CN101466142A - Synchronization method for layered time comparative clock in wireless sensor network - Google Patents

Abstract

A layered time comparison clock synchronization method in a wireless sensor network in the technical field of wireless networks comprises the following steps: Step 1, the network starts to initialize, and starts from the base station/coordinator according to the signal strength and device types; Step 2, the entire network It is divided into external clock source synchronization and internal clock source synchronization. Nodes are synchronized by time stamp comparison method, and upper and lower nodes are synchronized in server/client mode; Step 3, adjust the synchronization period and clock drift parameters according to the previous synchronization effects; Step 4 , to automatically discard stale timestamps. Compared with the prior art, the invention can better meet the accuracy of the wireless sensor network time synchronization method, and has the advantages of low overhead and robustness.

Description

Hierarchical Time Comparison and Clock Synchronization Method in Wireless Sensor Networks

technical field

The invention relates to a method in the technical field of wireless networks, in particular to a method for layered time comparison and clock synchronization in a wireless sensor network.

Background technique

In recent years, wireless sensor networks (WSNs) composed of low-power wireless sensors have been widely used. Time synchronization is an important part of WSNs applications. Control message conflicts, data fusion, and node positioning all require clock synchronization between nodes.

Many universities and research institutions at home and abroad have conducted in-depth research on the synchronization technology of wireless sensor networks and proposed a variety of time synchronization methods to meet the application needs of sensor networks from different aspects. So far, more than a dozen implementation methods have been proposed, typically RBS, TPSN, LTS, TS/MS, TSync, DMTS, FTSP, AD, TSS and TDP. Although these methods meet the application needs from different aspects, they have many defects correspondingly. For example, RBS has better precision, but it cannot be applied in some network protocols. TS/MS and Tsync have better accuracy, but the method is complicated, which is a bit too large for the limited computing power and network overhead in the wireless network.

After searching the existing technical documents, it was found that Rui Fan, Indraneel Chakraborty, and Nancy Lynch, published in Lecture Notes in Computer Science, Principles of Distributed Systems on August 23, 2005 The article "Clock Synchronization for Wireless Networks" (clock synchronization of wireless networks), this article proposes a synchronization method, which adopts a parallel method of external clock synchronization and internal clock synchronization, by continuously comparing the size of the time coordinates, whichever is the largest (actually It is also the logic time that may be the closest to the external reference time during synchronization, and the synchronization period adopts a fixed period. This method has less computational overhead and communication overhead, avoids the estimation of packet processing delay with the greatest randomness, and avoids the possibility that some algorithms may dial back the clock, but this method is only suitable for stable single-hop The network effect is better, but the synchronization accuracy is poor in a large-scale dynamic network, which is obviously not enough for some practical applications.

Contents of the invention

The object of the present invention is to address the deficiencies of the above-mentioned prior art, and provide a hierarchical time comparison clock synchronization method in a wireless sensor network, which reduces the overhead of synchronous communication through layering and time stamp comparison, and uses estimation to approach clock drift, The method of adjusting the parameter ρ further reduces the overhead of synchronization. At the same time, the invalid time stamp is discarded, and the overhead for node storage is reduced. The present invention can be applied to dynamic multi-hop networks.

The present invention is realized through the following technical solutions, and the present invention comprises the following steps:

Step 1, the network starts to initialize: with the general base station/coordinator as the top layer, the affiliation confirmation frame is sent out according to the signal strength and device type, and the corresponding device receives the confirmation frame and returns a response frame, and so on, to complete the whole Layered setup of the network, while confirming that the network has reached a steady state;

Step 2, start from the top layer and start synchronization layer by layer. The synchronization is divided into two parts: external clock source synchronization and internal clock source synchronization. During the external clock source synchronization process, if node i receives a GPS synchronization signal, the current valid Add one to the array coordinate i.current, and the synchronization of the external clock source provides a time reference for the synchronization of the internal clock source. Each node obtains the logical time of the node according to the comparison of the synchronization results of the internal time source and the external clock source;

Step 3: Each node uniformly sets the synchronization period, and then each node adjusts the clock drift parameter ρ and the synchronization period τ according to the initial synchronization results and synchronization accuracy. overhead; if not satisfied, adjust the clock drift parameter ρ, and reduce the synchronization period τ;

Step 4. Node i checks whether the current valid array coordinates are greater than the array length. If yes, the current valid array coordinates are cleared, and the invalid timestamp is discarded. If not, the current valid array coordinates continue to count, and step 2 is repeated.

Described each node obtains the logic time of this node according to the comparison of internal time source and external clock source synchronization result, comprises the following steps:

In the first step, if node i finds that the logical clock of node i is equal to the product of node i’s scheduled synchronization times and node synchronization period, node i starts to synchronize and checks whether the external clock source exists. If it exists, it sends an external synchronization service request. Enter the second step; no matter whether the external clock source exists or not, the internal clock synchronization is started, that is, the third step is executed, and the number of reservation synchronizations of node i is increased by one, but the 0th layer node does not perform internal clock synchronization;

In the second step, node i performs external clock source synchronization: if node i receives a GPS synchronization signal, the current effective array coordinate i.current of node i is increased by one, and the GPS timestamp is stored in the array i.local[i.current] and In i.max_gps, i.local[i.current] is updated synchronously with the physical clock, i.local[] represents node i’s best estimate of the time coordinate of the node, and i.max_gps represents the latest GPS timestamp;

In the third step, node i performs internal clock synchronization: node i sends a synchronization service request to the directly subordinate node j, if node i receives the synchronization frame sync_message(j.local[j.current], j .max_gps), node i checks whether node j’s GPS timestamp is as new as node i’s i.max_gps, then checks whether j.local[j.current] is greater than i.global[i.current], i.global[ ] indicates that node i best estimates the time of other nodes in the network to which the node belongs. If both are satisfied, it means that node i has an error in estimating i.global[i.current] of other nodes, and then update i.global[i. current] is j.local[j.current], and j.global[j.current] is updated by (1-ρ)/(1+ρ) times of the physical clock;

In the fourth step, the node i compares the elements of the array i.local[], and takes the largest one as i.mlocal, compares the elements of the array i.global[], and takes the largest one as i.mglobal;

In the fifth step, node i compares the size of i.mlocal and i.mglobal, and takes the larger one as i.logical to represent the logical clock of node i, completes node clock synchronization, and i.logical is updated synchronously with physical time.

Compared with the prior art, the present invention includes the following beneficial effects:

1. The present invention avoids the estimation of data packet transmission delay between nodes by comparing time stamps; from the actual wireless network construction and packet capture analysis, it can be known that the transmission delay between nodes is due to channel energy detection and channel conflict detection. The impact of task scheduling is very random, and the use of timestamp comparison bypasses the estimation of these factors, thereby reducing system overhead while ensuring a certain accuracy;

2. The present invention reduces the synchronization error limit between nodes by a layered method; in other words, the synchronization error limit between nodes and the synchronization accuracy of nodes all depend on the maximum delay of the network to which the node belongs, so a large network Layered, divided into several small networks for synchronization, which helps to reduce errors;

3. The present invention reduces the overhead of synchronous communication through the method of variable period; in actual engineering, the crystal oscillators used by each node are quite different, some have good precision, and some have poor precision. If a one-size-fits-all synchronization period is adopted, it will obviously add a lot of overhead to the limited network channel traffic and increase the energy consumption of the entire network, and allowing nodes to make statistics to change the synchronization period can avoid unnecessary energy consumption and traffic overhead;

4. The present invention further reduces the overhead of synchronization by estimating the approximate clock drift and adjusting the parameter ρ. In the previous node synchronization process, it can be seen that if the ρ estimation is accurate, it is obvious that the node has a better estimate of the clocks of other nodes. Accurate, the corresponding synchronization cycle can be set longer, and the natural overhead is small;

5. The present invention discards invalid time stamps by clearing i.Current to reduce the cost of node storage; although with the continuous development of semiconductor technology, on-chip resources are becoming more and more abundant, but for nodes used In terms of single-chip microcomputers, the situation that the resources are relatively limited will not be changed for a period of time in the foreseeable future. In this synchronization mechanism, the two best estimated arrays do not take up much resources in a short period of time, but it is obviously not possible for a long time, and at the same time, the old timestamp is meaningless. So discard the invalid timestamp.

In summary, the present invention can better meet the requirements of moderate accuracy, low overhead and robustness of the infinite sensor network time synchronization algorithm.

Description of drawings

Fig. 1 is the work flowchart of the inventive method;

Fig. 2 is the simulation result figure of embodiment of the present invention;

In the figure, (a) is the simulated unprocessed crystal frequency change diagram, (b) is the logical frequency change diagram after parameter adjustment, and (c) is the comparison histogram of the synchronization period before and after processing.

Detailed ways

The embodiments of the present invention are described in detail below in conjunction with the accompanying drawings: this embodiment is implemented on the premise of the technical solution of the present invention, and detailed implementation methods and specific operating procedures are provided, but the protection scope of the present invention is not limited to the following the described embodiment.

This embodiment is applied to the application of the clock synchronization research of the experimental network in the wireless sensor network construction project. It mainly uses the ASAP1810/20 development board based on Zixin Company and the CC2420 wireless sensor development board of TI, and develops on it Networking experiment.

In this embodiment, the wireless sensor network is composed of 126 slave devices and one master device, wherein the master device is provided with a GPS clock source, and the hierarchical layers of the network are divided by each node according to its physical location and function.

As shown in Figure 1, this embodiment includes the following steps:

Step 1, the network starts to initialize: with the general base station/coordinator as the top layer, the affiliation confirmation frame is sent out according to the signal strength and device type, and the corresponding device receives the confirmation frame and returns a response frame, and so on, to complete the whole Layered setup of the network, while confirming that the network has reached a steady state;

Step 2, start from the top layer and start synchronization layer by layer. The synchronization is divided into two parts: external clock source synchronization and internal clock source synchronization. During the external clock source synchronization process, if node i receives a GPS synchronization signal, the current valid Add one to the array coordinate i.current, and the synchronization of the external clock source provides a time reference for the synchronization of the internal clock source. Each node obtains the logical time of the node according to the comparison of the synchronization results of the internal time source and the external clock source;

Step 3: Each node uniformly sets the synchronization period, and then each node adjusts the clock drift parameter ρ and the synchronization period τ according to the initial synchronization results and synchronization accuracy. overhead; if not satisfied, adjust the clock drift parameter ρ, and reduce the synchronization period τ;

Step 4. Node i checks whether the current valid array coordinates are greater than the array length. If yes, the current valid array coordinates are cleared, and the invalid timestamp is discarded. If not, the current valid array coordinates continue to count, and step 2 is repeated.

Described each node obtains the logic time of this node according to the comparison of internal time source and external clock source synchronization result, comprises the following steps:

The first step is to create two arrays in node i: i.local[p], i.global[p], and eight variables: i.current, i.next_sync, i.mlocal, i.mglobal, i.max_gps, ρ, i.logical, τ, where i.current represents the current valid array coordinates of node i; i.local[] represents the best estimate of node i’s time coordinates for the node; i.global[] represents node i The best estimate of the time of other nodes in the network to which the node belongs; i.mlocal represents the maximum value in i.local[]; i.mglobal represents the maximum value in i.global[]; i.max_gps represents the latest GPS timestamp; ρ Indicates the clock drift, the unit is ppm (per million); τ indicates the node synchronization period; p indicates the length of the array; i.logical indicates the logical clock of node i; i.next_sync indicates the scheduled synchronization times of node i;

In the second step, if node i finds that i.logical=i.next_sync*τ, node i starts to synchronize, and checks whether the external clock source exists. If it exists, it sends an external synchronization service request and enters step 3; All existences start internal clock synchronization, that is, execute step 4, and at the same time increase the variable i.next_sync by 1, but the 0th layer nodes do not perform internal clock synchronization;

In the third step, node i performs an external clock source synchronization process: if node i receives a GPS synchronization signal, i.current is incremented by one, and the GPS timestamp is stored in i.local[i.current] and i.max_gps, i. local[i.current] is updated synchronously with the physical clock;

In the fourth step, node i performs the internal clock synchronization process: node i sends a synchronization service request to the directly subordinate node j, if node i receives the synchronization frame sync_message(j.local[j.current] from the superior node j, j.max_gps), node i checks whether node j’s GPS timestamp is as new as node i’s i.max_gps, then checks whether j.local[j.current] is greater than i.global[i.current], if both All are satisfied, which means that node i has an error in estimating i.global[i.current] of other nodes, then update i.global[i.current] to j.local[j.current], j.global[j.current] Update with (1-ρ)/(1+ρ) times of the physical clock;

In the fifth step, the node i compares the elements of the array i.local[] and takes the largest one as i.mlocal, compares the elements of the array i.global[] and takes the largest one as i.mglobal;

In the sixth step, node i compares the size of i.mlocal and i.mglobal, and takes the larger one as i.logical, that is, the logical clock has completed node clock synchronization, and i.logical is updated synchronously with physical time.

In this embodiment, the synchronization error between nodes is as follows: |i.logical(t)-j.logical(t)|≤2(D+ρ(T+D)), the synchronization error between nodes is approximately The signal delay between nodes is controlled within 2 times, and the network structure is recursively descending in a binary tree structure. The purpose of this embodiment is to use this method to meet the synchronization accuracy requirements of the network with as little network overhead and node computing overhead as possible.

Write the method of this embodiment into each development board in the form of software, and then use the development board as a wireless sensor network node to build a network, observe the network communication, mainly to confirm that the network has been established, and the top-level nodes are connected to the external clock source. In the embodiment, the drift ρ of the crystal oscillator is 10 ⁻⁴ seconds, and the synchronization period τ is 1 second. The number of hierarchical layers of the network depends on the communication capacity and signal coverage of the nodes.

On the basis of the actual networking experiment, the relationship between the adjustment parameter ρ and the adjustment synchronization period τ was simulated. The simulation results are shown in Figure 2. Figure (a) is the simulated unprocessed crystal oscillator frequency change diagram , (b) is the change diagram of logic frequency after parameter adjustment, and the ordinates in these two figures respectively represent the difference between the original frequency and the reference frequency, and the logic frequency and the reference frequency. Figure (c) is a histogram showing the comparison of the synchronization periods before and after processing. Among them, No. 1 column represents the synchronization cycle before processing, and No. 2 column represents the synchronization cycle after processing. It can be seen from the figure that the processed logic frequency is much more stable than the original frequency, so the synchronization period can be extended as much as possible under the condition of satisfying a certain timing accuracy, which has great advantages for energy-sensitive wireless sensor networks. Significance, especially when the transmission power consumption far exceeds the operation power consumption.

Claims (2)

1, synchronization method for layered time comparative clock in a kind of wireless sensor network is characterized in that, comprises the steps:

Step 1, network begins initialization: with total base station/telegon is top layer, send the membership acknowledgement frame according to signal strength signal intensity and device category, after corresponding equipment receives acknowledgement frame, and echo reply frame, by that analogy, finish, confirm that simultaneously network reaches stable state whole Network Layering setting;

Step 2, from top layer, successively begin synchronously, be divided into two parts synchronously: external clock source is synchronous and the internal clock source is synchronous, in the external clock reference synchronizing process, if node i is received the GPS synchronizing signal, current effective array coordinate i.current of node i adds one, external clock reference provides time reference synchronously for internal clock source synchronously, and each node is according to the logical time that relatively obtains this node of interior time source and external clock source synchronized result;

Step 3, each node is unified sets synchronizing cycle, and each node is adjusted clock drift parameter ρ, synchronizing cycle τ according to initial several times synchronized result and synchronization accuracy subsequently, satisfies synchronization accuracy if node is subsynchronous, then increase τ synchronizing cycle, reduce overhead; If do not satisfy, then adjust clock drift parameter ρ, and will reduce τ synchronizing cycle;

Whether step 4, node i check current effective array coordinate greater than array length, if set up, the timestamp of inefficacy is abandoned in current effective array coordinate zero clearing, if be false, current effective array coordinate continues counting, repeating step two.

2, synchronization method for layered time comparative clock in the wireless sensor network according to claim 1 is characterized in that, described each node comprises the steps: according to the logical time that relatively obtains this node of interior time source and external clock source synchronized result

The first step equals the reservation synchronization times of node i and the product of node synchronizing cycle if node i is found the logical timer of node i, and node i begins synchronously, and check whether external clock reference exists, if exist, then send the external sync service request, entered for second step; Whether external clock reference exists that all to begin internal clocking synchronous, promptly carries out for the 3rd step, and the reservation synchronization times of node i increases by one simultaneously, but that the 0th node layer does not carry out internal clocking all the time is synchronous;

Second step, it is synchronous that node i is carried out external clock reference: if node i is received the GPS synchronizing signal, current effective array coordinate i.current of node i adds one, gps time stabbed deposits array i.local[i.current in] and i.max_gps in, i.local[i.current] upgrade synchronously with phy clock, i.local[] represent the best estimate of node i to this node time coordinate of living in, i.max_gps represents that nearest gps time stabs;

The 3rd step, it is synchronous that node i is carried out internal clocking: node i to directly under superior node j send synchronization service request, if node i receive the synchronization frame sync_message that sends from superior node j (j.local[j.current], j.max_gps), node i checks whether the same new with the i.max_gps of node i the gps time of node j stabs, check j.local[j.current subsequently] whether than i.global[i.current] big, i.global[] the expression node i is to other node time best estimates of this node belonging network, if both are all satisfied, representation node i is to the estimation i.global[i.current of other nodes] error arranged, then upgrade i.global[i.current] for j.local[j.current], j.global[j.current] doubly upgrade with (1-ρ)/(the 1+ ρ) of phy clock;

In the 4th step, node i is array i.local[relatively] each element, getting the maximum is i.mlocal, relatively array i.global[] each element, getting the maximum is i.mglobal;

In the 5th step, node i compares the size of i.mlocal and i.mglobal, and getting the greater is the logical timer that i.logical represents node i, and it is synchronous to finish nodal clock, and i.logical upgrades synchronously with physical time.

Images