Research On Control And Coordination Of Multiple AUVs Based On Neural Network

In twenty-one century, resources in oceans will play a very important role in the development of human society. Autonomous Underwater Vehicle (AUV) is a very useful tool in exploring and utilizing resources in oceans. As the key technology, the research on how to control and coordinate multiple AUVs has important significance in theory and practice.Firstly, the application of neural network in the AUV's motion control is presented and discussed. The focus of research is on the optimization of neural network learning algorithm to increase the convergence velocity and the stability of learning. The experiment has verified that the neural network controller designed had good performance. The successful application of neural network controller in AUV's motion control has given a new way of controller design and will have positive effect in the research of intelligent control of AUV.Then the exploration of the distributed control of the multi-AUV system is made, in which we focused on the structure of multi-AUV team, the hybrid AUV control architecture and the behavior coordination between AUVs. The experiments have proved that a valid team control system can be achieved based on the hybrid architecture and well-defined behaviors, and better results by introducing Reinforcement Learning theory. New 6 DOF simulation model, restrictive condition modification, acoustic and non-acoustic sensor simulation and the MAS based framework make the new multi-AUV simulator more close to the real arena for the controller, which is a great promotion to the development ofAUV team.