A deep reinforcement learning integrated aerial RIS system for enhanced combat intelligence

The military operations demand effective communication and proper situational awareness especially in situations where the line-of-sight communication is compromised and performance of a cluttered situation can be impaired by the interference of the clutter. The combination of communication and sensing systems offers a good solution to the simultaneous transfer of data and target detection. Nevertheless, efficiency, reliability and minimum interference in these systems are still complex challenges to optimize these systems. In this paper, a new method of an aerial reconfigurable intelligent surface (A-RIS) in an integrated sensing and communication (ISAC) system is presented. It is installed with a central base station (BS), A-RIS, target sensing and a number of users. Using deep reinforcement learning (DRL) we suggest the mechanism to optimize jointly transmit beamforming at the BS, RIS phase shifts and A-RIS pathway under the strict signal-to-interference-plus-noise ratio (SINR) constraints. The results show that there are major improvements in the performance of the system such as a minimization in self-interference and clutter echoes as well as adaptive RIS phase shift to changing battlefield conditions. Our approach out-muscles the conventional approaches that obtain significant improvements in the reliability of communication and target identification.