CN103024799B - The method of wireless sense network delay analysis on a large scale - Google Patents

Abstract

The invention relates to the fields of wireless ad hoc networks and sensor networks, in particular to a delay analysis method for large-scale wireless sensor networks. The method includes steps: S1. Calculate the difference between the time when the network node receives the packet and the time when the packet is transmitted to obtain the first delay; S2. Remove errors caused by incorrect timestamps in the first delay to obtain the second delay; S3. Remove the second delay In the second delay, an error caused by time stamp overflow is obtained to obtain a third delay; S4. Combined with the third delay, an incorrect delay value is restored to obtain a final delay. The delay analysis method of the present invention is simple and easy, does not depend on network synchronization, can analyze the delay of a wide range of wireless sensor networks, and does not cause additional overhead, and can provide data support for the improvement of system service quality, and has a strong practicality.

Description

Latency analysis method for wide range wireless sensor network

technical field

The invention relates to the fields of wireless ad hoc networks and sensor networks, in particular to a delay analysis method for large-scale wireless sensor networks.

Background technique

Wireless sensor networks (WSNs, Wireless SensorNetworks) are widely used in building structures, health monitoring and many other aspects; and these applications often require service quality assurance to meet system requirements, such as: to meet the needs of real-time data transmission, etc.; in the impact Among the main factors of system service quality, delay is a very important factor.

In the prior art, for wireless sensor networks, there are a lot of research work on delay analysis and modeling; for example, in the random delay model, a discrete Markov process is used to combine real-time theory and queuing theory; another example, Some empirical network delay models are used as end-to-end delay measurements; another example is the single-hop channel access delay model, which is different from end-to-end delay models; these models and analyzes are based on ideal network conditions (such as high communication and fixed transmission channel) as the basis; however, these ideal conditions are often not realized in the actual wireless sensor network; moreover, these studies lack the comparison and proof of the actual large-scale network.

In the prior art, there is also a lot of research work on delay analysis and measurement of the Internet and data centers; for example, an accurate delay measurement method that prompts packet loss, which uses a lossy delta aggregator that can measure the delay of each packet in the network The very limited additional communication overhead caused by the delay; for another example, in order to measure the delay of each stream, the reference delay interpolation measurement method and so on.

Although there have been many research works on wireless sensor networks, the Internet and data centers, the measurement and analysis of end-to-end delay performance in practical large-scale operable wireless sensor networks is still missing; and, considering wireless In order to understand the requirements of various applications in sensing networks, it is important to understand the delay performance in practical large-scale networks.

However, there are great challenges in the measurement and analysis of latency performance in large-scale operational wireless sensor networks. First of all, unlike the Internet and data centers, there is a software part in the wireless sensor network that supports the delay measurement of each packet, and the delay measurement relies on network synchronization; while traditional network synchronization will cause additional overhead, and considering packet loss, Therefore, it is unreliable; at the same time, due to the limited node resources and the relatively large range of the network, it may not be able to undertake tasks such as network synchronization. Secondly, it is challenging to perform data analysis on the collected information, and a single delay change may be accompanied by many performance changes; in addition, collecting the required information from the network will also cause high network overhead; and, due to resource Due to limitations and packet loss, the information is usually incomplete; moreover, according to the protocol design, the delay itself is random, so it is difficult to automatically and effectively extract useful information from the collected data.

In summary, a practical delay analysis method for large-scale wireless sensor networks is urgently needed.

Contents of the invention

(1) Technical problems to be solved

The object of the present invention is to provide a kind of delay analysis method of large-scale wireless sensor network that is simple and easy and does not depend on network synchronization, is used for analyzing the delay of large-scale wireless sensor network, and does not cause additional overhead, for The improvement of system service quality provides data support.

(2) Technical solution

Technical scheme of the present invention is as follows:

A large-scale wireless sensor network delay analysis method, comprising steps:

S1. Calculate the difference between the time when the network node receives the packet and the time when the packet is transmitted to obtain the first delay;

S2. Remove errors caused by incorrect timestamps in the first delay to obtain a second delay;

S3. Remove errors caused by time stamp overflow in the second delay to obtain a third delay;

S4. Combining with the third delay, restore the incorrect delay value to obtain the final delay.

Preferably, said step S2 includes:

Build compensation constraints in conjunction with clock skew;

Delay values that do not satisfy the compensation constraints are removed.

Preferably, said step S3 includes:

Delay values greater than the maximum overflow time of the timestamp in the second delay are removed.

Preferably, the maximum overflow time of the timestamp is 4294967295 milliseconds.

Preferably, said step S4 includes:

calculating an offset and compensation for an incorrect delay value based on said third delay;

In combination with the offset and compensation, incorrect delay values are recovered.

(3) Beneficial effects

The present invention firstly calculates the difference between the time when the network node receives the packet and the time when the packet is transmitted to obtain a rough delay result, then removes the error caused by the incorrect timestamp and the error caused by the overflow of the timestamp, restores the incorrect delay value, and obtains Final delay results. The delay analysis method of the present invention is simple and easy to implement and does not depend on network synchronization, and can analyze the delay of a large-scale wireless sensor network without causing additional overhead; at the same time, the present invention has the characteristics of small error, mainly manifested as no data There is no error caused by packet error, no error caused by data packet overflow, and all error data packets can be recovered; therefore, the present invention can provide data support for the improvement of system service quality, and has strong practicability.

Description of drawings

Fig. 1 is a schematic diagram of the first delay calculation process of a large-scale wireless sensor network delay analysis method of the present invention;

Fig. 2 is a schematic diagram of the application effect of a large-scale wireless sensor network delay analysis method of the present invention.

Detailed ways

The specific implementation manner of the invention will be further described below in conjunction with the accompanying drawings and embodiments. The following examples are only used to illustrate the present invention, but not to limit the scope of the present invention.

The present invention is mainly based on the time stamp technology of the data access control layer (MAC, Media Access Control), and the time stamp of the data access control layer can accurately record the time of packet sending and receiving.

A large-scale wireless sensor network delay analysis method mainly includes steps:

S1. Calculate the difference between the time when the network node receives the packet and the time when the packet is transmitted to obtain the first delay;

In this embodiment, take the node 3 shown in Figure 1 receiving packets from node 1 and node 2 as an example; assuming that this event occurs at the local time t ₁ of node 1, we first measure the event time and node 2 and node 2 3 local time: Suppose node 1 sends a packet at time t ₂ , the event time contained in the packet is t ₁ , and then node 2 receives the packet with a time stamp of t ₃ , assuming that the transmission time is trivial, receiving node 2 can base it on The local clock of t ₃ calculates the event time, and subtracts (t ₂ -t ₁ ) from t 3 , that is, t ₃ -(t ₂ -t ₁ ). From an intuitive point of view, the event is at t ₃ -(t ₂ -t ₁ ) time; similarly, node 3 calculates the event time in the same way after receiving the packet from node 2. Therefore, assuming that a packet is sent at time t ₁ , the sink node (i.e. node 3 in Figure 1) can calculate the sending time of the packet t ₅ -(t ₄ -(t ₃ -(t ₂ -t ₁ )) ) and the reception time t ₅ , both of which are referenced to the clock of node 3; then the delay t ₄ -(t ₃ -(t ₂ -t ₁ )) of the packet is calculated. The time when the source node transmits the packet is recorded as the source time of the packet (SourceTime), and the time when the sink node receives the packet is recorded as the sink time (SinkTime). According to the clock of the sink node, the packet transmission time (the time of the source node relative to the sink node ) is recorded as SourceTimeAtSink, correspondingly, the delay delay can be calculated as follows:

delay=SinkTime-SourceTimeAtSink. (1)

The first delay calculated according to formula 1 is shown in the first column in FIG. 2 , and it can be seen that the delay is distributed in a large range.

The clocks of the source node and the sink node are synchronized, and the linear model used by the sensor node can be:

SourceTimeAtSink＝α ₁ SourceTime+offset ₁ (2)

where α ₁ and offset ₁ are the associated offset and offset;

According to formula 2, the difference between SourceTimeAtSink and SourceTime is:

Offset ₂ = SourceTimeAtSink-SourceTime

=α ₂ SourceTime+offset ₂ (3)

where _α2 and _offset2 are related offsets and offsets.

S2. The first type of error in the first delay is caused by the packet overflow and packet loss of the limited receive buffer, and the error of the underlying driver of the sensor node, which cannot ensure that the correct timestamp is provided; therefore, the error in the first delay is removed. Errors caused by time stamps to obtain the second delay; this step mainly includes: constructing compensation constraints combined with clock offset; removing delay values that do not satisfy the compensation constraints. In this embodiment, it is specifically: calculate the compensation Offset ₂ =SourceTimeA tSink-SourceTime, that is, it can be calculated according to SourceTimeAtSink and SourceTime contained in the received data packet, such as formula 3; assuming that when t ₁ <t ₂ , Offset ₂ ( t ₁ ) and Offset ₂ (t ₂ ) are two compensations calculated at time t ₁ and t ₂ ; when the clock offset is the maximum value e, according to formula 3, the compensation should satisfy |Offset ₂ (t ₂ ) -Offset ₂ (t ₁ )|≤(t ₂ -t ₁ )e; Based on this compensation constraint, judge whether the time stamp calculated by each data packet and the time stamp calculated by all previous data packets meet the compensation constraint; All delay values satisfying the compensation constraints are left and put into the same group. At the same time, since the incorrect delay is randomly distributed, the error data does not satisfy the compensation constraints that most data satisfy, so only the group that satisfies the most compensation constraints is retained, and other groups that are less than this group are removed. The resulting second delay is shown in the second column of Figure 2.

S3. The second type of error comes from the overflow of the timestamp, so it is necessary to remove the error caused by the overflow of the timestamp in the second delay to obtain the third delay; this step is mainly to remove the maximum overflow greater than the timestamp in the second delay The delay value for time. For example, the SourceTimeAtSink provided by timestamp technology is a four-bit timestamp based on a 32KHz timer, so the maximum overflow time is 0Xffffffff/32ms (4294967295 milliseconds), which means about 1.5 days; through formula 1, it can be found that the timestamp overflow causes The particularly large delay value and remove it. The removal method is to subtract 0Xffffffff/32ms from the calculated delay until the delay is smaller than the maximum overflow time; the above processing is based on the fact that most normal delays are less than the maximum overflow time ( 4294967295 milliseconds); the resulting third delay is shown in the third column in Figure 2.

S4. Combining the third delay, restore the incorrect delay value to obtain the final delay; this step mainly includes: calculating the offset and compensation of the incorrect delay value according to the third delay; combining the offset and compensation, Restore incorrect delay value. Specifically in this embodiment: first, calculate α ₁ and offset ₁ according to formula 2 and the correct time stamp and the calculated delay, thereby obtaining the linear model of SourceTimeAtSink; then, for the incorrect delay value at t2 , we can first calculate SourceTimeAtSink according to Formula 2, and then calculate the delay according to Formula 1, and the final delay obtained is shown in the fourth column in Figure 2. The basic principle of this step is to use the correct delay to synchronize the source node and the sink node, and then calculate the delay of the two synchronization nodes; the error caused by this method is at most error=e×(t ₂ -t ₁ ), e is the corresponding t ₁ is the latest correct delayed SourceTime; if error≤δ is required, only the source time t ₂ of the timestamp to be recovered satisfies t ₂ ≤t ₁ +δ/e.

The above embodiments are only used to illustrate the present invention, but not to limit the present invention. Those of ordinary skill in the relevant technical field can make various changes and modifications without departing from the spirit and scope of the present invention. Therefore, all Equivalent technical solutions also belong to the protection category of the present invention.

Claims (4)

1. the method for wireless sense network delay analysis on a large scale, is characterized in that, comprise step:

S1. the difference of computing network node receiving package time and packet transmission time, obtains the first delay;

S2. remove first postpone in the mistake that causes because of incorrect timestamp, obtain the second delay;

S3. remove second postpone in overflow the mistake caused because of timestamp, obtain the 3rd delay;

S4. in conjunction with described 3rd delay, recover incorrect length of delay, obtain final delay;

Described step S2 comprises:

Build in conjunction with clock skew and compensate constraint;

Remove and do not meet the described length of delay compensating constraint;

Wherein, described compensation is constrained to:

|Offset ₂(t ₂)-Offset ₂(t ₁)|≤(t ₂-t ₁)e，

Offset ₂(t ₁) be at time t ₁the compensation calculated, Offset ₂(t ₂) be at time t ₂the compensation calculated, e is the maximum of clock skew, t ₁<t ₂.

2. method according to claim 1, is characterized in that, described step S3 comprises:

Remove the length of delay being greater than the timestamp maximum spilling time in described second delay.

3. method according to claim 2, is characterized in that, the described timestamp maximum spilling time is 4294967295 milliseconds.

4. method according to claim 1, is characterized in that, described step S4 comprises:

According to skew and the compensation of the incorrect length of delay of described 3rd Delay computing;

In conjunction with described skew and compensation, recover incorrect length of delay.