Towards Plan Recognition in Hybrid Systems

Hybrid systems model software-based control systems to deal with real-world applications that exhibit an interaction of discrete and continuous dynamic behaviour. Hybrid automata (HA) is a widely used formalism for model checking of hybrid systems including reachability analysis, safety and stability verification. Given a HA model and a set of state observations received at certain times, we focus on determining if the HA accepts a trajectory that recognizes the observations. We formulate this task as a planning problem encoded in PDDL+ and use a planner to find a hybrid plan that satisfies the observations while following the HA dynamics. Ultimately, we show the adequacy of PDDL+ to model HAs with two well-known examples in the literature on HAs.