Comparative Study of Multi-Agent Actor-Critic Algorithms in Parameterized Action Reinforcement Learning