SACNS: Sibelobe Control of Collaborative Beamforming Using Deep Reinforcement Learning in Wireless Sensor Network

Collaborative Beamforming (CB) based on Wireless Sensor Network (WSN) is a promising technology for low-power consumption, long-range data transmission. Due to the random distribution of nodes in WSN, unpredictable sidelobe may interfere with unintended direction. It is possible to control sidelobe level in unintended direction by selecting an appropriate combination of nodes. However, as the network scale in creases, the complexity of node selection grows exponentially, requiring existing methods to spend significant computation time when addressing the node selection problem in large-scale WSN, which results in high computational delay. To tackle this challenge, we first formulate a nonlinear optimization problem that aims to reduce the sidelobe level by selecting collaborative nodes within a given time while minimizing delay. Then based on deep reinforcement learning (DRL), we propose a SAC-based node selection algorithm (SACNS) to solve this problem. Subsequently, we propose a potential CB implementation scheme to validate the performance of the SACNS. The simulation results show that, compared with the baseline method, the proposed SACNS algorithm can effectively reduce the sidelobe level in CB optimization and significantly shorten the computation time.