IRS-Assisted UAV Secure Communications via Joint Collaborative and Passive Beamforming

Unmanned aerial vehicle (UAV) networks play a crucial role in 5G/6G wireless communications. However, the presence of known and potential unknown eavesdroppers poses security risks to UAV communications. To overcome this issue, we construct part UAVs within a UAV swarm as a virtual antenna array (VAA) and then introduce intelligent reflecting surface (IRS) to avoid eavesdropping from these eavesdroppers. By adopting joint collaborative and passive beamforming of VAA and IRS, our objective is to jointly optimize the secrecy rate, maximum sidelobe level, and total energy consumption of the system, by determining appropriate excitation current weights and trajectories of the UAVs, and phase shifts of IRS elements. Considering the dynamic and heterogeneity of the system, we transform the problem into a heterogeneity Markov decision process (MDP). Then, a heterogeneous multi-agent control approach (HMCA) consisting of an IRS control policy and a multi-agent soft actor-critic UAV control policy is proposed. Simulation results show that the proposed HMCA effectively solves the optimization problem and it has better performance than other baseline approaches.