Learning-Based Power Minimization in Secure Multiple UAV-Aided MIMO Networks

In modern communication systems, the demand for efficient and secure data transmission has become increasingly important due to the rapid growth of wireless technologies and connected devices. This paper addresses the power minimization problem in secure multiple unmanned aerial vehicle (UAV)-aided multiuser multiple-input multiple-output (MU-MIMO) networks, where balancing secrecy and minimizing energy consumption poses a significant challenge. We formulate a non-convex problem with the secrecy target signal-to-interference-plus-noise ratio (SINR) constraints and UAV movement. We establish the optimality conditions, and the problem is then decoupled into the following sub-problems: power allocation, combiner optimization, precoder optimization, and UAV trajectory optimization, which are jointly optimized. We obtain the necessary and sufficient conditions for the first three sub-problems to show their equivalence to the established conditions. The power allocation, combiner, and precoder optimization are solved via the Jacobi recursion-based algorithm, and closed-form solutions are obtained for each. UAV trajectory is solved via successive convex approximation (SCA)-based algorithm. Aiming at providing real-time solution, we design a learning-based real-time power minimization (LRPM). LRPM consists of two parts; the first one is a recurrent convolutional neural network (RCNN) to optimize the UAV trajectory. The second part is designed using deep unfolding with the help of the closed-form solutions of the first three sub-problems. Simulation results show the efficiency of the proposed framework in minimizing power consumption and improving secrecy in real-time scenarios.