Multi-Agent Reinforcement Learning using the Deep Distributed Distributional Deterministic Policy Gradients Algorithm
Wael Farag · 2020
In this paper, the Deep Distributed Distributional Deterministic Policy Gradients (D4PG) reinforcement learning algorithm is adopted to train a multi-agent action in a cooperative game environment. The algorithm is experimented on training the agents to play a game of tennis against each other. The architectures of the actor and cretic networks are meticulously designed and the D4PG hyperparameters are carefully tuned. The trained agents are successfully tested in the Unity Machine Learning Agents environment. The testing shows the powerful performance of the D4PG algorithm in training multi-agents in complex environments.