Reinforcement Learning-Based Resource Allocation for Streaming in a Multi-Modal Deep Space Network

With the development of 5G communication, users' data requirements are increasing rapidly, and the requirements for real-time data communication are increasing. These requirements were embodied as streaming services. This requirement can also be applied in deep space environments, where space communication, which is a relatively constrained environment, places importance on recovering the error rate. Typical channel coding techniques include turbo code, which is used to reduce Bit Error Rate. Taking this into account, this paper constructs deep space virtual environments, and applies various turbo-coding-based modulations for each link. In addition, streaming services use Scalable Video Coding to provide smoother streaming services. In this paper, we propose an algorithm that measures complexity and BER according to various modulation techniques of turbo codes to verify trade-off relationships and to assign data to appropriate links by learning them with Deep Q-learning Network.