Unmanned Aerial Vehicle Optimal Cooperative Obstacle Avoidance in a Stochastic Dynamic Environment

Effective target-tracking and obstacle avoidance strategies are essential to the success of unmanned aerial vehicle missions. This paper presents an extended Kalman-filter-based algorithm that predicts the optimal three-dimensional trajectory and position prediction error of a dynamic object (obstacle or target) detected by an unmanned aerial vehicle. This trajectory prediction scheme is thereafter tested in a three-dimensional path planner for multiple unmanned aerial vehicles, which relies on decentralized model-based predictive control to calculate the optimal unmanned aerial vehicle setpoints that will lead each unmanned aerial vehicle to the interception of a single dynamic ellipsoidal target while avoiding dynamic ellipsoidal obstacles detected en route. A novel model-based predictive control collision avoidance algorithm is also presented in this paper. The method first computes the unmanned aerial vehicle collision probability with an obstacle by convolving the statistical distribution of the obstacle center of mass position with the obstacle shape. The method then seeks to minimize the unmanned aerial vehicle collision probability with all known obstacles on a future horizon, all while ensuring that the collision probability with any given obstacle at each prediction step does not surpass a preset threshold. Simulations are presented to demonstrate the effectiveness of the proposed approach.