An efficient time synchronization method based on air interface to application layer TSN

Abstract

The application relates to an efficient time synchronization method based on air interface to application layer TSN, comprising the following steps: constructing an end-to-end TSN transmission model from the air interface to the application layer; reconstructing based on clock source historical data of each transmission node; constructing an optimal master clock selection model based on a Light GBM model; and constructing an efficient time synchronization model based on DDQN. The application constructs a reinforcement learning model for each clock source by using DDQN, combines the interaction process of each clock source with the environment, and completes the training of the time synchronization model. While ensuring the accuracy of clock synchronization, the time delay required in the synchronization process is greatly reduced.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of energy internet TSN time synchronization, and particularly relates to an efficient time synchronization method based on air interface to application layer TSN. BACKGROUND

[0002] With the rapid development of power internet of things and big data industry, time sensitive network represented by TSN is widely used in the energy internet of power system. Meanwhile, in the traditional time sensitive network, data transmission only considers the limited transmission mode of Ethernet, while in the development of modern communication, wireless communication is often indispensable, therefore, it is an urgent problem to be solved to extend TSN to the field of wireless communication. With the increase of network node devices, how to intelligently select the master clock source to ensure the low delay characteristics of the whole network, and efficiently and accurately synchronize all clocks is still an open and difficult problem to be solved. SUMMARY

[0003] The purpose of the embodiment of the present application is to provide an efficient time synchronization method based on air interface to application layer TSN, to realize intelligent master clock source selection and efficient clock synchronization.

[0004] To achieve the above purpose, the present application provides the following technical scheme:

[0005] The embodiment of the present application provides an efficient time synchronization method based on air interface to application layer TSN, comprising the following steps:

[0006] An end-to-end TSN transmission model from air interface to application layer is constructed;

[0007] Based on the historical data reconstruction of the clock source of each transmission node;

[0008] Optimal master clock selection model based on Light GBM model;

[0009] Efficient time synchronization model construction based on DDQN.

[0010] The end-to-end TSN transmission model from air interface to application layer is constructed, which considers K wireless terminals, two wired routing nodes and a server. The K wireless terminals collect data and transmit wirelessly, when reaching the first routing node, they transmit by wired Ethernet, and a routing node is added in the middle to strengthen data transmission, and finally the data collected by the wireless terminal is transmitted to the server. Among them, the K wireless terminals are represented as R k ∈{R1,…,R K}, and the corresponding clock source is represented as clock k ∈{clock1,…,clockK}。In the process of Ethernet transmission, the first route is represented as RL1, and the corresponding clock source is clock K+1 , the second route is represented as RL2, and the corresponding clock source is clock K+2 .

[0011] The reconstruction based on the clock source history data of each transmission node is specifically that, in the clock source adjustment process of each transmission node, the clock adjustment amount is denoted as ΔT, and the corresponding signaling operation is denoted as M. In the reconstruction based on the clock source history data of each transmission node, the clock adjustment amount and the corresponding signaling operation M constitute a tuple, that is, data_clock k =(ΔT, M). The data of all clock sources at each time is combined with the corresponding signaling operation to obtain the final data, that is, data={data_clock1,…, data_clock K+2}.

[0012] The optimal master clock selection model based on Light GBM is supervised learning, and the data obtained also needs to be labeled. The processing method is to add a field label after each group of data, that is, the master clock selected at the current time. Subsequently, the most optimal master clock selection model based on Light GBM is constructed, and the model is represented as:

[0013] clock lightgbm =f lightgbm (data) (1)

[0014] Wherein, clock lightgbm is the master clock selected by the Light GBM model, and data is the reconstructed data. The optimization objective in the training process is represented as:

[0015] minimize L(clock opt ,clock lightgbm ) (2)

[0016] Wherein, clock opt is the optimal master clock selected in the historical data, and clock lightgbm is the master clock selected by the Light GBM model. Through the above method, the optimal master clock selection can be obtained.

[0017] The efficient time synchronization model based on DDQN is constructed as follows:

[0018] After obtaining the optimal master clock according to the method of claim 2, for each clock source, the efficient time synchronization model based on DDQN is constructed, and the corresponding Markov problem is represented as:

[0019] max Q π (s,a; θ) = E[R1+ γR2+ … | S, A, π; θ] (3)

[0020] Where Q π (s,a; θ) is the maximization of cumulative expected reward, R represents the reward under each corresponding action, A is the action taken at each time, π is the policy formulated by DDQN, and γ is the decay coefficient. The update method of the corresponding selection network and evaluation network is:

[0021] Y t Q = R t+1 + γQ(S t+1 , argmaxQ(S t+1 , a; θ t ); θ t ) (4)

[0022] Y t DoubleQ = R t+1 + γQ(S t+1 , argmaxQ(S t+1 , a; θ t ); θ t ') (5)

[0023] Therefore, after designing the selection network and evaluation network under DDQN, combining the master clock determined in claim 2, each signaling is converted into the corresponding state, that is, S. The time adjustment amount each time is converted into the corresponding action, that is, A. After training, the time synchronization model based on DDQN can be obtained.

[0024] Compared with the prior art, the beneficial effects of the present application are:

[0025] The deficiencies of the prior art time-sensitive network in few wireless network applications, low efficiency of master clock determination and clock synchronization are solved, and an efficient time synchronization method based on air interface to application layer TSN is provided, which effectively improves the time synchronization efficiency of air interface to application layer TSN. BRIEF DESCRIPTION OF DRAWINGS

[0026] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following will briefly introduce the drawings needed to be used in the embodiments of the present application. It should be understood that the following drawings only show some embodiments of the present application, and therefore should not be regarded as a limitation on the scope, and for those skilled in the art, other related drawings can also be obtained without creative labor on the basis of these drawings.

[0027] Fig. 1A TSN network diagram based on air interface to application layer is proposed;

[0028] Fig. 2 A high-efficiency time synchronization method based on air interface to application layer TSN is proposed,

[0029] Fig. 3 A performance comparison diagram of the method and all device clock updating proposed in the patent. DETAILED DESCRIPTION

[0030] The technical solutions in the embodiments of the present application will be described below with reference to the drawings in the embodiments of the present application. It should be noted that similar reference numerals and letters represent similar items in the following drawings, so once an item is defined in one drawing, it does not need to be further defined and explained in subsequent drawings.

[0031] The term "comprising" or "including" or any other variant thereof is intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements not only includes those elements, but also includes other elements not explicitly listed or inherent to such a process, method, article or device. Without more limitations, the element defined by the statement "including a" does not exclude the presence of another identical element in the process, method, article or device including the element.

[0032] Reference Figs. 1-3 A high-efficiency time synchronization method based on air interface to application layer TSN mainly includes three links of constructing an end-to-end TSN transmission model from air interface to application layer, reconstructing based on clock source history data of each transmission node and selecting an optimal master clock based on a Light GBM model, and constructing a high-efficiency time synchronization model based on DDQN.

[0033] In the implementation method of the present application, an end-to-end TSN transmission model from air interface to application layer is first constructed. In the high-efficiency time synchronization method based on air interface to application layer TSN designed in the present patent, K wireless terminals are considered to collect information for wireless transmission. When the remote routing node receives the data, it starts to perform wired transmission in the Ethernet mode, and strengthens data transmission through an intermediate routing, and finally transmits the data to the server. Among them, the K wireless terminals are represented as R k ∈{R1,…,R K}, and the corresponding clock source is represented as clock k ∈{clock1,…,clock K}. In the Ethernet transmission stage, the first routing is represented as RL1, and the corresponding clock source is clock K+1, the second route is represented as RL2, and the corresponding clock source is clock K+2 .

[0034] In the implementation method of the present application, secondly, the clock source history data reconstruction based on each transmission node and the optimal master clock selection model based on the Light GBM model are completed. In the clock source history data reconstruction based on each transmission node, for each clock source, in each clock adjustment process of it, the clock adjustment amount ΔT and the corresponding signaling operation M form a tuple, that is, data_clock k =(ΔT,M). The data combination of the clock adjustment amount of all clock sources at each moment and the corresponding signaling is carried out, and the final data obtained is data={data_clock1,…,data_clock K+2}. Since the optimal master clock selection model based on Light GBM is adopted in the present application, it is supervised learning, and the obtained data also needs to be labeled. The processing method is to add a field label after each group of data, that is, the master clock selected at the current moment. Subsequently, the optimal master clock selection model based on Light GBM is constructed, and the model is represented as:

[0035] clock lightgbm =f lightgbm (data) (1)

[0036] Wherein, clock lightgbm is the master clock selected by the Light GBM model, and data is the reorganized data. The optimization objective in the training process is represented as:

[0037] minimize L(clock opt ,clock lightgbm ) (2)

[0038] Wherein, clock opt is the optimal master clock selected under the historical data, and clock lightgbm is the master clock selected by the Light GBM model. Through the above method, the optimal master clock selection can be obtained.

[0039] Finally, for each clock source, the efficient time synchronization model based on DDQN is constructed, and the corresponding Markov problem is represented as:

[0040] max Q π (s,a;θ)=E[R1+γR2+…|S,A,π;θ] (3)

[0041] Wherein, Q π(s, a; θ) is the maximization of cumulative expected reward, R represents the reward under each corresponding action, A is the action taken at each time, π is the policy formulated by DDQN, and γ is the decay coefficient. The update method of the corresponding selection network and evaluation network is:

[0042] Y t Q = R t+1 + γQ(S t+1 , argmaxQ(S t+1 , a; θ t ); θ t ) (4)

[0043] Y t DoubleQ = R t+1 + γQ(S t+1 , argmaxQ(S t+1 , a; θ t ); θ t ') (5)

[0044] Therefore, after designing the selection network and evaluation network under DDQN, combined with the master clock determined in claim 2, each signaling is converted into the corresponding state, that is, S. The time adjustment amount each time is converted into the corresponding action, that is, A. After training, the time synchronization model based on DDQN can be obtained.

[0045] The above only describes the embodiments of the present application and does not limit the protection scope of the present application. For those skilled in the art, the present application can have various changes and modifications. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. An efficient time synchronization method based on air interface to application layer TSN, characterized in that, Comprise the following steps: An end-to-end TSN transmission model from the air interface to the application layer is constructed; Reconstruction based on the clock source history data of each transmission node; Optimal master clock selection model based on Light GBM model; Construction of an efficient time synchronization model based on DDQN; The Light GBM-based optimal master clock selection model is supervised learning, and the data obtained also needs to be labeled. The processing method is to add a field after each set of data , which is the selected master clock at the current time. Subsequently, a Light GBM-based optimal master clock selection model is constructed, and the model is represented as: (1) wherein, The selected master clock for the Light GBM model is data, i.e., the reorganized data, and the optimization objective in the training process is represented as: (2) wherein, is the optimal master clock selected under historical data, is the master clock selected under the Light GBM model, and the optimal master clock selection can be obtained by the above method.

2. The efficient time synchronization method based on air interface to application layer TSN according to claim 1, characterized in that, The end-to-end TSN transmission model from the air interface to the application layer is constructed, which considers K wireless terminals, two wired routing nodes and a server, the K wireless terminals collect and transmit data, when reaching the first routing node, the wired transmission is performed in the form of Ethernet, and a routing node is added in the middle to strengthen data transmission, and finally the data collected by the wireless terminal is transmitted to the server, wherein the K wireless terminals are represented as , and the corresponding clock source is represented as , in the process of Ethernet transmission, the first routing is represented as , the corresponding clock source is , the second routing is represented as , and the corresponding clock source is .

3. The efficient time synchronization method based on air interface to application layer TSN according to claim 1, characterized in that, The reconstruction based on the clock source history data of each transmission node is specifically that, in the clock source adjustment process of each transmission node, the clock adjustment amount is recorded as The corresponding signaling operation is recorded as In the reconstruction based on the clock source history data of each transmission node, the clock adjustment amount and the corresponding signaling operation , constitute a tuple, that is The data combination of the clock adjustment amount and the corresponding signaling of all clock sources at each moment is obtained, and the final data is .

4. The efficient time synchronization method based on air interface to application layer TSN of claim 1, wherein, Construction of an efficient time synchronization model based on DDQN is as follows: After obtaining the optimal master clock, for each clock source, an efficient time synchronization model based on DDQN is constructed, and the corresponding Markov problem is represented as: (3) wherein, is the maximization of the cumulative expected reward, represents the reward under each corresponding action, is the action taken at each time instant, is the policy formulated by the DDQN, is the decay coefficient, and the update mode of the selection network and the evaluation network corresponding thereto is: (4) (5)。

Images