Minimizing Age of Information in Multi-party UAV-Assisted Data Collection with Overlapping Areas

Unmanned Aerial Vehicles (UAV) are widely used to assist the data collection in the Internet of Things (IoT), especially when IoT devices are deployed in critical environments without infrastructures. In such scenarios, the Age of Information (AoI) is a key metric to evaluate the performance of the UAV-assisted IoT data collection system. However, current studies usually consider all IoT devices and UAVs belong to one institution. In practice, there may be IoT devices and UAVs belonging to different institutions are deployed in an overlapping area, where the UAVs only collect the data from the IoT devices of its own institution. Actually, the cooperation among different institutions can substantially improve the performance of the whole system. Based on Graph Neural Networks (GNN) and Transformer, this paper proposes a cooperative optimization method to minimize the average AoI in multi-institutional UAV-assisted data collection with common areas. By leveraging GNNs to capture and encode the positional features of IoT devices, and then passing the encoded features to a Transformer decoder using an attention pointer mechanism, we generate optimized UAV trajectory for efficient IoT data collection with the minimum average AoI. Simulation results show our proposed scheme significantly improve the system performance. Compared to the benchmark schemes, our proposed scheme can reduce the average AoI by up to 23.6%, and shorten the UAV trajectory length by up to 21.0%.