Ensemble consensus representation deep reinforcement learning for hybrid FSO/RF communication systems

Hybrid FSO/RF system requires an efficient FSO and RF link switching to improve the capacity by realizing complementary benefits of both links. The dynamics network conditions, such as fog, dust, sand storms compound problem control complexity. To address this problem, we initiate study deep reinforcement learning (DRL) for hybrid systems. Specifically, focus on actor–critic called Actor/Critic-FSO/RF Deep-Q (DQN) DQN-FSO/RF under atmospheric turbulences. formulate define state, action, reward function a system. frequently updates deployed policy that interacts with environment in system, resulting high costs. overcome this, lift ensemble consensus representation learning-based DRL DQNEnsemble-FSO/RF. proposed DQNEnsemble-FSO/RF approach uses learned features based asynchronous threads update policy. Experimental results corroborate DQNEnsemble-FSO/RF’s consensus-learned demonstrate better performance than Actor/Critic-FSO/RF, DQN-FSO/RF, MyOpic while keeping cost low. provide interesting insights into prediction received signal strength indicator (RSSI) switching.