Joint Speed Control and Energy Replenishment Optimization for UAV-assisted IoT Data Collection with Deep Reinforcement Transfer Learning

Unmanned aerial vehicle (UAV)-assisted data collection has been emerging as a prominent application due to its flexibility, mobility, and low operational cost. However, under the dynamic and uncertainty of IoT data collection and energy replenishment processes, optimizing the performance for UAV collectors is a very challenging task. Thus, this paper introduces a novel framework that jointly optimizes the flying speed and energy replenishment for each UAV to significantly improve the data collection performance. Specifically, we first develop a Markov decision process to help the UAV automatically and dynamically make optimal decisions under the dynamics and uncertainties of the environment. We then propose a highly-effective reinforcement learning algorithm leveraging deep Q-learning, double deep Q-learning, and a deep dueling neural network architecture to quickly obtain the UAV's optimal policy. The core ideas of this algorithm are to estimate the state values and action advantages separately and simultaneously and to employ double estimators for estimating the action values. Thus, these proposed techniques can stabilize the learning process and effectively address the overestimation problem of conventional Q-learning algorithms. To further reduce the learning time as well as significantly improve learning quality, we develop advanced transfer learning techniques to allow UAVs to ``share'' and ``transfer'' learning knowledge. Extensive simulations demonstrate that our proposed solution can improve the average data collection performance of the system up to 200% compared with those of current methods.