Promela Planning

In this paper a compiler from a restricted subset of SPIN’s input language Promela into an action planning description language is presented. It exploits the representation of protocols as communicating finite state machines. The work targets the transfer between the state space exploration areas AI planning and model checking by making protocols accessible for action planners. It provides a practical study of the expressiveness of different input specifications, introduces concurrency and quality metrics to counterexample traces, and compares search, pruning and acceleration methods. Indirectly, refined estimates for improved error detection in directed protocol validation are introduced. For example, the relaxed plan heuristic, which comes along with an enforced hill climbing search engine. The experimental results are encouraging. In some sample protocols, planners perform close to state-of-the-art model checkers.