AI Planning in More Expressive Domains

	Amol D. Mali
	Electr. Engg & Computer Science
	University of Wisconsin, Milwaukee

	Significant advances have occurred in plan
	synthesis under classical assumptions in last seven years.
	This has increased the interest in an efficient synthesis
	of high quality plans in more expressive domains.
	Such domains have numerical variables, conditional
	effects, quantifiers, metric time and actions

        Such domains have numerical variables, conditional
        effects, quantifiers, metric time and actions
        whose pre-conditions and effects are functional.
        In addition, different pre-conditions and effects of actions
        are true at different times and some pre-conditions may
        need to be true over time intervals. I will provide a
        brief introduction to AI planning and more expressive
        domains. Most of the remaining talk will be
        about the more expressive planner being developed in my group.
        Specifically, I will talk about the search algorithm, various
        expressivity features handled, various solving options supported
        by the planner, experimental results and the heuristics used. 
        At the end of the talk, I will briefly discuss other AI work
        in my group like quantified weighted MAX-SAT and improvements 
        to local search for SAT solving.