An adaptive controller system using mnSOM

Underwater vehicles are expected to be attractive tools in the deep ocean. In order to realize useful and practical robots, underwater vehicles should take their action by judging the changing condition from their own sensors and actuators, and are desirable to make their behavior by themselves in the hazardous working environment. We have investigated the application of neural networks into the Autonomous Underwater Vehicles (AUVs). AUVs have complicated non-linear dynamics in six degrees of freedom. In our previous adaptive control method using NNs, the information of initial states is getting lost gradually during the process of adaptation. If the environment of the robot is changed, the former environmental information was not effectively reflected in our previous NNs controller. Therefore, the new method which keeps the information of initial state or previous environment and adapt to the new environment should be developed to increase the efficiency of learning and reduce the learning cost with the use of the former environmental information which the robot had learned. The new adaptive control system for AUV using modular network SOM by K. Tokunaga et al. has been proposed. The efficiency of the system is investigated through identification of dynamics of an underwater robot.