Planning as constraint satisfaction: solving the planning graph by compiling it into CSP

• Authors: Minh Binh Do;Subbarao Kambhampati
• Venue: Artificial Intelligence
• Year: 2001

The idea of synthesizing bounded length plans by compiling planningproblems into a combinatorial substrate, and solving the resultingencodings has become quite popular in recent years. Most workto-date has however concentrated on compilation to satisfiability(SAT) theories and integer linear programming (ILP). In this paperwe will show that CSP is a better substrate for the compilationapproach, compared to both SAT and ILP. We describe GP-CSP, asystem that does planning by automatically converting Graphplan'splanning graph into a CSP encoding and solving it using standardCSP solvers. Our comprehensive empirical evaluation of GP-CSPdemonstrates that it is superior to both the Blackbox system, whichcompiles planning graphs into SAT encodings, and an ILP-basedplanner in a wide range of planning domains. Our results show thatCSP encodings outperform SAT encodings in terms of both space andtime requirements in various problems. The space reduction isparticularly important as it makes GP-CSP less susceptible to thememory blow-up associated with SAT compilation methods. The paperalso discusses various techniques in setting up the CSP encodings,planning specific improvements to CSP solvers, and strategies forvariable and value selection heuristics for solving the CSPencodings of different types of planning problems. Copyright 2001.Elsevier Science B.V.