Semi-autonomous underwater vehicles for shallow water mine-clearing

The authors present an intelligent control architecture which promises to be useful for controlling a semi-autonomous undersea robot for shallow water mine hunting. The architecture integrates reaction plans and layered competences, providing for reactive behaviors and goal-oriented sequencing of tasks. Control theoretic techniques can be smoothly integrated into the architecture, providing for robust transit and hovering with reactive obstacle avoidance. The architecture could be used in an underwater task to attach detonators to moored mines. Included in this architecture were selective perception routines for high-frequency forward scanning sonars and vision processing from a low-light charge-coupled device (CCD) camera. The search routines involved are designed to be interrupted by survival behaviors and subsequently resume processing in the new situation. The reaction plan allows the sonar search, tracking, and vision processing to be retried whenever the task environment disrupts the normal flow of events