New treatment planning systems adapted to proton spot scanningtechniques have been recently proposed to simultaneouslyoptimize several fields and thus improve dose delivery.The only parameters usually optimized are the field fluenceswhereas other parameters (e.g. beam directions) are chosenby medical physicists and are limited by technological constraints.We investigate a genetic algorithm approach designed tooptimize the number of beams, target points and irradiationdirections at the same time. This tool is intended to pave theway towards optimal treatment ballistics. The optimizationframework is designed to be versatile and to account formany degrees of freedom, without any a priori technologicalconstraint.