The representation and the reasoning about simultaneous actions are necessary in solving the problems of multi-agent domains. Because the previous planners based on situation calculus can not deal with the simultaneous actions directly, they use interleaving approximation. But, it is impossible for them to manage the situations as see-saw in which simultaneous actions are essential.
In this thesis, we design a planner which can represent and reason about the simultaneous actions. To describe actions, we define an action description language similar to STRIPS representation. In addition, we define a meta language based on the extended situation calculus to describe the independency and causality between actions for reasoning about concurrent activities. The domain knowledge and state descriptions are represented by first order clauses. To generate a plan, the nonlinear planning technique is used because its data structure and problem solving method are appropriate to parallel actions.
It is shown that the plan generated by our planner is correct.
