Chaos-based reinforcement learning with TD3