Deep Reinforcement Learning Based Information Age Minimization for Unmanned Aerial Vehicle Data Collection

In recent years, unmanned aerial vehicle (UAV)-based Mobile Crowd Sensing (MCS) systems have emerged as ideal platforms for data collection due to their rapid deployment and controllable maneuverability. This paper proposes a UAV trajectory planning problem under charging environments to minimize the average age of information (AoI) at nodes using deep reinforcement learning. We first formulate the UAV data collection trajectory planning optimization problem to minimize information age and further model it as a Markov decision process. Subsequently, a deep reinforcement learning algorithm based on deep deterministic policy gradient is introduced, which integrates deep deterministic policy gradient algorithm, convolutional neural networks, long short-term memory networks, and prioritized experience replay technique to accelerate the convergence speed of the training process and exhibit superior performance of the trained model. Through comparative experiments with different parameters, we validate the superiority of the proposed algorithm over heuristic methods in minimizing the average AoI at nodes.