Study On Hybird Scheduling Methods For Perodic/Aperiodic Data Based On Age Of Information

Age of Information(Ao I)is a new data freshness metric,which describes the difference between the current time and the generation time of the latest data received by the target node,focusing on measuring the freshness of the delivered data from the perspective of the target node.In industrial wireless sensor networks,there are both periodic data for control tasks and aperiodic data for observation tasks.These two types of industrial field data need to be delivered to the control center in real time to ensure the safety and efficiency of industrial production process.Given that Ao I can simultaneously quantify the generation time and transmission delay of data at the control center.Therefore,compared with traditional data real-time measurement indicators such as delay and delivery time interval,Ao I can more comprehensively measure the real-time performance of data delivery in the industrial wireless sensor network with hybrid data updates.At present,for the network with data hybrid,optimizing the overall freshness of network data based on Ao I has become a research hotspot.However,the existing research does not consider the network throughput while improving the freshness of network data.In particular,there are a large number of sensor nodes in industrial wireless sensor networks,and these nodes need to deliver data as much as possible under the condition of limited channel resources.In addition,due to the periodic data needs to be delivered within a certain time threshold in industrial wireless sensor networks,there is no research on both optimizing the real-time performance of data delivery and improving the timeliness of periodic data.Therefore,in view of the above problems,this thesis studies the wireless network scheduling method based on Ao I for the industrial wireless sensor network with periodic and aperiodic data hybrid update.The main work includes:1.The research significance of Ao I-based network optimization data delivery in industrial wireless sensor networks is analyzed,and the relevant theoretical basis of Ao I-optimization is introduced.Further,according to the characteristics of network data hybrid update,the Ao I iterative model of data is established.2.To improve the freshness of network data and optimize network throughput,a scheduling method based on deep reinforcement learning is proposed in this thesis.Firstly,the optimization problem of minimizing the weighted sum of system average Ao I and throughput is expressed as Markov Decision Process and solved by relative value iteration.Secondly,for the problem that cannot be implemented due to dimensional disaster in the process of solving,the deep reinforcement learning method is used to learn the optimal scheduling strategy while reducing the state space dimension of the optimization problem,so as to obtain a scheduling method based on deep reinforcement learning.3.Considering that periodic data needs to be delivered within the time threshold,this thesis reduces the overdue probability of periodic data exceeding the threshold while optimizing the real-time performance of network data delivery,and proposes a deep reinforcement learning scheduling method based on optimal decision exploration.Firstly,deep reinforcement learning is used to solve the joint optimization problem of minimizing the weighted sum of average Ao I and periodic data overdue probability.Secondly,to optimize the speed of learning the optimal scheduling strategy by the traditional deep reinforcement learning method,the decision exploration mechanism of the method is improved,and a deep reinforcement learning scheduling method based on optimal decision exploration is proposed.Simulation results show that the scheduling method based on deep reinforcement learning can effectively optimize the freshness of network delivery data and improve network throughput,while the deep reinforcement learning scheduling method based on optimization decision exploration can improve the real-time performance of network data delivery and ensure the timeliness of periodic data.Therefore,the research work in this thesis has certain reference value for the development of real-time scheduling technology in industrial wireless sensor networks.