Predictive Guidance for Automated Velocity Obstacle Collision Avoidance

As the use of unmanned systems becomes more prevalent in both commercial and military applications, increasing performance requirements have led to a greater demand for automation. An autonomous unmanned system's ability to perform basic tasks reliably reduces the operator's cognitive tasks and workload, which could facilitate the control of multiple systems by a single operator. A key component for autonomous systems in many applications is the ability to perform reliable collision avoidance. This paper presents a predictive velocity obstacle approach integrated into the Automated Velocity Obstacle Collision Avoidance (AVOCA) algorithm. AVOCA is a multi-agent velocity obstacle based collision avoidance system that includes Kinematic Velocity Constraints (KVCs) to select feasible collision free velocities in a computationally efficient manner. The predictive velocity obstacle algorithm serves as a near-term forecasting component, with the goal of enabling the algorithm to make better decisions based on the trend of changes in the problem space. Results from simulation testing of the predictive component are presented, which are focused on the overall performance enhancement and parameter sensitivity.