Research On Low-latency And High-reliability Transmission Mechanisms For Industrial Internet Scenarios

With the rapid development of the industrial internet,latency-sensitive applications have put forward more stringent requirements for communication networks,such as high reliability,deterministic low-latency,ultra-low jitter,etc.At present,some industrial network communication protocols have provided low-latency transmission services.But due to many types of communication protocols,it is difficult for devices between different protocols to communicate with each other.In recent years,Ethernet technology,based on "best-effort" transmission mechanism,has been widely used in the consumer Internet.However,how to meet the deterministic low-latency requirements in latency-sensitive applications is still a challenging problem.Time-sensitive networks(TSN)has become a key technology to address the above problems.This technology precisely synchronizes the clocks of devices on the entire network,uses the traffic shaping and scheduling mechanism to shape and schedule different types of traffic,and reserves resources in advance for latency-sensitive applications through resource reservation,so as to achieve deterministic low-latency,low jitter and zero packet loss.Therefore,TSN technology has become one of the most important future orientations in the areas of network communication.The deterministic low-latency communication can be guaranteed by using the traffic shaping and scheduling mechanism in TSN effectively.How to shape traffic and optimize scheduling is a research hotspot to improve network performance.At present,there are a large number of wireless devices(such as mobile robots,automated guided vehicles,etc.).In order to meet the mobility of equipment and improve flexibility,it is essential to extend the determinacy of TSN from the wired field to the wireless field.Therefore,the integration of TSN technology into the 5th Generation Mobile Communication Technology(5G)is another research hotspot.This thesis studies the low-latency and high-reliability transmission mechanisms for industrial internet scenarios.The researches conduct in-depth research from two aspects of TSN traffic scheduling technology and TSN integrated 5G technology(5G-TSN)to ensure deterministic low-latency for key industrial applications and highly reliable transmission.And the network performance and the scheduling success rate of latency-sensitive flow can be improved simultaneously.Then,a feasible solution can be provided for the bounded low-latency and high-reliability data transmission of the future industrial Internet.The main research works and innovation points of the thesis are as follows:1.Research on no-wait traffic scheduling strategy for time-sensitive networksIn the TSN scenario with known routes,a heuristic-based and precise-based traffic scheduling strategy relies heavily on human experience,resulting in poor scalability.Therefore,a no-wait traffic scheduling strategy based on deep reinforcement learning is proposed to solve this issue.First,a no-wait traffic scheduling model is proposed to solve the problem of network bandwidth waste caused by guard bands.In this model,the time for each latency-sensitive flow to inject data at its source sender,is reasonably arranged to eliminate the queuing delay of latency-sensitive flows in the network so as to realize the no-wait transmission of latency-sensitive flows in the network.Second,a traffic scheduling algorithm,using a deep reinforcement learning,is designed to improve network resource utilization.In this algorithm,the scheduling completion time of latency-sensitive flows is optimized to compress their transmission at the beginning of the schedule.Then,the number of guard bands can be reduced and the network bandwidth resources can be saved.Finally,the performance of this strategy is proved by simulations.The experimental results show that the effective scheduling scheme can be achieved within an acceptable time by using the proposed strategy,which solves the problem of poor scalability of traditional scheduling strategies.2.Research on joint optimization strategy of routing and scheduling for time-sensitive networksWhen calculating the traffic schedule,the route is taken as a known condition,which seriously reduces the scheduling success rate and network performance of latency-sensitive flows.To solve this problem,a joint optimization strategy of routing and scheduling,based on load balancing,is proposed in this thesis.First,a latency-sensitive flow sorting mechanism is developed to reduce the impact of different flow scheduling orders on the scheduling success rate.In this mechanism,flow sequencing is performed by evaluating the scheduling priority weight of each latency-sensitive flow.Second,a joint routing and scheduling optimization algorithm,based on load balancing,is designed by using the proposed latency-sensitive flow sorting mechanism.In this algorithm,the injection slot offset mechanism and dynamic routing mechanism of latency-sensitive flows are considered jointly,and delay sensitive flows are scheduled one by one to improve the scheduling success rate.Finally,the scheduling success rate and resource utilization rate obtained by the proposed joint optimization strategy are analyzed.Experimental results represent that this strategy not only improves the scheduling success rate of latency-sensitive flows,but also increases the network resource utilization rate.3.Research on end-to-end traffic scheduling strategy for 5G-TSN networkIn the 5G-TSN network scenario,it is vital to ensure the deterministic scheduling of end-to-end traffic.However,current works lack research on end-to-end traffic scheduling for 5G-TSN networks.Thus,an end-to-end traffic scheduling strategy for5G-TSN networks is devised in this thesis to solve this problem and fill the gap in this field.First,an industrial network architecture,integrating 5G-TSN,is proposed.And a uniform time slot length,based on this architecture,is designed for the entire network,so that the 5G-TSN network can apply the cyclic queuing and forwarding mechanism for traffic scheduling.Second,a hierarchical particle swarm optimization algorithm,based on Double Q-Learning,is designed to solve the end-to-end traffic scheduling problem in the 5G-TSN network.The algorithm adopts a hierarchical population structure,and uses Double Q-Learning to dynamically adjust the population level after each iteration,making the population jump out of the local optimum,improving the search efficiency of the algorithm,and optimizing the injection time slot of each flow.Finally,the performance of this end-to-end traffic scheduling strategy is demonstrated by simulations.The experimental results demonstrate that this scheduling strategy can improve the scheduling success rate of delay-sensitive flows effectively.