Underwater Vehicle Control: Minimum Requirements for a Robust Decision Space

Development of automated tactical control in underwater vehicles is motivated by the growing need for both decision aids in manned vehicles, as well as for controllers in unmanned autonomous vehicles. While these systems may vary widely in architectural design, ultimately the system interacts with the world via a sequence of decisions or control variable assignments. The choice of these variables, and their domains, ultimately affects both the complexity (i.e. quickness) of the decision process and the precision of the individual decisions or actions. Operating a vehicle situated in the real, dynamic world means there are both real-time requirements and precision requirements. In this paper we examine the precision required in both submarine tactical decision aids and autonomous underwater vehicles, to identify a minimum set of control/decision variables. We also argue that existing techniques for handling the resulting large decision space are inadequate and often lead to damaging oversimplifications. We discuss these common simplifications and their drawbacks, and suggest new techniques that may handle the large spaces, preserving the necessary level of precision in control.