Research On Multimodal Transmission Mode In Ad Hoc Network

Node movement and changes in geographical and electromagnetic environments in MANET networks cause frequent changes in wireless link quality and network topology,resulting in unstable end-to-end communication path quality in the network.Fixed transmission modes are difficult to adapt to these changes and cannot meet the Qo S requirements of the services.Therefore,this thesis designs a multimodal transmission framework,which can jointly decide the suitable transmission mode according to the service Qo S requirements and current path quality to optimise the service transmission performance.Meanwhile,there are usually multiple reachable paths between MANET network nodes,and the traditional multipath data transmission protocol uses a fixed load allocation policy that cannot be adjusted for the dynamic variability of the paths;therefore,this thesis uses a deep reinforcement learning framework to optimise the multipath load allocation algorithm and improve the service transmission efficiency.The main work of this thesis is as follows.First,this thesis implements multiple transmission modes based on the number of available paths from the source node to the destination node and generalised fountain coding design: single-path transmission,single-path coded transmission,multipath load balancing transmission,multipath redundancy and multipath redundancy coded transmission.On this basis,two transmission mode decision algorithms based on decision trees and neural networks are designed and proposed,both of which can select the transmission mode that satisfies the user according to the current path quality and service Qo S requirements.Secondly,to address the problem that the load allocation policy cannot be adaptively adjusted under the multipath load balancing transmission mode,this thesis proposes an asynchronous reinforcement learning-based load allocation method,which takes the current path quality as the input data,integrates three metrics of end-to-end throughput rate,delay and submission rate as the reward function,and uses the self-learning feature of reinforcement learning to adjust the load allocation ratio of each subpath,and combines the decision network and The method improves the efficiency of data transmission in multi-path load balancing mode.This thesis then develops a multimodal transmission module,a transmission mode adaptive selection module and a multipath load allocation optimisation module based on a laptop Linux system,builds several network nodes,and simultaneously combines a wireless channel emulator or a wireless network card to build various experimental scenarios to test and study the performance of the multimodal transmission protocol and the load allocation algorithm of the multipath load balancing mode.The test results show that the coded transmission improves the reliability of service transmission;the transmission mode decision algorithm can adapt to path quality changes to switch the transmission mode that meets user requirements,ensuring that the performance indicators of service transmission delivery rate and throughput rate are better than those of a single transmission mode;the multipath load allocation algorithm shortens service transmission time and improves end-to-end throughput rate compared with the traditional allocation strategy,and achieves optimized service delivery.