Energy-efficient Predictive Deployment Strategy of UAVs Based on ConvLSTM with Attention Mechanism

Unmanned Aerial Vehicle (UAV) can be used as the air base station to cover flexibly hotspots by its mobility. However, it is challenging for the network operators to forecast the distribution of network traffic and optimize the deployment of UAVs. To solve this problem, an energy-efficient predictive deployment strategy of UAVs based on ConvLSTM with Attention mechanism (A-ConvLSTM) is proposed: a convolutional long short term memory deep spatio-temporal network model A-ConvLSTM with attention mechanism is proposed to forecast the spatio-temporal distribution of users and cellular traffic. Then based on the forecast, the coverage and locations of UAVs are optimized. On the premise of meeting the requirements of user access rate, an optimization formulation is established with the goal of minimizing the transmission power of UAVs. The formulation is decoupled into two subproblems and an energy-efficient deployment algorithm is proposed for iterative solution. The experimental results show that the performance of A-ConvLSTM is better than that of each baseline model. Energy-efficient deployment algorithm can effectively reduce the transmission power consumption of UAVs, and achieve the overall area coverage with fewer UAVs.