Hybrid Modeling of Velocity Obstacle Method in Autonomous Collision Avoidance System for Unmanned Aerial Vehicles

This paper presents the modeling works of autonomous collision avoidance system (ACAS) of unmanned aerial vehicles (UAV). We took and extend the so-called Velocity Obstacle method as the base of the avoidance algorithm, due to its simplicity on handling dynamic environment. The outputs of the method are used to conduct avoidance, which is set as combinations between three modes, i.e., (1) avoid, (2) maintain, and (3) restore. The interactions between the continuous vehicles kinematics and the discrete decision making are described by a hybrid system model. A hybrid automaton is constructed, mapping criteria required for mode changes of the system, including when collisions will occur. To demonstrate algorithm and model performance, several simulations were conducted. This includes scenario where there is just one vehicle avoiding, and where all involved vehicles are avoiding. All simulations were conducted smoothly, resulting in zero collision, even for multiple encounters. Several concluding remarks are presented as a baseline for the research continuation.