Probabilistic, Temporal Projections in ConGolog

The ability of mobile robots to project plans, i.e. to predict what will happen when they execute a plan, is crucial to flexibly plan and react in uncertain environments. While classical formalizations of this problem make strong assumptions about the world and restrict robot plans to simple, partially ordered sets of actions, real world robot controllers make use of concurrency, priorities and use probabilistic models. In this paper we present an extension of the situation calculus, representing time, concurrency, probabilistic belief, probabilistic action effects and complex plans, integrating and extending several previous proposal into a single formal framework. We apply this framework to model and reason about the behavior of a mobile service robot.