Section: New Results

Packing curved objects

Participants : Nicolas Beldiceanu, Gilles Chabert, Ignacio Salas Donoso.

The development of algorithms to pack curved objects has continued. The filtering algorithm developed in 2013 for generic objects shapes has been published in the CP 2014 conference. Based on this result, we have started the design of a generic (nonlinear) packing solver in 2014. The strategy for packing is directly inspired from a successful approach recently proposed by our project partners (see On solving mixed shapes packing problems by continuous, T. Martinez et al., first BRICS countries congress on Computational Intelligence). It makes use of a stochastic optimization algorithm (CMA-ES) with a fitness function that gives a violation cost and equals zero when objects are all packed. We have generalized their approach by replacing the ad-hoc formulas (for measuring the overlapping between two objects) with an automatic calculation based on our filtering algorithm. The solver is done and the experiments have started.