Influence of the reward function on the selection of reinforcement learning agents for hybrid electric vehicles real-time control