UAV Network Routing With Deep RL Relay and Power Control

The patent implements self-service by enabling UAVs to autonomously select relay nodes and adjust transmission powers through deep reinforcement learning algorithms. Each UAV independently learns optimal routing decisions and power control strategies based on real-time network conditions, eliminating the need for complex centralized control while improving transmission quality and network reliability.

The patent applies dynamics by transitioning from static fixed relay nodes to dynamic relay selection. UAVs continuously adapt relay choices based on real-time network topology and interference conditions through deep reinforcement learning. This dynamic approach maintains routing stability while maximizing UAV utilization efficiency, as any UAV can serve as a relay when needed rather than relying on pre-designated fixed relays.

The patent implements feedback mechanisms where UAVs continuously monitor network conditions including channel states, interference levels, and transmission quality metrics. The deep reinforcement learning algorithms process this feedback in real-time and automatically adjust transmission powers to maintain optimal performance. This closed-loop feedback system achieves both precise power control and rapid response to network changes, eliminating the delays associated with manual power control.

The patent applies preliminary action by pre-training deep reinforcement learning models offline using simulated UAV network environments. The algorithms learn optimal routing and power control strategies in advance through extensive training before deployment. During actual operation, the pre-trained models make rapid decisions by evaluating current conditions against learned patterns, achieving both fast processing speed and high adaptability to dynamic topology and interference without real-time computational overhead.