A Prototype for General Planning

Planning problems, in which a sequence of actions are sought to go from a given state to a desired final state, or goal, have traditionally been hard to handle computationally. In particular and in the most usual case, the goal is not just to find a satisfying plan, but rather to find the best one. Searching for such optimum, i.e. shortest, plan, increases complexity even further. Frequently, specialised systems are developed for particular problems with ad-hoc solutions, losing generality and flexibility. Alternatively, a generalised naive solution is adopted, losing efficiency. Constraint programming (CP) has been used successfully to declaratively and efficiently solve large combinatorial problems. Unfortunately, modelling such problems in CP, requires declaring, a priori, a fixed number of variables with specified types and domains. This makes it hard to express planning problems, where the number of actions, possibly of different types, to perform is not known a priori. In this paper, we describe how planning problems can be declaratively and efficiently solved with an example system we developed based on logic programming and term rewriting. Additionally we show that such a system can also be used for general declarative programming, gaining in ease of use, flexibility, maintenance and debugging