A real-time goal-interaction free task planner

The authors describe a real-time task planning algorithm, PLANNER. The high planning speed achieved in PLANNER is due to three techniques adopted in the planner. First, priority numbers are assigned to each literal of the conjunctive goals and each literal of the preconditions of the robot operations; the priority assignment is combined with a depth first search technique to make the goal-interaction problem vanish in task planning. Second, binary numbers are used to represent facts and robot operations, which greatly increases the speed of operation searching. Third, simple hardware circuits are used to match a subgoal to all operations and facts in a parallel fashion. The total speed gain of PLANNER over the traditional task planner is the product of the speed gains of these three techniques. PLANNER can be used for a fairly large subset of conjunctive problems that have priority among its subproblems