Optimized Dynamic TDMA Scheduling Method Based On Service Requirements

With the development of communication technology,wireless ad-hoc networks(AdHoc)are widely used in military and civilan fields with its characteristics of centerless,distributed,scalability and flexible networking.In the diversified application scenarios,the parallel transmission of multiple types of services with different characteristics and demands becomes the norm,so the traditional Medium Access Control(MAC)protocol cannot meet the differentiated quality of service(Qo S)requirements of services.This thesis focuses on the characteristics and requirements of services,studies the distributed dynamic TDMA time slot allocation algorithm constrained by the limited wireless channel resources.The main research contents are as follows.Firstly,this thesis studies the service-driven hybrid time slot allocation algorithm.Since different services can be roughly classified into low-bandwidth services with high delay requirements or high-bandwidth services according to delay and bandwidth,we divide the data slots into node-oriented fixed time slots and shared data time slots.he former mainly ensures the transmission of services with high latency requirements,while the latter adopts service-oriented dynamic time slot allocation to meet the high bandwidth requirements of services.We design a frame structure with interleaved control time slots and data time slots to reserve information processing for nodes and achieve multiple time slot application and continuous occupation.In addition,we design a mechanism to resolve conflicts according to service priorities to ensure the transmission of high-priority services.The simulation results show that compared with the single slot allocation algorithms,the algorithm can meet the differentiated demand of services and improve success rate and average delay of service.Due to the dynamic nature of the services in ad-hoc network,with the change of the network environment,the statically planned fixed time slot resources may not match the actual demands of the nodes.Therefore,in order to prioritize the high-priority service transmission and reduce the extra overhead,this thesis proposes a prediction-based time slot resource reconfiguration algorithm.We use the long short time memory neural network algorithm to predict the traffic flow and reconfigure the fixed time slot resources of nodes according to the prediction results to maximize the time slot utilization while ensuring the transmission of high priority services.The simulation results show that the LSTM algorithm can effectively predict the service traffic and the time slot resource reconfiguration algorithm has better network performance compared with the static planning approach.Finally,we study the time slot application mechanism in the service-oriented dynamic time slot allocation algorithm and propose a slot number adaptive algorithm combined with Deep Reinforcement Learning(DRL)algorithm.The nodes are able to sense the changes of the network environment,learn and predict the slot occupancy,service access and service queue in the network,adaptively adjust the number of slots applied for each service,and integrate the dynamic time slot allocation algorithm we mentioned before to determine the final slot occupancy to achieve efficient utilization of slot resources and improve network performance.Numerical results show that the algorithm helps to reduce the average delay of services and can prioritize to ensure the successful access of highpriority services.