Simulation environments for the design and test of an intelligent controller for autonomous underwater vehicles

Abstract

Intelligent controllers usually consist of a hybrid system that includes both discrete and continuous processes. This hybrid construction poses difficulties in validating and verifying their design. As the use of intelligent controllers proliferates throughout society, the development of simulation techniques that support both the construction and testing of these controllers becomes increasingly important. At the Applied Research Laboratory (ARL) of the Pennsylvania State University, we have gained insight over the last ten years (1989-99) into the design, implementation and testing of intelligent controllers for Autonomous Underwater Vehicles (AUV). However, as AUV missions become more complex, simulation environments must be provided that achieve complete state coverage of the discrete processes of the controller while still fully exercising the continuous processes with high fidelity Monte Carlo simulations. As an illustrative example, the paper describes the current utilization of simulation in the development and testing of intelligent mission controllers for AUVs using ARL's own intelligent control architecture. Then our new paradigm for simulation based design and testing for intelligent controllers is formulated and discussed.