In this chapter, we tackle uncertain information in dynamic wireless networks due to time-varying channel conditions, mobility, or measurement delays, which motivates the Bayesian-enhanced DRL for more robust learning by leveraging Bayesian inference. The case study focuses on a UAV-assisted wireless network to help relay GUs’ sensing information to remote BS. We use MADRL to adapt UAVs’ trajectory planning and network formation strategies that minimize overall energy consumption and end-to-end transmission delay. Considering limited information exchange among UAVs, we further leverage Bayesian optimization for each UAV to infer a more preferable hovering location for data collection based on its historical trajectory observations. This reduces inefficient action exploration during the early stages of training and significantly improves convergence speed in the MADRL framework.