Hey all, we're making a robot that teaches itself to walk using a genetic algorithm(took a hiatus with this but have returned). So i coded a simulator that to see why the GA wasn't working well with realistic physics. The GA works like this "A++,B---,D+" would mean motor A's angle increase by 2*num, and B's angle decrease by 3*num value, angle D's, will be increased by 1*num. But now we can't use standard crossover otherwise they will be a HUGE change between the new and old chromosomes, So our current idea is to cross just the +'s and -'s ie A++,B--- A--,B++ Would cross to A+-,B-+ But then motor changes have to be in a specific order, so unless they're in the optimum configuration which is cheating) it can never get to an optimum, Any advice? this is like the travelling salesman Genetic algorithm, which needed its own unique crossover and mutation. Any help appreciated.