Integrating policy transfer, policy reuse and experience replay in speeding up reinforcement learning of the obstacle avoidance task