Learning to Rest: A Q-Learning Approach to Flying Base Station Trajectory Design with Landing Spots
| Autor: | David Gesbert, Harald Bayerlein, Rajeev Gangula |
|---|---|
| Rok vydání: | 2018 |
| Předmět: |
Artificial neural network
Computer science business.industry Deep learning 05 social sciences Real-time computing Q-learning 050801 communication & media studies 020206 networking & telecommunications Context (language use) 02 engineering and technology Base station 0508 media and communications Channel state information 0202 electrical engineering electronic engineering information engineering Wireless Artificial intelligence business Power control |
| Zdroj: | ACSSC |
| Popis: | Several key wireless communication setups call for coordination capabilities between otherwise interfering transmitters. Coordination or cooperation can be achieved at the expense of channel state information exchange. When such information is noisy, the derivation of robust decision-making algorithms is unfortunately known to be very challenging via conventional optimization method. In this paper we introduce a learning-based framework which allows the agents, aka. the transmitters, to produce as-relevant-as-possible messages to each other on the basis of arbitrarily partial and noisy local channel state information. The messages are produced via distributed deep neural networks (DNNs) which are trained for a specific coordination purpose. The message-passing DNNs are completed with decision-making DNNs which are trained for a network metric maximization. Promising preliminary results are obtained in the context of sum-rate maximizing decentralized power control. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|