Addressing the issue of stochastic environments and local decision-making in multi-objective reinforcement learning