An implemented universal mission controller with run time ethics checking for autonomous unmanned vehicles—A UUV example

摘要

The authors have been involved for several decades in the development and testing of both remotely controlled and autonomous subsea and ground vehicles. This experience has led us to view autonomous mobile robot control problems from both a bottom up and a top down perspective. Specifically, in our work, we have developed and tested a three-level software architecture called Rational Behavior Model (RBM), in which a top (strategic) level mission control finite state machine (FSM) orders the rational execution, at an intermediate (tactical) level, of vehicle behaviors in such a way as to carry out a specified mission. This implementation experience and these principles have led us to believe that human-like intelligence and judgment are not required to achieve a useful operational capability in autonomous mobile robots. Furthermore, we are convinced that a primitive but useful type of robot ethical behavior can also be attained, even in hazardous or military environments, without invoking concepts of artificial intelligence. To support our views, we present a software invention called a mission execution engine (MEE), implemented in the Prolog logic programming language. This MEE can be shown to represent an extension of the idea of a universal Turing machine and is therefore well grounded in existing mathematical automata theory. We further show how human readable mission orders, also written in Prolog, can specialize an MEE to any desired mission control FSM. An important aspect of our work is that mission orders can be tested exhaustively in human executable form before being translated into robot executable form. This provides the kind of transparency and accountability needed for after action review of missions, and possible legal proceedings in case of loss of life or property resulting from errors in mission orders.