Sites Inria

Version française

Scientific award

24/08/2011

DESCARWIN ANR wins the 7th annual international planning competition

The DAEX planning system, developed by the DESCARWIN ANR (French National Research Agency), has won 1st prize in the 7th annual International Planning Competition (IPC) in the Deterministic Temporal Satisficing track.

This international competition  is organised once every two years in the context of the ICAPS congress (International Conference on Automated Planning and Scheduling ). The purpose of the competition is to provide a forum for empirical comparisons of planning systems, to launch new challenges for the scientific community by testing the limitations of current systems, to propose new research avenues and, finally, to provide the basis for reference problems as well as formal representation rules in order to facilitate the assessment and comparison of planning systems.

Divide and Evolve  (DAE), a combination of Descartes and Darwin, is a planning system that optimises a "divide and conquer" strategy using an evolutionary algorithm.  The main components of DAEx are: the principle of sequential space of states decomposition, an evolutionary algorithm controlling the decomposition process and an embedded planner X, used to consecutively solve successive sub-problems resulting from the decomposition proposed by the evolutionary algorithm. The version that was submitted to the competition was DAEYAHSP, the instantiation of DAEX with the heuristic planner YAHSP. The combination of DAE and YAHSP results in a clear distinction of roles: after several attempts YAHSP quickly sends back a solution while DAE applies its optimisation strategy through successive evolutions.
 The DAEX planner is developed within the framework of the DESCARWIN ANR project, which brings together Thales Research & Technology (Pierre Savéant and Johann Dréo), Inria Saclay – Île-de-France (Marc Schoenauer, Tao project team) and ONERA Toulouse (Vincent Vidal).

Keywords: Planning Saclay - Île-de-France Tao Marc Schoenauer

Top