An Ethology-Based Hybrid Control Architecture for an Autonomous Underwater Vehicle for Performing Multiple Tasks

In this paper, we propose a control architecture for an autonomous underwater vehicle (AUV), implemented in a hybrid architecture with two layers: a hierarchical planning layer and a reactive execution layer. Most of its tasks are interpreted as a set of waypoints and then specified actions at the points. Thus, in the planning layer, the task planner is designed as a waypoint planner using a genetic algorithm. This planner generates an optimized plan, considering given constraints such as positions of obstacles, current velocities, and task priority. In addition, the execution of the task plan is monitored by a mission supervisor, which determines ongoing tasks and can change the original plan if exceptional events occur. In the execution layer, a behavior-based control with an ethology-based action selection mechanism is implemented. As a result, the AUV can always choose the most appropriate behavior, maximizing its motivation, and the robot is controlled by the output of the selected behavior. Consequently, the proposed control architecture has an open and modular structure. Numerical simulations were conducted to verify its performance.