UAVs and Mobile Sensors Trajectories Optimization with Deep Learning Trained by Genetic Algorithm Towards Data Collection Scenario

In challenging environments like deserts and forests, without communication facilities, data transmission is a key challenge. To solve this problem, data collection using Unmanned Aerial Vehicles (UAVs) is widely applied in such scenarios. However, when high-density data collection is required, traditional methods using UAVs are fatally flawed due to the limited energy of UAVs and the dense distribution of collection points. Aiming to this, a data collection strategy employing mobile acquisition devices (mobile sensors), fixed sensors, and UAVs cooperation is chosen. In this strategy, mobile sensors collect data and transmit it to fixed sensors. Then, UAVs collect and carry the returned data to the data center for further processing. Correspondingly, considering data collection using mobile devices, path planning for the mobile sensors and UAVs largely determines the system’s performance. In this kind of data acquisition scheme, the previous works lack efficiency, because the genetic algorithm (GA) consumes a lot of time to get the local optimal solution and deep reinforcement learning (DRL) highly depends on the selection of reward given by experience value. Hence, deep learning trained by genetic algorithm (DL-GA) is proposed as path planning method for mobile sensors and UAVs in this paper. Simulation results show that, in data collection strategy for high-density data collection, the proposed path planning algorithm can achieve higher efficiency compared to the other algorithms.