Non-Terminating Processes in the Situation Calculus

By their very design, many robot control programs are non-terminating. To give a simple example one we shall use in this paper an office coffee-delivery robot might be implemented as an infinite loop in which the robot responds to exogenous requests for coffee that are maintained on a queue. Since a future coffee request is always possible, the program never terminates. As is the case for more conventional programs, we want some reliability assurances for robot controllers. This paper describes the approach being taken by our Cognitive Robotics Group to expressing and proving properties of non-terminating programs expressed in GOLOG, a high level logic programming language for modeling and implementing dynamical systems. The kinds of properties we have in mind are traditional in computer science: liveness, fairness, etc. We differ from the "classical" approaches ([LS87, Cou90, MP95]) for reasons dictated by the following characteristics of GOLOG: