Abstract
A decentralized control system is proposed for the control architecture of an autonomous underwater vehicle (AUV). The software concept is based on the object-oriented model and its specialization, a “agent oriented model”. Such control system was implemented on a network of transputers. First tests of such control system were related to the collision avoidance guidance for an AUV which follows the sea bed profile. The guidance system combines a pursuit guidance algorithm and a grid map built in real-time, which is based on the information acquired from a sonar system. First tests concerned simulations using a model of the vehicle Pteroa-150. Experiments were also carried out using the carriage system of a model basin adapted to a test bed for underwater robotics. A robotic system was achieved by using an automatic pursuing system aimed originally to experiments with free running ship models, and attaching underwater acoustic sensors to one of the carriages. Thus, automatic guidance in collision avoidance experiments was made possible for investigating the response of the cooperative control system when it faces the input of a real sensory system. Computer simulations and experimental tests have shown the usefulness and feasibility in real-time of the map assisted guidance system.
