A high information timeliness data transmission method based on two-stage polling

Abstract

The present invention provides a high information timeliness data transmission method based on two-stage polling. The industrial sensor network model based on two-stage polling and data aggregation includes sensors, aggregators and monitors. The high information timeliness data transmission method based on two-stage polling includes the following steps: Step 1: Divide the sensor nodes into K groups according to the spatial correlation between the sensor nodes; Step 2: Configure an aggregator for each group of sensors; Step 3: The sensor nodes send information updates to the aggregator in polling. The sensor nodes in each group transmit data information to the aggregator according to the polling cycle of the aggregator. This polling is a first-level polling; Step 4: The aggregator collects the data of the nodes in the group and aggregates them; Step 5: The aggregator forwards the aggregated data to the monitor by polling. This polling is a second-level polling. The present invention effectively improves the timeliness of the sensor network system.

Description

Technical Field

[0001] The invention relates to the field of data transmission, and in particular to a polling method for data transmission. Background Art

[0002] Information Age

[0003] Information age is quite different from the time delay in traditional communication systems. They measure the timeliness of a system from different perspectives. Time delay is considered from the perspective of the transmitter, describing the time from the generation of information to the completion of transmission, while information age quantifies the timeliness of the system at the receiving end, which refers to the time from the generation of information received by the receiving end to the present. In real-time monitoring or control systems, not only fast information transmission speed is required, but more attention is paid to the timeliness of the information received by the terminal, because outdated information cannot provide sufficient and effective information for the terminal's next decision. Therefore, information age can effectively quantify the timeliness of terminal information and play an important role in real-time systems.

[0004] Consider Figure 1 The information update system shown in the figure consists of three parts: a sending end, a channel, and a destination node D. The sending end includes a source node S, a sending server, and a buffer queue. Assume that the real-time sampling information generation of the source node obeys the random process X(t), and the destination node needs the real-time information of the source node. Therefore, the source node needs to continuously send information update packets through the sending server, and then reach the destination node through the communication link. The data of each information update packet includes the sampling information X(t of the source node). i ) and the generation time t of the i-th information update package i When the sending server is in a busy state, that is, the sending server is currently transmitting the information update packet, the newly arrived information update packet will enter the buffer queue and enter the sending server in the order of arrival for data transmission. Assuming that the information update packet is transmitted by the source node transmitter, it will be sent to the destination node through an error-free communication link. When the information update packet arrives at the destination node, it means that the receiving end has completed the information update, that is, the information age of the system is updated at the same time.

[0005] exist Figure 1 In the information update system in , all information update packets i=1,2,3… generated by the source node need to be transmitted to the destination node. Assume that the time from the generation of the information update packet to the arrival queue can be ignored, that is, the generation of the information update packet and the arrival of the information update packet to the queue have the same meaning in this model. The generation of the information update packet is modeled as a Poisson process with an average rate of λ, and the transmission time of the data packet follows an exponential random process with an average rate of μ.

[0006] The information age of the system at any time describes the time that has passed since the latest information update packet of the destination node was generated. Assume that the information update packet is generated at time t1, t2, …, t i ,…generate, accordingly, at time t′1,t′2,…,t′ i ,…arrives at the destination. At time t, the maximum index of the information update packet received by the destination is:

[0007] k=max{i|t′ i ∈t} (0-10)

[0008] The latest update timestamp is:

[0009] U(t)=t k (0-11)

[0010] At time t, the mathematical expression of information age is as follows:

[0011] Δ(t)=tU(t) (0-12)

[0012] Figure 1 The system information age changes corresponding to the system are as follows Figure 2 The time when the i-th information update packet is generated and received is t i and t′ i To indicate. i S represents the arrival interval between the i-th and i+1-th information update packets, and its size is related to the arrival rate λ of the information update packets. i Y represents the service time of the i-th information update packet, and its size depends on the service rate of the sending server. i Indicates the time interval between the receiving end receiving the i-th and i+1-th information update packets. i Indicates the maximum information age reached when the receiver receives the i-th information update packet. i represents the trapezoidal area generated when the i-th data packet is received, for example Figure 2 The shaded area Q3 represents the trapezoidal area generated when the third information update packet is received. Δ0 represents the system initial information age.

[0013] Depend on Figure 2 It can be seen that the information age curve of the system is sawtooth-shaped. When a new information update package arrives, the information age of the system is updated and the information age decreases. In the process where no information update package arrives, the information age tends to increase linearly.

[0014] The average information age of the system is expressed as follows:

[0015]

[0016] Polling Scheduling Strategy

[0017] When there are multiple update sources in the information update system, all update sources send information update packages in sequence according to a predetermined order to perform information update.

[0018] In industrial intelligent production and monitoring, it is necessary to ensure the real-time collection of production environment data and the accuracy and reliability of control instructions. Information age is proposed in the application scenario of high information timeliness. Compared with delay, it can measure the timeliness of industrial sensor networks more comprehensively and effectively. Due to the high density of sensor nodes (the high density here means as many as possible, but not wasteful, and does not need to be a fixed value. It is a numerical range selected by technicians in the technical field according to actual needs), there is a lot of redundancy in the data collected by adjacent sensor nodes. The transmission of a large amount of redundant data will lead to problems such as decreased system information timeliness. At present, the research on information age mainly focuses on different system models and information transmission strategies, as well as the impact of information update packet loss caused by interference on the system information age performance. Moreover, the current research in this field considers using data correlation to improve the system information age performance, and the current research is limited to single-stage information update systems. Summary of the invention

[0019] In order to solve the problems in the prior art, the present invention provides a high information timeliness data transmission method based on two-stage polling. The industrial sensor network model based on two-stage polling and data aggregation includes a sensor, an aggregator and a monitor. The high information timeliness data transmission method based on two-stage polling includes the following steps:

[0020] Step 1: Divide the sensor nodes into K groups according to the spatial correlation between the sensor nodes;

[0021] Step 2: Configure an aggregator for each group of sensors;

[0022] Step 3: The sensor nodes poll and send information updates to the aggregator. The sensor nodes in each group transmit data information to the aggregator according to the polling cycle of the aggregator. This polling is a first-level polling. In each aggregator polling cycle, the selected node of each group of sensor polling transmits data information to the aggregator.

[0023] Step 4: The aggregator collects data from nodes in the group and aggregates them;

[0024] Step 5: The aggregator forwards the aggregated data to the monitor through polling. This polling is a secondary polling.

[0025] As a further improvement of the present invention, step 6 is also included: modeling the system as a finite-state discrete Markov process, analyzing the detailed parameter information corresponding to the state transition of the system and the Markov process, and solving and analyzing the average information age of the system with the help of random hybrid system theory.

[0026] As a further improvement of the present invention, a random hybrid system models a queuing system as a finite state (q(t), x(t)) change process, where q(t)∈Q={0,1,2,…,m} represents a continuous time finite state Markov chain used to describe various queuing situations of the queuing system, and x(t)=[x0(t),x1(t)...x n (t)]∈R 1x(n+1) It is also a continuous process, which is used to record the age change of the corresponding process at the receiving end. Therefore, the source to be solved can be marked, and then the information age of the marked source can be tracked at the receiving end with the update of the system state. x(t) is a vector composed of linearly growing piecewise functions, and its growth rate can be solved by the following formula:

[0027]

[0028] where b q =[b q,0 b q,1 …b q,n ] is a vector, where b q,j ∈{0,1}, if in state q, the age process x j (t) linear growth, b q,j The value is 1, otherwise it is 0;

[0029] The above Markov chain is represented by a graph G(Q,L), where the node q∈Q represents the Markov state q, and the line l∈L represents the state from q′ l Transfer to q′ l ; The difference from the traditional Markov process is that there is the possibility of self-transition in this Markov chain, that is, from the current state to the current state; when the information update packet arrives or the transmission is completed, the state transition will occur, and the continuous state x(t) will update x'(t), that is, according to this formula x'=xA l For the transfer matrix A l To solve;

[0030] In order to solve the average information age using the random mixture model, it is necessary to calculate the steady-state probability π of each state of the Markov chain. q(t) and the correlation vector V between the discrete states q(t) q(t) To solve:

[0031] πq(t) =E[δ q,q(t) ]=P(q(t)=q) (0-14)

[0032] V q(t) =E[x(t)δ q,q(t) ]=[v q0 (t)…v qn (t)] (0-15)

[0033] Using L' q represents all transition links into state q, L q represents all the transition links out of state q, that is:

[0034] L' q ={l∈L|q′ l =q} (0-16)

[0035] L q ={l∈L|q l =q} (0-17)

[0036] The information age analysis will follow the ergodicity assumption of the Markov chain q(t), and the Markov state vector π(t) = [π0(t), π1(t)…π m (t)] will converge to the steady-state state vector at the same time The following relationship is satisfied:

[0037]

[0038]

[0039] Under the assumption of Markov chain ergodicity, the state-related vector v q(t) Converges to a non-negative limit For any state q, when t→∞, the following conclusion always holds:

[0040]

[0041] The final average information age can be solved by the following formula:

[0042]

[0043] As a further improvement of the present invention, the sensors are grouped according to their position characteristics.

[0044] As a further improvement of the present invention, in step 4, since the data of the same group of sensor nodes have a strong correlation, each time the aggregator collects data information of the sensor nodes, it aggregates the information using the collected data information and the stored historical data information.

[0045] As a further improvement of the present invention, the data transmitted from the sensor node to the aggregator is transmitted using short packets.

[0046] As a further improvement of the present invention, the sensor nodes are deployed in high density.

[0047] As a further improvement of the present invention, the secondary polling is driven by the primary polling. When the information update package of the primary polling is transmitted to the aggregator, the aggregator processes the received data in a timely manner and immediately sends it to the monitor.

[0048] The beneficial effects of the present invention are:

[0049] This paper clusters sensor nodes according to spatial correlation and obtains an industrial sensor network model based on two-level polling and data aggregation. The correlation between update sources is used to assist a series of related sources to update information, which can effectively improve the timeliness of the sensor network system. In addition, considering the small amount of industrial production data, the system uses short packet technology for data transmission, and considers three coping strategies when errors occur in data packet transmission: no retransmission, retransmission without resampling, and retransmission with resampling. The theoretical expression of the average information age of the system under the three retransmission strategies is solved, and the research results can effectively guide engineering practice.

[0050] Due to the high-density deployment of sensor nodes, the data between adjacent nodes has a strong correlation. The two-stage polling industrial sensor information update system reduces the amount of transmitted data by utilizing the correlation between sensor nodes to achieve the purpose of reducing the timeliness of the system. The sensor nodes are clustered according to spatial correlation, and the aggregator polls to send information updates to the monitor. Since the data between sensor nodes in each cluster has a strong correlation, the polling selects the nodes in the cluster to send the collected data to the aggregator. The aggregator aggregates the received data and the saved data to obtain valid data, and then sends it to the monitor. Therefore, in one polling cycle, each related cluster only needs one node to update the information, and all sensor nodes do not need to send information updates. Compared with the single-stage polling system, all sources must send information updates in one polling cycle. The two-stage polling information update system reduces the amount of information transmission and improves the system information age performance. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Figure 1 It is a model diagram of a single-source single-server information update system;

[0052] Figure 2 It is a schematic diagram of the instantaneous information age of the system;

[0053] Figure 3It is a sensor information update system model diagram of a high information timeliness data transmission method based on double-stage polling of the present invention;

[0054] Figure 4 It is a schematic diagram of the structure of the dual-stage polling information updating system model of the present invention;

[0055] Figure 5 It is the age graph of the secondary polling information updating system of the present invention;

[0056] Figure 6 It is a Markov state transition diagram of the no-retransmission strategy system of the present invention;

[0057] Figure 7 It is a comparison chart of information age performance between the dual-stage polling and single-stage polling systems of the present invention;

[0058] Figure 8 It is a graph showing the effect of the retransmission strategy of the present invention on the system information age performance;

[0059] Fig. 9 is a simulation diagram of the system information age under the non-perfect channel condition of the present invention;

[0060] Fig.10 It is a graph showing the effect of the packet length of the data packet of the present invention on the system information age performance;

[0061] Fig.11 It is a secondary polling information update system model. DETAILED DESCRIPTION

[0062] The present invention will be further described below in conjunction with the accompanying drawings.

[0063] Industrial sensor network model based on two-level polling and data aggregation Figure 3 As shown in the figure, it consists of sensors, aggregators and monitors. In this system, sensors need to send the collected data information to the monitor as quickly as possible. Since the sensor data in the smart factory has a strong spatial correlation, the sensors are evenly divided into K groups according to their location characteristics. After the aggregator aggregates the collected information updates and the saved information, it uses a polling scheduling strategy to cyclically send the information updates of the sensor group to the monitor. Since the sensor nodes in each cluster have a strong correlation, the sensor nodes in the cluster use a polling scheduling strategy to cyclically send information updates to the aggregator. The aggregator polls to aggregate the collected sensor data and then passes it to the monitor. When the transmission from the aggregator to the monitor fails, assuming that the feedback returned from the monitor to the aggregator is error-free, this article considers three packet management strategies:

[0064] (1) No Retransmission: When an error occurs in the transmission of an information update packet, the next sensor directly updates the information as if no error occurred.

[0065] (2) Retransmission Without Resampling: The retransmission without resampling strategy simply retransmits the originally collected data packets without resampling. This is the main difference from the retransmission with resampling strategy. Similarly, when an information update packet fails, it will continue to be retransmitted until the information update packet is successfully transmitted.

[0066] (3) Retransmission With Resampling: When an error occurs in the transmission of an information update packet, the sensor data will be collected again to ensure that the latest data is transmitted. If an error occurs, the retransmission and resampling strategy will be repeated until the information update packet is successfully transmitted.

[0067] In a one-level polling information update system, the aggregator is mainly used to collect and save information updates of a certain related sensor group. Assume that the aggregator is located near the sensor group and can be placed in a location with a better channel to ensure that the transmission from the sensor to the aggregator is reliable, that is, there will be no loss of information update packets. Assume that the sensor has real-time information collection capabilities, that is, the sending server can immediately collect information for the next information update packet after sending the previous information update packet. Therefore, the arrival rate μ of the information update packet is the same as the service rate μ of the sending server. That is to say, in the system considered in this article, the buffer queue of the sending server will always remain idle. Accordingly, the service time (or transmission time) for the sending server to send an information update packet is T b Since the information update packet is encoded as a data packet with a length of m, and the unit can be represented by symbols or channel uses, the transmission time is affected by the length of the data packet, that is, T b =mT u (T u Indicates the duration of a symbol). Since the amount of data transmitted for each update is small, the data packets sent are very suitable for transmission using short packets. The transmission rate of short packets in a Gaussian white noise channel is approximately:

[0068]

[0069] Where C is the channel capacity under Shannon's formula, that is, C = log2(1+γ), which is related to the signal-to-noise ratio γ;

[0070] V——channel dispersion,

[0071] Q -1 (·)——the inverse function of the Q(·) equation, that is,

[0072] By analyzing and processing formula (3-1), we can obtain the relationship between the packet error probability and the packet length m, data volume k and channel signal-to-noise ratio γ as follows:

[0073]

[0074] From formula (3-2), we can know that, assuming that an information update packet uses a short packet to transmit k bits of information, the probability of data packet error will also be affected by the destination receiving signal-to-noise ratio γ and the packet length of the information update packet. When the signal-to-noise ratio remains unchanged, if you want to reduce the probability of packet transmission error, you need to increase the packet length of the data packet. Similarly, when the packet length of the information update packet is fixed, the signal-to-noise ratio at the receiving end can be increased to reduce the probability of data packet transmission error. In addition, through careful observation, it can be found that when the channel conditions are good, for example, when the signal-to-noise ratio at the receiving end reaches 10dB, the data packet will be able to achieve error-free transmission, so the position of the aggregator can be adjusted to make the transmission from the sensor to the aggregator error-free.

[0075] The model of the two-level information update system from aggregator to monitor is as follows Fig.11 As shown. In the two-level polling information update system, the aggregator polls and aggregates the information collected from the sensor group and sends it to the monitor. The aggregation operation of the aggregator means that the aggregator aggregates the data representing the sensor group based on the currently received information update data with the assistance of historical data. In addition, since the aggregation time is extremely small, it will not be considered in this paper. In other words, the aggregator has the ability to aggregate data in real time, so this paper only considers the transmission delay of the system. Therefore, without considering retransmission, the polling period of the system is related to the number of aggregators, denoted as T s =KT b . In the first-level polling information update system, since the aggregator is placed near the sensor group to collect sensor information, the distance is short and the channel conditions are good to ensure reliable transmission. However, the transmission path from the aggregator to the monitor is long, the channel conditions are unstable, and the performance of using short packets to transmit information update packets is greatly affected by channel conditions. On the other hand, due to the regular movement of robots and equipment in the factory production environment, there is a certain probability p that the propagation path is blocked. In order to qualitatively analyze the impact of the blocked information transmission path on the system information age performance, it is assumed that there is a fixed channel attenuation δ due to the blockage.

[0076] Based on the literature research, the relevant information sources can effectively assist in information updating, in order to effectively measure the information age performance of the information update system with a large number of relevant sources. It is assumed that the destination will update the information after receiving the information update package from the same source. Within a certain error range, the information of the relevant update source is also valid and can be fully utilized by the destination. That is, when the destination receives an update from the relevant information source, it will also update the information, and the information age will decrease accordingly. Therefore, the information age of all update sources for the same relevant information source is the same. For example Figure 3 In the sensor group of aggregator 1, when the information update package of any source in the sensor group reaches the destination, the destination will update the information age of all relevant update sources corresponding to the sensor group.

[0077] In summary, Figure 3 In the two-stage polling information update system model shown in the figure, since the information update sent by any sensor in the sensor group is valid for the entire sensor group, the sensors in the sensor group poll for information update, which is the first-level polling mentioned above. The second-level polling is driven by the first-level polling, that is, when the information update package of the first-level polling is transmitted to the aggregator, the aggregator will process the received data in time and send it to the monitor immediately. Since the data aggregation time can be ignored, the overall system model can be simplified as follows Figure 4 Two-level information update system.

[0078] Next, the information age performance in the secondary polling information update system under the no-retransmission strategy will be analyzed. Since the information transmission from the aggregator to the monitor is unreliable, although each aggregator sends information updates to the monitor in a fixed order and periodically polls, the information age of the monitor does not change periodically. On the one hand, due to the use of short packets for data transmission, the probability of data packet error ε(k, m, γ) depends on the channel conditions, and the error probability increases rapidly when the channel conditions are poor. On the other hand, due to the interference of robots in the factory, there is a certain probability q that the channel gain will be attenuated δ. Since the occurrence of data packet error events depends on the mechanism of using short packets for data transmission, and the event of channel obstruction depends on the factory environment, these two events are independent. Therefore, the probability of successful transmission from the aggregator to the monitor is solved as follows:

[0079]

[0080] The effective information age of the sensors in the monitor changes as Figure 5 As shown, Δ0 represents the initial age of the information update packet, that is, the time it takes for the information update packet to reach the aggregator from the sensor. The time it takes for the information update packet to be aggregated in time by the aggregator and then transmitted again to reach the monitor is represented by 2T bSince the transmission from the aggregator to the monitor is unreliable and there is a possibility of data update packet loss, the effective information age observed by the monitor will not be updated periodically like the information age observed at the aggregator, for example Figure 5 The information update packet sent from the aggregator to the monitor at time t2 is lost, so the effective information age at the monitor is not updated and will continue to grow linearly.

[0081] for Figure 5 To solve the effective information age of the sensors monitored by the monitor, a random hybrid system will be used. For the no-retransmission strategy, we can model the information update system as Figure 6 The Markov chain shown in the figure. State 1 in the figure indicates that the cluster where the sensor we want to solve is located is updating information, and state 2-K indicates that the clusters where other sensors are located are updating information. Figure 6 In the i , i=1,2,…,K+1 indicates that the state of the system has changed, and the transfer rate is denoted as λ (l) , and the system state changes, that is, x'=xA l From formula (3-13), we can know that in order to solve v q , we need to change the state transfer matrix A l To analyze, Figure 6 The detailed changes of various parameters during the Markov chain transfer process are summarized in Table 3-1.

[0082] Table 3-1 No retransmission strategy state transition

[0083]

[0084]

[0085] The physical meanings of each transfer process in Table 3-1 are as follows:

[0086] (1) l1 indicates that the information update packet of the sensor whose information age is to be determined is successfully transmitted, and the information age of the receiving end is updated to x'0 = x1. Since after the information update packet of the current sensor is transmitted, a new information update packet will arrive in the next polling cycle, so x'1 = 0.

[0087] (2) l2 indicates that the information update packet of the sensor whose information age is to be solved fails to be transmitted, and the information age of the receiving end is not updated and continues to grow linearly, that is, x'0 = x0. Since after the information update packet of the current sensor is transmitted, a new information update packet will arrive in the next polling cycle, so x'1 = 0.

[0088] (3)l iIndicates that the information update packets of other sensors have completed transmission, where 2≤i≤K. Since the transmission of these sensor information update packets does not affect the information age of the sensor to be solved, and the states before and after the failed and successful transfers are the same, successful transmissions and transmission errors are considered together. The information age of the receiving end is not updated and continues to grow linearly, that is, x'0=x0. Since the information update packets of the sensors to be solved have not arrived after the transmission of these sensors' information update packets, x′1=0.

[0089] (4)l K+1 The information update packet of the irrelevant sensor has completed the transmission, and it also does not affect the information age of the sensor to be solved. Successful transmission and transmission errors are also considered together. The information age of the receiving end is not updated and continues to grow linearly, that is, x'0 = x0. Since the information update packet to be solved has just arrived at the sending server at this time, the age is 0, that is, x'1 = 0.

[0090] From formula (3-5), we can see that the growth of the information age x(t) at the receiving end depends on b q , that is, x'=b q Because the information age to be solved is tracked in real time by x0(t), there is b in all states q1 = 1. If the information update packet being transmitted comes from the sensor to be solved, then b q2 = 1, because when the information update packet of the sensor to be solved is transmitted, the information age of the receiving end will be updated to x1(t), that is, b q2 The role of x1(t) and x1(t) is to record the transmission time of the information update packet to be solved, as follows:

[0091]

[0092] Through the above analysis of the transfer process, we get the information age tracking process x' after the state transfer, and then we can get x'=xA l Find A in Table 3-1 l Finally, through simple calculation, we can get the last column v in Table 3-1 ql A l .

[0093] Substituting the corresponding data in the table into equations (3-10) and (3-11), we can get the Markov chain steady-state probability:

[0094]

[0095] Substituting equations (3-15) and (3-16) into equation (3-12), we can obtain ql The equation is as follows:

[0096]

[0097] By solving the equations represented by formula (3-17), we can get v q0 :

[0098]

[0099] Substituting equation (3-18) into equation (3-13), we can obtain the average information age Δ of the system under the non-retransmission strategy: 不重传 The expression is:

[0100]

[0101] Similarly, the expressions of the average information age of the system under the retransmission without resampling strategy and the retransmission with resampling strategy can be obtained:

[0102]

[0103]

[0104] Stochastic Hybrid Systems Theory

[0105] Since there is a possibility of error in the information update from the aggregator to the monitor, in order to simplify the computational complexity of the information age solution process, this section will use a random hybrid system to solve the information age of the two-stage polling system. Therefore, this section will give a brief introduction to the random hybrid system.

[0106] The random hybrid system models a queue system as a finite state (q(t), x(t)) change process, where q(t)∈Q={0,1,2,…,m} represents a continuous time finite state Markov chain used to describe various queue conditions of the queue system, and x(t)=[x0(t),x1(t)...x n (t)]∈R 1x(n+1) , R represents the vector space of 1*(n+1), which is also a continuous process, used to record the age change of the corresponding process at the receiving end. Therefore, the source to be solved can be marked, and then the information age of the marked source can be tracked at the receiving end with the update of the system state [59,60]. x(t) is a vector composed of linearly growing piecewise functions, and its growth rate can be solved by the following formula:

[0107]

[0108] where b q =[b q,0 b q ,1…b q,n ] is a vector, where b q,j∈{0,1}, if in state q, the age process x j (t) linear growth, b q,j The value is 1, otherwise it is 0.

[0109] The above Markov chain is represented by a graph G(Q,L), where the node q∈Q represents the Markov state q, and the line l∈L represents the state from q l Transfer to q′ l The difference from the traditional Markov process is that this Markov chain has the possibility of self-transition, that is, transition from the current state to the current state. When the information update packet arrives or the transmission is completed, the state transition will occur, and the continuous state x(t) will be updated to x'(t), that is, according to this formula x'=xA l For the transfer matrix A l To solve.

[0110] In order to solve the average information age using the random mixture model, it is necessary to calculate the steady-state probability π of each state of the Markov chain. q(t) and the correlation vector V between the discrete states q(t) q(t) To solve:

[0111] π q(t) =E[δ q,q(t) ]=P(q(t)=q) (3-6)

[0112] V q(t) =E[x(t)δ q,q(t) ]=[v q0 (t)…v qn (t)] (3-7)

[0113] Using L' q represents all transition links into state q, L q represents all the transition links out of state q, that is:

[0114] L' q ={l∈L|q′ l =q} (3-8)

[0115] L q ={l∈L|q l =q} (3-9)

[0116] The information age analysis follows the ergodicity assumption of the Markov chain q(t)[8], where the Markov state vector π(t) = [π0(t), π1(t)…π m (t)] will converge to the steady-state state vector at the same time The following relationship is satisfied:

[0117]

[0118]

[0119] It can be obtained from the literature that under the assumption of Markov chain ergodicity, the state-related vector v q(t) Converges to a non-negative limit For any state q, when t→∞, the following conclusion always holds:

[0120]

[0121] The final average information age can be solved by the following formula[8]:

[0122]

[0123] In order to compare the average information age performance gap between the two-stage polling system and the single-stage polling system and highlight the superiority of the two-stage polling system, the simulation assumes that the number of sensors in both systems ranges from 10 to 3000, the amount of data transmitted each time is 150 bits, the length of the information update packet is 200 channels used, the number of groups is 5, and other simulation parameters are set to γ ​​= 10dB, p = 0.1, δ = 5dB, T u =0.006ms, some parameters of which were selected based on references.

[0124] Figure 7 The information age performance of the two-stage polling system and the single-stage polling system is simulated and analyzed and compared. It can be seen from the figure that the system information age is approximately linearly related to the number of sensors, which is consistent with the conclusion of formula (3-19). Compared with the information timeliness in multi-source information update based on polling scheduling in the literature, the two-stage polling strategy in this paper reduces the transmission of redundant data and shortens the overall polling time of the system by aggregating data using correlation after the first-stage polling. Therefore, the two-stage polling scheduling strategy assisted by the aggregator has significantly improved the information timeliness. With the increase of information update sources, the information timeliness improvement effect of the method adopted in this study is more obvious.

[0125] Figure 8 The simulation analyzes the impact of three retransmission strategies, namely, no retransmission, retransmission with resampling, and retransmission without resampling, on the information age performance of the two-stage polling system. The parameters in the simulation process are set as N = 100, K = 5, k = 150bit, m = 200 channeluses, γ = 10dB, p = 0.1, σ = 5dB, T u =0.006ms. Figure 8It can be seen that when the channel conditions are good, even if a transmission error occurs, the transmission will most likely succeed in the next polling, which has little impact on the system information age performance. Therefore, the system information age performance under the three strategies is basically the same. In addition, since the retransmission resampling strategy performs resampling each time it retransmits, the performance of the retransmission resampling strategy is slightly better than the retransmission without resampling strategy. When the channel conditions are poor, the retransmission strategy can ensure the timely delivery of information update packets and effectively improve the system information age performance.

[0126] Fig. 9 The simulation analyzes the influence of channel conditions on the system information age in actual industrial production. The parameters in the simulation process are set as N = 3000, K = 5, k = 150bit, m = 200channel uses, γ = 10dB, p = {0.1, 0.2, 0.3, 0.4}, δ = 0-20dB, T u =0.006ms. Fig. 9 It can be seen that when the probability of channel interference p is constant, when the channel gain loss is small, it has almost no effect on the average information age of the system. When the channel loss reaches 6dB, the information age of the system increases rapidly. When the channel loss is close to 9dB, the information age of the system does not change with the attenuation of the channel gain. The fundamental reason for the above phenomenon is that the second level of the two-stage polling information update system uses short packets for data transmission. Since the success rate of using short packets for data transmission changes rapidly when the channel conditions reach the critical value, when the channel conditions drop to a certain threshold, the probability of data packet errors increases rapidly, resulting in a rapid increase in the information age of the system.

[0127] Fig.10 The simulation analyzes the effect of data packet length on the effective information age of the system. The experiment simulates the effective information age of the system under different data packet lengths. The parameters are set as N = 3000, K = 5, k = 150bit, m = 150-300channel uses, γ = 10dB, p = 0.2, δ = 5dB, T u=0.006ms. It can be seen from the figure that when the data packet length is large, the information age performance of the system is less affected by the channel conditions. When the data packet length is small, the system timeliness drops sharply as the channel conditions deteriorate. In addition, as the signal-to-noise ratio increases, the length of the transmission data packet with the lowest system timeliness gradually decreases. From formula (3-2) and formula (3-19), it can be found that the root causes of the above phenomenon are mainly the following two aspects. In a system using short packets for data transmission, when the data packet length increases, on the one hand, the transmission error probability of the information update packet can be reduced, thereby reducing the average information age of the system. On the other hand, when the data packet increases, the transmission time and polling time of the data update packet will increase, thereby increasing the average information age of the system. Therefore, the optimal data packet length will depend on the information transmission conditions in the actual scenario.

[0128] The above contents are further detailed descriptions of the present invention in combination with specific preferred embodiments, and it cannot be determined that the specific implementation of the present invention is limited to these descriptions. For ordinary technicians in the technical field to which the present invention belongs, several simple deductions or substitutions can be made without departing from the concept of the present invention, which should be regarded as falling within the protection scope of the present invention.

Claims

1. A high information timeliness data transmission method based on two-stage polling, characterized in that: The industrial sensor network model based on two-stage polling and data aggregation includes sensors, aggregators and monitors. The high information timeliness data transmission method based on two-stage polling includes the following steps: Step 1: Divide the sensor nodes into K groups according to the spatial correlation between the sensor nodes; Step 2: Configure an aggregator for each group of sensors; Step 3: The sensor nodes poll and send information updates to the aggregator. The sensor nodes in each group transmit data information to the aggregator according to the polling cycle of the aggregator. This polling is a first-level polling. In each aggregator polling cycle, the selected node of each group of sensor polling transmits data information to the aggregator. Step 4: The aggregator collects data from nodes in the group and aggregates them; Step 5: The aggregator forwards the aggregated data to the monitor through polling. This polling is a secondary polling. Step 6: Model the system as a finite-state discrete Markov process, analyze the detailed parameter information corresponding to the state transition of the system and the Markov process, and solve and analyze the average information age of the system with the help of random hybrid system theory.

2. The high information timeliness data transmission method based on double-stage polling according to claim 1 is characterized in that: The random hybrid system models a queue system as a finite state (q(t), x(t)) change process, where q(t)∈Q={0,1,2,…,m} represents a continuous time finite state Markov chain, which is used to describe various queue conditions of the queue system, and x(t)=[x0(t,x1(t)...x n (t)]∈R 1x(n+1) It is also a continuous process, which is used to record the age change of the corresponding process at the receiving end. Therefore, the source to be solved can be marked, and then the information age of the marked source can be tracked at the receiving end with the update of the system state. x(t) is a vector composed of linearly growing piecewise functions, and its growth rate can be solved by the following formula: where b q =[b q，0 b q，1 …b q，n ] is a vector, where b q，j ∈{0, 1}, if in state q, the age process x j (t) linear growth, b q，j The value is 1, otherwise it is 0; The above Markov chain is represented by a graph G(Q, L), where the node q∈Q represents the Markov state q, and the line l∈L represents the state from q l Transfer to q′ l ; The difference from the traditional Markov process is that there is the possibility of self-transition in this Markov chain, that is, from the current state to the current state; when the information update packet arrives or the transmission is completed, the state transition will occur, and the continuous state x(t) will update x′(t), that is, according to this formula x=xA l For the transfer matrix A l To solve; In order to solve the average information age using the random mixture model, it is necessary to calculate the steady-state probability π of each state of the Markov chain. q(t) and the correlation vector V between the discrete states q(t) q(t) To solve: π q(t) =E[δ q，q(t) ]=P(q(t)=q) (0-2) V q(t) =E[x(t)δ q，q(t) ]=[v q0 (t)…v qn (t)] (0-3) Using L′ q represents all transition links into state q, L q represents all the transition links out of state q, that is: L′ q ={l∈Lq′ l =q} (0-4) L q ={l∈L q l =q} (0-5) The information age analysis will follow the ergodicity assumption of the Markov chain q(t), and the Markov state vector π(t) = [π0(t), π1(t)…π m (t)] will converge to the steady-state state vector at the same time The following relationship is satisfied: Under the assumption of Markov chain ergodicity, the state-related vector v q(t) Converges to a non-negative limit For any state q, when t→∞, the following conclusion always holds: The final average information age can be solved by the following formula:

3. A high information timeliness data transmission method based on double-stage polling according to any one of claims 1 to 2, characterized in that: Sensors are grouped according to location characteristics.

4. A high information timeliness data transmission method based on double-stage polling according to any one of claims 1 to 2, characterized in that: In step 4, since the data of the same group of sensor nodes have a strong correlation, each time the aggregator collects data information of the sensor nodes, it aggregates the information using the collected data information and the saved historical data information.

5. A high information timeliness data transmission method based on double-stage polling according to any one of claims 1 to 2, characterized in that: The data transmitted from the sensor node to the aggregator is transmitted in short packets.

6. A high information timeliness data transmission method based on double-stage polling according to any one of claims 1 to 2, characterized in that: The sensor nodes are deployed in high density.

7. A high information timeliness data transmission method based on double-stage polling according to any one of claims 1 to 2, characterized in that: The secondary polling is driven by the primary polling. When the information update package of the primary polling is transmitted to the aggregator, the aggregator will process the received data in a timely manner and send it to the monitor immediately.

Images