Extending QMIX with Simple Techniques from Single- Agent Reinforcement Learning

Key Words: Multi-Agent Reinforcement Learning (MARL), Deep Learning

Link:

The Aim: To seek the potential of improving the performance and stability of QMIX, a value-based deep MARL algorithm in the paradigm of Centralized Training with Decentralized Execution (CTDE) by investigate its extensions.

Contributions:

(1) Proposed two extensions of QMIX by using Multi-step Learning and Dueling Networks. (2) Evaluated … (3) Discovered … hyphothesis (4) extend other in the ctde family