Research On A Novel MAC Protocol For The Distributed Satellite System

Distributed satellite systems are widely used in many areas,and become an important part of the space information network.There is a real-time data that is generated sporadically and a low priority data that is generated periodically in distributed satellite systems.The conventional multiple access protocols in inter-satellite networks cannot meet the stringent delay requirement of the real-time data.Most of the existing communication protocols which support data with multiple priorities reserve dedicated resources for real-time data to guarantee its delay constrains,which will result in the waste of resources.Therefore,it is vital to design a novel communication protocol for distributed satellite systems.In this thesis,we studied the transmission problem of the burst real-time data and the selection problem of the back-off slots for delay-sensitive data.The specific researches and contributions are summarized as follows.First,we studied the transmission problem of distributed satellite systems and introduced a novel multiple access protocol to meet the characters of distributed satellite networks.Compared to the conventional mechanism,the proposed protocol combined time division multiple access and random access to pre-allocate the whole resources to each node.Each node transmits data in its own slots.When a node generates burst data with high priority,the protocol supports the node contending for resources immediately.The protocol also supports many data with multiple priorities contending for resources randomly.The proposed protocol avoided reserving dedicated resources to improve the resource utilization.We made a theoretical analysis for the proposed protocol using Discrete-Time Markov Chain.The results provide the guidance to set the optimum parameters.The simulation scenario is a satellite cluster which consists of 15 satellites.The simulation results provide the performance gains of average waiting delay and resource utilization in two different scenarios.One is the scenario that only one node generates data with high priority,such as quick message report of position in Bei Dou system,and the other is the scenario with five nodes generating high priority data simultaneously.Then,we proposed a mechanism that supports multiple data contending for resource based on reinforcement learning theory.Current mechanisms which support multiple priorities data allow high priority data to access to the channel first,which results in the large delay for the low priority data.The data with low priority also has delay limitation.Therefore,this mechanism cannot meet the requirement of delay-sensitive data.The proposed mechanism fully considered priority of data and delay limitation.The nodes select the back-off slots automatically,according to the requirement of current data and the condition of channel.The proposed mechanism is based on Q-Learning algorithm.A reward function which takes comprehensively into account three factors of the priority of data,the waiting time of data and the feedback of channel condition was designed.The simulation results provide performance gains of the learning rate,average waiting delay of delay-sensitive and packet loss rate in the scenario with seven delay-sensitive data in the same slot.