonomic vehicle using occupancy grids

In this paper, we outline the strategy used for obstacle avoidance on our small, non-holonomic test vehicle, the Autonomous Tractor. This strategy relies on the fusion of data from a stereo camera, a scanning laser range-finder and the vehicle odometry to create occupancy grids which describe the traversability of the terrain in the current vicinity of the vehicle. Knowledge of the vehicle’s kinematics and its response to control inputs is then used to derive obstacle free paths, if they exist, where preference is given to those commands which are ‘closest’ to those issued by the overarching navigation system.