Seventh International Planning Competition (IPC 2011)
DAEYAHSP Winner of the Seventh International Planning Competition (Deterministic Temporal Satisficing track)
YAHSP2-MT Runner-up of the Seventh International Planning Competition (Deterministic Temporal Satisficing track)
Divide-and-Evolve: the Marriage of Descartes and Darwin
DAEx, the concrete implementation of the Divide-and-Evolve paradigm, is a domain-independent satisficing planning system based on Evolutionary Computation. The basic principle is to carry out a Divide-and-Conquer strategy driven by an evolutionary algorithm. The key components of DAEx are a state-based decomposition principle, an evolutionary algorithm to drive the optimization process, and an embedded planner X to solve the sub-problems. The release that has been submitted to the competition is DAEYAHSP, the instantiation of DAEx with the heuristic forward search YAHSP planner. The marriage of DAE and YAHSP matches a clean role separation: YAHSP gets a few tries to find a solution quickly whereas DAE controls the optimization process. The DAEx planner is developed at TRT/LMTD by Pierre Savéant and Johann Dréo and stems from a collaboration started in 2006 with Marc Schoenauer (INRIA) and Vincent Vidal (ONERA). It is currently supported by the French National research Agency (ANR) through the COSINUS programme, under the research contract DESCARWIN (ANR-09-COSI-002).
YAHSP2: Keep It Simple, Stupid
YAHSP2-MT: the Multi-Threaded Release of YAHSP
The International Planning Competition
The International Planning Competition (IPC) is a biennial event organized in the context of the International Conference on Automated Planning and Scheduling (ICAPS). The objectives of the competition are to provide a forum for empirical comparison of planning systems, to highlight challenges to the community in the form of problems at the edge of current capabilities, to propose new directions for research and to provide a core of common benchmark problems and a representation formalism that can aid in the comparison and evaluation of planning systems.
After a 3 years gap, the 2011 edition of the IPC involved a total of 55 planners, some of them versions of the same planner, distributed among four tracks: the sequential satisficing track (27 planners submitted out of 38 registered), the sequential multicore track (8 planners submitted out of 12 registered), the sequential optimal track (12 planners submitted out of 24 registered) and the temporal satisficing track (8 planners submitted out of 14 registered). A total of 55 people were participating, grouped in 31 teams. Participants came from Australia, Canada, China, France, Germany, India, Israel, Italy, Spain, UK and USA.
For the sequential tracks, 14 domains, with 20 problems each, were selected, while the temporal one had 12 domains, also with 20 problems each. Both new and past domains were included. As in previous competitions, domains and problems were unknown for participants and all the experimentation was carried out by the organizers.
To run the competition a cluster of eleven 64-bits computers (Intel XEON 2.93 Ghz Quad core processor) using Linux was set up. Up to 1800 seconds, 6 GB of RAM memory and 750 GB of hard disk were available for each planner to solve a problem. The detailed results of the competition, the software used for automating most tasks, the source code of all the participating planners and the description of domains and problems can be found at the competition’s web page: http://www.plg.inf.uc3m.es/ipc2011-deterministic
The work on the DAEx planner, with Jacques Bibaï, Pierre Savéant, Marc Schoenauer and Vincent Vidal, got a Silver Medal at the 7th Annual Humies Awards for Human-Competitive Results Produced by Genetic and Evolutionary Computation (HUMIES'10) held at the ACM-Genetic and Evolutionary Computation Conference (GECCO 2010), Portland.
More details on SIGEVO web site