CN101692743A - Method for realizing time synchronization in wireless sensor network - Google Patents

Abstract

The invention discloses a method for realizing the time synchronization in a wireless sensor network, which comprises the following steps: initializing a counter; starting the counter, opening an interrupt and starting a message queue; and calculating the increment of the counter according to a received data packet and modifying the value of the counter according to the increment. The method for realizing the time synchronization in the wireless sensor network achieves the aim of time synchronization in the wireless sensor network by periodically modifying a counting value of a linear oscillator and solves the problem of loss of synchronization in network node time due to the timing rate error of nodes in the network caused by that a crystal oscillator is susceptible to the influences of external factors during operation.

Description

Realize the method for wireless sensor network time synchronized

Technical field

The present invention relates generally to areas of information technology, relates in particular to a kind of method that realizes the wireless sensor network time synchronized.

Background technology

The application of wireless sensor network more and more widely, at characteristics such as the uncertainty of challenge-transmission delay that the wireless sensor network time synchronized faced, low-power consumption, low cost, extensibilities, some typical time synchronization protocols have appearred in the wireless sensor network field.Typical wireless sensor network event synchronization agreement can be divided into synchronous and synchronous two classes of recipient-recipient of sender-recipient.TPSN (TimeProtocol for Sensor Networks) agreement is typical sender-recipient's synchronous protocol, and it suppresses by MAC layer timestamp technology and bi-directional switching method and reduces transmission delay and postpone uncertain influence.RBS (Reference Broadcast Synchronization) agreement is typical recipient-recipient's synchronous protocol, except directly the time migration between the recipient being estimated, it further reduces synchronous error by dynamic estimation is carried out in recipient's clock drift (clock skew).In the time synchronization protocol of multihop network, " time route " method that proposes in the RBS agreement, representative work also has LTS (Lightweight Tree-based Synchronization) agreement, HRTS (HierarchyReferencing Time Synchronization) agreement, FTSP (Flooding Time SynchronizationProtocol) agreement and GCS (Global Clock Synchronization) agreement.These typical time synchronized algorithms stress the demand of synchronization accuracy and synchronous energy consumption.

The distributed system that described wireless sensor network is made up of sensor node.Separate and between the node with wireless mode communication, local clock of each node maintenance.The timing signal of local clock is generally provided by crystal oscillator (abbreviation crystal oscillator).

In implementing process of the present invention, the inventor finds the restriction owing to the crystal oscillator manufacturing process, and crystal oscillator is subject to the extraneous factor influence in running, and then causes the deviation of the clocking rate of node in the network, has caused the step-out of network node time.Therefore, time synchronization problem is the major issue that wireless sensor network will solve.

At present, in the location, nearly all occasion such as range finding, data fusion, mac-layer protocol, sleep scheduling, Routing Protocol, cooperation transmission, database synchronization all has clear and definite requirement to time synchronized.Different with the time synchronized of traditional distributed system, the time synchronized of wireless sensor network not only has high-precision requirement, but also is faced with new challenges such as energy consumption, extensibility, wireless transmission unreliability.

Summary of the invention

The object of the present invention is to provide a kind of method that realizes the wireless sensor network time synchronized, prior art is because the influence of crystal oscillator causes the problem of network node time step-out.

A kind of method that realizes the wireless sensor network time synchronized of the present invention may further comprise the steps: the count initialized device; The enabling counting device is opened interruption, the initiation message formation; Revise the count value of linear oscillator according to the packet of receiving.

The method of realization wireless sensor network time synchronized of the present invention reaches the purpose of wireless sensor network time synchronized by the count value of revising linear oscillator, solved in the prior art, because crystal oscillator is subject to the extraneous factor influence in running, and then cause the deviation of the clocking rate of node in the network, caused the problem of the step-out of network node time.

Description of drawings

Fig. 1 is the flow chart of the method for the described realization wireless sensor network of specific embodiment of the invention time synchronized;

Fig. 2 is for receiving the process chart of packet in the method for the described realization wireless sensor network of specific embodiment of the invention time synchronized;

The process chart that Fig. 3 overflows for counter in the method for the described realization wireless sensor network of specific embodiment of the invention time synchronized;

Fig. 4 is T described in the method for the described realization wireless sensor network of specific embodiment of the invention time synchronized ₀The standard deviation of synchronized result is with the situation of change of d when getting 2T, 4T, 8T, 16T;

Fig. 5-7 got respectively 1/8,1/4,1/2 o'clock for ε described in the method for the described realization wireless sensor network of specific embodiment of the invention time synchronized, and the standard deviation of synchronized result is with the situation of change of d.

Embodiment

For making the present invention clearer, below in conjunction with specific implementation process and the result of the present invention in wireless sensor network, and with reference to accompanying drawing, the present invention is described in more detail.

Following implementation process is to realize on the wireless sensor network test envelope based on the GAINS-3 node.Described GAINS-3 node mainly is made up of an ATMega128 single-chip microcomputer and a CC1000 wireless transceiver, also has a serial ports simultaneously, is responsible for the reliable transmitting-receiving of programming and data.The serial ports of all nodes all is connected in the Ethernet by the modular converter of serial ports to Ethernet, and a Web server in this Ethernet is responsible for the programming to node, to the data that the node transmission is ordered and receiving node returns.In addition, sample for the clock (counter) to all nodes simultaneously, the INT4 pin of ATMega128 is joined together on all nodes, and the trailing edge on this pin can trigger all nodes and send sampled result to Web server.

When implementing, with oscillator phase φ from [0,1] interval is amplified to [0, T], the phase place that is about to oscillator uses 32 counter to represent, this counter dullness be increased to T (being the cycle), send a packet then, described packet comprises serial number, node number and period T, and the node that receives this packet uses the formula mistake according to the T in the packet! Do not find Reference source.Calculate the coupling increment, the phase place that then this increment is added to oneself gets on.Below in conjunction with specific embodiment, and with reference to accompanying drawing, the present invention is described in more detail.

As mistake! Do not find Reference source.Shown in, the method for realization wireless sensor network time synchronized of the present invention comprises following steps:

Step 101, the count initialized device comprises:

Mode of operation, the maximum count value (being T) of counter are set;

Node number and period T and stiffness of coupling are set.

In implementation process with the period T of node i _iBe made as:

T _i＝T ₀+(i-1)*d

Wherein, T ₀Radix for the cycle; D describes the cycle differentiation of adjacent numbered node.

Step 102, enabling counting device begin counting, open interruption, allow counter to overflow, and the interruptions such as reception of packet, last initiation message formation.

Node is after receiving packet, and according to the data computation phase increment that excites node that is comprised in the packet, the phase place that then this increment is added to oneself gets on.Next judge whether oneself excites, and carry out corresponding processing procedure.As shown in Figure 2, concrete steps are as follows:

Step 201: from the packet that receives, obtain the period T excite the node number of node and to excite node.

Step 202: whether judge phase place greater than a certain threshold value L, if just calculate phase increment greater than this threshold value.The reason that threshold value L is set is, when certain node A sends packet and Node B is excited, node A can receive the excitation data bag that B sends, and can not cause the phase place of A to increase because of exciting of B if this threshold value is not set.Therefore, the generation that threshold value can prevent this situation is set.

A step 203: by the cycle that excites node that step 201 obtained, according to the formula mistake! Do not find Reference source.Calculate phase increment.Stop counter then, phase increment is increased to counter gets on, at last enabling counting device once more.

Step 204: judge whether the counter after increasing through step 203 overflows, if overflow then illustrate that this node excites, invocation step 3 is correlated with overflows processing.Otherwise processing procedure finishes.

When counter overflows, need counter reset, tissue comprises the packet of serial number, node number and period T, then this packet is sent.Concrete steps are as follows as shown in Figure 3:

Step 301: stop counter, the count value of counter is made as 0, restart counter then.

Step 302: serial number from adding 1, then with itself and node number, period T and be saved in the packet related data structure, is broadcasted away this packet then.

Because wireless senser adopts the processor of low-cost low-power consumption, therefore can't carry out complicated function calculation.The linear oscillator of Shi Yonging when guaranteeing precision, has been avoided a large amount of NONLINEAR CALCULATION in the present invention, has saved synchronous energy consumption.

Simultaneously, this method has been avoided identical this hypothesis of oscillator crystal oscillator frequency, and particularly for large-scale wireless sensor network, this point seems particularly important.

In addition, this method only needs a counter in synchronous process, do not need to preserve other state information, also need be based on any topology information, therefore compare with the method for synchronous of classics and saved computing time and memory space, also have autgmentability preferably simultaneously.

For the validity of this method is described, and further discuss how to choose parameters in actual applications, we have carried out actual test to this method on the wireless sensor network test envelope of being made up of 15 nodes.Frequency (the T of node oscillator _Counter) be 115200Hz.No. 0 node can send an interrupt signal when the cycle is carried out roughly half makes all nodes send the counter values of oneself simultaneously to Web server.The standard deviation of the counter of all nodes of calculating except No. 0 node is weighed the precision of these group data, and after having collected all data, the arithmetic mean value of calculating all standard deviations again is as final result.In addition, when calculating last arithmetic mean, some groups of maximins are taken out from the result, to avoid stochastic factor.

At first, discuss how to influence synchronous precision experimental period.ε is fixed as 1/4, for each group T ₀, make d with step-length 500 traversals interval [500,40000], when calculating final result, cast out 5 groups of minimum and maximum data.Fig. 4 is T ₀Get T respectively _Counter2,4,8,16 times, the standard deviation of synchronized result is with the situation of change of d.As can be seen from the figure, for T ₀Each value, change procedure is divided into three phases.When d was very little, standard deviation was with T ₀Very little with the correlation of d, along with the increase synchronization criterion difference of d at first is stabilized in a certain fixed value, increase along with the increase of d then.This be because, when d was very little, the cycle of each node was very approaching, because the effect of threshold value L can make counter enter the state of a kind of lock-step (phase-locked); Along with the increase of d, enter normal synchronous phase between node; And it is excessive to work as d, and makes synchronous conservation condition mistake! Do not find Reference source.When can't be satisfied, the synchronization criterion difference then can significantly increase.In this experiment, have

T _min＝T ₀，T _max＝T ₀+14d?andε＝1/4

A substitution mistake! Do not find Reference source.In, can get

$\frac{T_0}{{\mathrm{<figure-callout\; id="0"\; label="T"\; filenames="H2009101769142E0000032.png,H2009101769142E0000042.png"\; state="\{\{state\}\}">T</figure-callout>}}_0+14d}(1+\frac{1}{4})\&GreaterEqual;1$

Solve

$d\&le;\frac{{\mathrm{<figure-callout\; id="0"\; label="T"\; filenames="H2009101769142E0000032.png,H2009101769142E0000042.png"\; state="\{\{state\}\}">T</figure-callout>}}_0}{56}$

Work as T ₀Equal 2T respectively _Counter, 4T _Counter, 8T _CounterAnd 16T _CounterThe time, the d that is calculated by following formula is respectively: 4114,8228,16457 and 32914.It is very good that these results and Fig. 4 coincide.

In order further to study T ₀4 kinds of values at the synchronization accuracy in above-mentioned synchronism stability stage.We have carried out further experiment.The relevant parameter of experiment is identical with a last experiment, and just d is with step-length 10 traversals interval [3000,3500], and more stable in order to make the result, we have removed 15 minimum and maximum results.Experimental result as shown in Figure 5, as can be seen from the figure less T ₀Can obtain higher precision, this is that it is frequent more that node is adjusted own phase place because the cycle is short more, thus can be fast more reach synchronous.

Next, the influence of stiffness of coupling ε to synchronization accuracy is discussed.In this experiment, with T ₀Be fixed on 2T _Counter, for different ε, make d with step-length 100 traversals interval [100,10000], when calculating final result, cast out 15 groups of minimum and maximum data.Fig. 6 got respectively 1/16,1/8,1/4,1/2 o'clock for ε, and the standard deviation of synchronized result is with the situation of change of d.Similar with first experiment, the change procedure of standard deviation is divided into three phases.But two, the calculating of triphasic intersection point difference slightly, have for above-mentioned experiment condition

T _min＝2T _counter＝230400，T _max＝T ₀+14d＝230400+14d

A substitution mistake! Do not find Reference source.In, can get

$\frac{230400}{230400+14d}(1+\mathrm{\&epsiv;})\&GreaterEqual;1$

Solve

$d\&le;\frac{115200}{7}\mathrm{\&epsiv;}$

When ε was respectively 1/16,1/8,1/4,1/2, the d that is calculated by following formula was respectively 1028,2057, and 4114,8228.This and experimental result are also coincide.

As shown in Figure 7, under same experiment condition, with the experimental result of step-length 10 traversals [1000,2000].For the displaying that makes the result clearer, ε=1/2, ε=1/4 and ε=1/8 is drawn in a) respectively, b) and c) in these three subgraphs.In this three width of cloth figure relatively, can find that the result is more stable when ε=1/4.Reason may be, when ε is very little, and a little less than the coupling too between two nodes, therefore can be destroyed easily; Therefore and when ε was too big, it is synchronous with other node that node is more prone to, and the coupling between two nodes can be influenced by other node again easily.Therefore, a stable synchronized result be obtained, a suitable ε value must be selected.

T has been discussed in above-mentioned experiment ₀If ε and d influence the precision of synchronized result.Analysis by to experiment can draw as drawing a conclusion:

Not too little at d, and under the situation that synchronously conservation condition also can be satisfied, by can well be synchronous, best precision 100 delicate about.

T ₀Determine the synchronous cycle, less T ₀Can obtain the higher synchronous precision.The size of selection ε that must be careful, less and bigger ε can reduce synchronization accuracy.In sum, verified that by experiment this method can carry out obtaining synchronized result preferably synchronously to wireless sensor network node.

Claims (7)

1. a method that realizes the wireless sensor network time synchronized is characterized in that, may further comprise the steps:

The count initialized device;

The enabling counting device is opened interruption, the initiation message formation;

Revise the count value of linear oscillator according to the packet of receiving.

2. method according to claim 1 is characterized in that, described count initialized device comprises:

Mode of operation, the maximum count value of counter are set;

Node number and period T and stiffness of coupling are set.

3. method according to claim 1 and 2 is characterized in that, the period T of node i is set _iFor:

T _i＝T ₀+(i-1)*d

Wherein, T ₀Radix for the cycle; D describes the cycle differentiation of adjacent numbered node.

4. according to the method described in the claim 1, it is characterized in that, also comprise:

Counter linear counting on the node sends one when count value arrives T and has comprised the packet that changes the node cycle, then counter is put 0, repeats this process;

Receiving node is received after the described packet that comprises the cycle, is used following formula to obtain the new count value of oneself

${\mathrm{\&phi;}}_j^{\&prime;}={\mathrm{\&phi;}}_j+\frac{T_i}{T_j}\mathrm{\&epsiv;}$

Wherein: φ ' _jBe new count value; φ _jBe the count value of receiving node counter before receiving packet; T _iBe the cycle of sending node; T _jBe the cycle of receiving node; ε is a stiffness of coupling.

5. method according to claim 4 is characterized in that: if described φ ' _jResult of calculation exceed the periodic quantity of described receiving node, described receiving node is got back to initial value and is sent a packet that comprises the described receiving node cycle.

6. according to the method described in the right 1, it is characterized in that: the probability density function bounded in the cycle of described linear oscillator.

7. method according to claim 6 is characterized in that: the cycle of described linear oscillator satisfies

$\frac{T_{\min }}{T_{\max }}(1+\mathrm{\&epsiv;})\&GreaterEqual;1$

Wherein: T _MinMinimum value for the described linear oscillator cycle; T _MaxMaximum for the described linear oscillator cycle; ε is a stiffness of coupling.

Images