Improving UAV Obstacle Avoidance by Environment Dependent Cost for Path Planning

Most local path planning algorithms rely on several cost functions representing different aspects of the planned path, e.g., path smoothness and deviation from goal, to evaluate alternative flight paths and define the next waypoint(s). The cost functions are weighed against each other by fixed or user predefined weighting parameters. Previous investigations showed that an optimal choice of weights differs significantly between various environments. This work presents a method to automatically choose weighting parameters for flight velocity, minimum obstacle distance, yaw, and pitch cost depending on the environment. The presented strategies are formulated such that they fit various path planning algorithms. The performance of the proposed method is tested with the 3DVFH* in the px4-Avoidance implementation. The adaptions show significant improvement in failure probability for environments with small to medium sized, rectangular obstacles and environments containing different kinds of forests. The failure probability of city-like environments is very sensitive to the correct implementation of the avoidance strategies. However, a proper choice leads to a reduction of failure probability as well.