A. Brogi Et Al. / the Logic of Totally and Partially Ordered Plans 34 7 { Conclusions

In this paper, we have shown that complex planning problems and non-monotonic strategies nd natural logic-based formulations in the framework of deductive database languages. In particular, we have shown how totally and partially ordered plans can both be modeled using XY-stratiied Datalog 1S programs. The framework proposed is exible, amenable to eecient implementation, and supportive of variations and improvements. For instance, notions such as rigid plans and surprises can also be included in the model 29]. In 5, 7], we generalize our logic-based framework to handle parallel plans, that is, plans where actions that are compatible under some concurrency model are executed in parallel. Since the Datalog-oriented approach discussed in this paper represents a minimalist approach to the complex problems occurring in the planning domain, there is a need to investigate and understand its limits in modeling other complex planning problems. In particular, several new concepts and techniques have recently emerged in the eld of planning research. For instance, in least-commitment planning , search is performed in the space of partially ordered plans, rather than in the space of world states (see 33] for a good discussion of the method and its beneets). Current work is oriented in this direction. Another interesting extension is to consider the problems of temporal projection with incomplete information, as well as reasoning about the past. Gelfond and Lifschitz 17] addressed these issues by representing properties of actions in a logic programming language that uses both classical negation and negation by failure. Future work will be devoted to investigate the extendibility of our approach to such more complex forms of temporal reasoning.