Planning Under Time Pressure

Heuristic search is a technique used pervasively in the fields of artificial intelligence, automated planning and operations research to solve a wide range of problems from planning military deployments to planning tasks for a robot that must clean a messy kitchen. An automated agent can use heuristic search to construct a plan that, when executed, will achieve a desired task. The search algorithm explores different sequences of actions that the agent can execute, looking for a sequence that will lead it to a desired goal state. In many situations, an agent is given a task that it would like to solve as quickly as possible. The agent must allocate its time between searching for the actions that will achieve the task and actually executing them. We call this problem planning under time pressure. Classic heuristic search algorithms do not address planning under time pressure. A* (Hart, Nilsson, and Raphael 1968), one of the most well-known heuristic search algorithms, finds optimal plans that have the minimum execution time. (In general A* optimizes any cost metric, however, we are concerned with time so we assume that cost = time.) Unfortunately, it is often impractical or intractable to find an optimal plan. Other algorithms, such as greedy best-first search (Doran and Michie 1966), will quickly find solutions of unbounded sub-optimality. Unfortunately, these greedy solutions often lead to plans that are too time consuming to execute. Finally techniques, such as weighted A* (Pohl 1970) or more recently Explicit Estimation Search (Thayer and Ruml 2011), try to find a balance between optimality and unbounded sup-optimality by returning solutions that are guaranteed to be within a user-specified factor of optimal. It is not clear, however, how to choose a suboptimality factor to properly trade-off planning time and execution time. The thesis of my dissertation is: when under time pressure, an automated agent should explicitly attempt to minimize the sum of planning and execution times, not just one or just the other. My plan for tackling the problem of planning under time pressure has three stages. The first stage focuses on parallel heuristic search, where the parallel capabilities of modern computing hardware are used in order to decrease search