Optimization of multiple-goal plans with limited interaction