Handling Uncertainty in Least Committed Graphplan: A Conformant Approach

An artificial intelligent planner called conformant least committed graphplan (CLCGP) is proposed in this paper. This planner can handle uncertainty even without sensory information, which means it is possible to find valid plans no matter which of the allowed states the world is actually in. CLCGP is based on the famous planner LCGP, which has been proved to have great success in solving classic planning domains. The basic idea of this algorithm is to develop separate least committed planning graph for each possible world. The planner is implemented in common Lisp and tested on a IBM RS6000 machine, empirical results show that CLCGP performs significantly better than the famous conformant planner CGP