Rate-Aware Quantum-Inspired Trajectory Learning for Interference-Limited Multi-UAV Networks

As Unmanned Aerial Vehicles (UAVs) are increasingly used to provide on-demand connectivity, optimizing their trajectories to maximize network throughput while minimizing interference becomes computationally prohibitive. This paper addresses the curse of dimensionality in multi-UAV trajectory optimization, where the vast search space and interference-limited nature of wireless environments make real-time coordination difficult for traditional reinforcement learning models.

The authors propose the Rate-Aware Quantum-Annealed Graph Condensation (RA-QAGC) framework, which operates in two stages. First, it performs offline global state-space compression. Instead of using simple geometric clustering, it uses a rate-aware cost function that accounts for SINR dynamics and co-channel interference. A quantum-inspired probabilistic annealing mechanism then explores this cost landscape to identify a set of optimal, discrete waypoints. Second, these waypoints form a connectivity graph that restricts the movement of UAVs. This allows decentralized Independent Q-Learning (IQL) agents to navigate the network efficiently, as they only need to select paths between pre-validated, high-performance locations.

By decoupling global coordination from local control, the RA-QAGC scheme significantly reduces the computational burden of trajectory planning. Simulations show that the proposed method achieves a total network throughput of 59.4 Mbps and a priority-user throughput of 23.9 Mbps. These results represent performance improvements of approximately 15% and 34%, respectively, compared to baseline schemes that do not utilize rate-aware state condensation.

This research provides a scalable solution for managing multi-UAV networks in complex, interference-limited environments. By integrating quantum-inspired optimization with reinforcement learning, the framework effectively balances the need for high-quality service with the practical constraints of real-time, decentralized UAV control, offering a pathway toward more resilient aerial communication networks.