Research On Routing Algorithm Of Wireless Mesh Network In Video Transmission Scenario

The rapid development of 5G communication and the Internet of Things has promoted the rapid growth of multimedia services,making it an indispensable part of our daily lives.Therefore,the demand for wireless communication networks that can operate in all scenarios and environments has reached an unprecedented height.Although traditional wireless local area networks are stable and have high bandwidth,they cannot meet the fast and stable transmission requirements for large-scale video services.Wireless mesh networks are a new type of wireless Ad Hoc network that can be widely used in education,healthcare,security,and other fields to effectively handle the increasing demand for multimedia services.Although they have advantages such as low deployment costs and easy installation,they also have various shortcomings.Firstly,the nodes in the network are randomly distributed and usually mobile,which leads to a complex network topology.If unsuitable routing protocols and algorithms are used,it can result in serious waste of communication resources.Secondly,the number of available channels is limited based on the network topology,and each node’s load capacity is limited.If an unreasonable routing selection strategy is designed,it can cause local network congestion and significantly reduce data transmission efficiency.Therefore,this thesis proposes two improved routing algorithms for wireless mesh networks to overcome these limitations.The first algorithm uses Q-learning and introduces Qo E perception for intelligent selection of multiple paths.The second algorithm combines Q-learning with feedforward neural networks when there is heavy traffic in the network topology,and designs a load balancing reward function as the evaluation indicator for path selection.Aiming at the path selection problem of video stream transmission,a multi-interface multi-channel wireless Mesh network multi-path routing algorithm based on Q-learning and Qo E perception is designed.(1)Q-learning algorithm is used to train the network.A self-adaptive Q-learning routing learning strategy is loaded for each node in the mathematical model of the wireless Mesh network topology.When video stream data is sent to a network node,the node selects an action based on the current state and interacts with the environment,interacting with the next neighboring node according to the routing learning strategy.In this process,the network gives real-time feedback to the node with corresponding reward and updates it to the Q-value table.The objective of Q-learning strategy is to find the optimal path to maximize the expected cumulative reward between the source node and the destination node.(2)Design the Qo E perception reward function.In the training process of Q-learning,a Qo E perception model is introduced to evaluate the transmission quality of video stream data,and a Qo E-based reward function is designed.This reward function can effectively consider the Qo E indicators of multimedia data transmission during the training process,supplement the transmission data characteristics,and improve the user’s experiential value of multimedia data after transmission is completed.(3)Improve the master-slave path intelligent selection strategy.The objective correlation between the optimal path obtained by Q-learning and other available transmission paths is calculated to select the path with the lowest objective correlation as the slave path.This strategy can adaptively select the transmission status of the link with better performance for video stream data,effectively improve the network’s risk resistance during transmission and alleviate the network’s load congestion problem to some extent.Aiming at the local congestion when the video stream is overdosed in the network,a multi-interface multi-channel wireless Mesh network load balancing routing algorithm based on deep Q-learning is designed.(1)The feedforward neural network is used to replace the Q value table mechanism in Q-learning.When there are a large number of video stream data in the network,the Q-value table in the Q-learning algorithm can occupy a lot of memory space,which can easily lead to local congestion of some topology nodes with good link quality in the network.Combining Q-learning algorithm with neural network can optimize the training process of the network.Updating Q-values in the neural network can effectively reduce the memory consumption of the training process,thus improving the performance of the network.(2)Design the load balancing reward function for training.The link load conditions and Qo E-awareness measurements are used as factors for the reward function,targeting the improvement of wireless link load conditions in the network topology,and improving the reliability of video stream data transmission in the network.The two routing algorithms proposed in this article are compared with other traditional routing algorithms,and the proposed algorithms show improved network performance for video streaming data transmission compared to the compared algorithms.