Observation of guarded command programs

We consider the problem of estimating the internal state of a class of guarded-command programs. Such programs model dynamical systems that may have both continuous and discrete states. We propose a way to construct an observer that takes advantage of the guarded-command structure of the program. We then apply our ideas to the "RoboFlag Drill" example wherein two teams of robots compete in a simplified capture-the-flag-like game. This system is complicated enough that the proposed observer is practicable only for small numbers of agents. We then propose an approach for reducing the complexity of the observer that takes advantage of the particular structure of the RoboFlag Drill example.