Project 1 - Efficient Transmission in Deep Reinforcement Learning
Individual Research Project:
-
Compared the impact of packet loss (latency) and noise (reliability) on the transmission between DRL agent and environment, and demonstrated that latency is more crucial for the overall DRL accuracy
-
Addressed the problem that some observation channels are more critical than others by proving that an importance-based resource allocation scheme is needed for a DRL agent
-
Proposed an Efficient Transmission Scheme for DRL agents to identify the essential observation channels when making a decision, and proposed further research to adapt this scheme to a time-varying hostile channel condition