Pursuit and evasion in a recursive nested behavioral control structure for unmanned aerial vehicles

This work tackles the problem of integrating a pursuit-evasion game (PEG) in a recursive nested behavioral control (RNBC) structure for unmanned aerial vehicles (UAV). The UAV agents' architecture is the RNBC structure, considering that the collision avoidance, velocity control and attitude control are already implemented behaviors in this structure. The new high-level behaviors pursuit and evasion are introduced, describing the PEG and providing an appropriate reference velocity vector for the underlying layers of the RNBC structure of the pursuer and the evader, respectively. The reference velocities for the agents are determined by solving the pursuit-evasion problem in a game-theoretical manner, while collision avoidance behavior is implemented one level below. The repulsion force approach is used as collision avoidance technique and provides a velocity vector, which is added to the reference velocity vector of the upper level, if there is a collision risk. The resulting reference input is processed by the subjacent layers in the RNBC structure of the UAV. In simulations it is shown that this approach provides a collision-free pursuit-evasion between UAV agents with dynamic constraints in a general three-dimensional environment with previously unknown obstacles.