Abstract: Reinforcement Learning (RL) for underactuated mechanical systems presents unique challenges due to the limited control inputs and complex dynamics of the system. Efficient exploration poses ...