Wireless sensor network clustering rapid time synchronization method for forest fire monitoring

Abstract

The invention discloses a wireless sensor network clustering rapid time synchronization method for forest fire monitoring, and belongs to the technical field of remote sensing. According to the invention, the problems of communication delay and packet loss are considered, clock drift is compensated, and a clustering fast time synchronization algorithm is provided. According to the algorithm, a network is divided into two layers through topology generation, namely bidirectional exchange synchronization of a main network and interception of a secondary network. The method can realize convergence in finite time, and has expandability. And when the secondary network node has a fault, the synchronization of other nodes is not influenced, and the robustness of the algorithm is enhanced. According to the method, the communication overhead is effectively reduced, the convergence speed is increased, and the influence of random delay is reduced.

Description

Technical Field

[0001] The present invention belongs to the field of remote sensing technology and proposes a wireless sensor network clustering fast time synchronization algorithm for forest fire monitoring. The algorithm can achieve faster convergence speed, lower communication overhead and better robustness, while effectively reducing the impact of random delays in the communication process. Background Art

[0002] As the carrier of global ecosystem stability and climate regulation, forests are an indispensable part of human life and the natural environment. However, global forest resources are being severely reduced, of which about one-third is related to forest fires. There are hundreds of thousands of forest fires around the world every year. They are highly destructive, fierce and difficult to extinguish, posing threats and damage to people's lives and property.

[0003] Forest fire monitoring is an important part of forest fire prevention and control. With the development of information technology, the establishment of information-based forest fire monitoring technology has become a hot research topic today. Due to the complex terrain, large area and sparsely populated forest areas, traditional forest fire monitoring methods can effectively prevent the serious harm of fires, but they are still restricted by factors such as cost and region. Wireless sensor network (WSN) technology has the advantages of low power consumption, easy deployment and self-adaptation. It can be deployed on a large scale in the forest to achieve long-term real-time monitoring. Among them, time synchronization is one of the key technologies and necessary conditions to support the stable and reliable operation of WSN. WSN is composed of an indefinite number of nodes, each of which works independently and uses a local clock to keep time. In actual processes, due to the influence of factors such as environment and voltage changes, the clocks of each node will differ over time. Therefore, WSN needs a time synchronization algorithm to correct the local clock so that the nodes working independently have consistent time.

[0004] Considering the hardware resources, node energy and other factors in practical applications, communication delay cannot be ignored. Sometimes packet loss may even occur, which has a negative impact on the convergence speed and synchronization accuracy of the algorithm. Therefore, WSN needs a time synchronization algorithm with low power consumption, fast convergence speed and good robustness. Most of the early WSN time synchronization algorithms calculate the time deviation and correct the local time of the node by exchanging data packets between nodes. However, this algorithm depends on the topological structure and reference nodes. When new nodes join the network or some nodes fail, these algorithms need to rebuild the topological structure of the entire network, resulting in additional energy consumption. In order to enhance robustness, researchers have proposed a distributed synchronization algorithm that adjusts its own clock according to neighbor information to achieve global relative synchronization. Although this type of algorithm does not require topology management, has no reference nodes, and does not rely on routing, it also has the disadvantages of slow convergence speed and low efficiency. Based on this, the present invention designs a clustered fast time synchronization algorithm for wireless sensor networks for forest fire monitoring. Summary of the invention

[0005] Aiming at the scenario of WSN applied to forest fire monitoring, the present invention proposes a cluster-based fast time synchronization algorithm (CFTS), which can effectively reduce network communication overhead and reduce the impact of random communication delay.

[0006] The technical solution of the present invention is a wireless sensor network clustering fast time synchronization method for forest fire monitoring, the method comprising:

[0007] Step 1: Topology generation;

[0008] First, the cluster head node is elected from the gateway node outward, and the timer is set using the node's own remaining energy and distance. The timer is used to determine which node wins the election. If a node fails to win the election, it will mark itself as an ordinary node and select the nearest cluster head node as the parent node based on its own recorded cluster head information, and join the cluster headed by the node. Only when the number of time synchronization cycles reaches the preset number or a node fails, a new round of topology generation will be performed.

[0009] Step 2: Main network time synchronization;

[0010] The cluster head nodes form the main network and achieve time synchronization through a two-way exchange mechanism; let node A be the time reference node, node B be the node to be synchronized, and the timestamp τ A (t2) and τ A (t3) is the local time of node A, timestamp τ B (t1) and τ B (t4) is the local time of Node B;

[0011] At time t1, node B initiates a synchronization request to node A. Node A receives the request at time t2 and replies with a message packet at time t3. Node B receives the message packet at time t4 and records the timestamp τ. B (t4); Taking the local time of node A as the reference time, the timestamp of node B is represented by node A; node B estimates the relative clock drift by using the synchronization reply message packets received twice adjacently, and further estimates the drift compensation parameter;

[0012] Using the drift compensation parameters and deviation compensation parameters replied by node A, the clock of node B is adjusted to be consistent with the logical clock of node A. In the case of multi-hop in the main network, the first-level cluster head node is within the single-hop range of the sink node, and synchronizes with the sink node using this method. In turn, all nodes in the main network are directly or indirectly synchronized with the sink node, so that the logical clock of the entire main network is consistent with the sink node.

[0013] Step 3: Secondary network time synchronization;

[0014] The synchronization of the secondary network adopts a one-way listening mechanism. When a cluster head node in the main network initiates a message containing a synchronization request to the cluster head node of the upper level, it will broadcast the request. At this time, the cluster nodes in the cluster where the cluster head node is located will listen to and record the time information broadcast by the cluster head node; the relative time drift is estimated based on the two consecutive time information received, and the drift compensation parameters are updated in combination with the low-pass filter; the low-pass filter coefficient is further updated to reduce the impact of packet loss on the convergence speed of the algorithm; after the drift compensation converges, the logical clock drift of the nodes in the cluster is consistent with the cluster head node, and the logical clock deviation of the nodes in the cluster gradually converges.

[0015] Furthermore, the specific method of using the timer in step 1 to determine which node wins the election is as follows:

[0016] The gateway node broadcasts an information packet, which contains information such as node number and layer number; when other nodes within the single-hop communication range of the gateway node receive the packet, they calculate the distance between themselves and the gateway node based on the RSSI strength, and start the timer and set the timing time to elect the cluster head node based on their own relative remaining energy and distance; after the timer times out, a message packet is broadcast to the surrounding nodes, announcing that they have successfully run for the cluster head node; nodes that have not successfully run for the election clear the timer and become ordinary nodes.

[0017] Furthermore, the specific method of step 2 is:

[0018] Taking the local time of node A as the reference time, node B estimates the relative clock drift using the synchronization reply message packets received twice in succession; node B adjusts its own time to be consistent with the local time of node A through the following calculation.

[0019] τ A (t2) = ρ BA (τ B (t1)+d)+θ BA (1)

[0020] τ A (t3) = ρ BA (τ B (t4)-d)+θ BA (2)

[0021] τ A (t3-1) = ρ BA (τ B (t4-1)-d)+θ BA (3)

[0022]

[0023]

[0024] τ A (t) = τ′ B (t) = ρ BA τ B (t)+θ BA (6)

[0025] Among them, ρ BA represents the relative clock drift of node B relative to node A, d represents the fixed delay during transmission, and θ BA is the relative clock deviation of node B relative to node A, τ A (t3-1) represents the time when node A last replied to the message packet, τ B (t4-1) represents the time when node B last received the reply message packet, τ B (t) represents the local time of node B, τ′ B (t) represents the corrected node B time consistent with the local time of node A;

[0026] Use the drift compensation parameters replied by node A and deviation compensation parameters Adjust the clock of Node B To the logical clock of node A Consistent, that is Complete synchronization;

[0027]

[0028] τ A(t) represents the local time of node A. Taking into account the influence of random delay, the joint maximum likelihood clock drift and deviation estimator corresponding to the Gaussian random delay model is used on this basis; at this time, the expression is as follows:

[0029] τ A (t2) = ρ BA (τ B (t1)+d+X i )+θ BA (8)

[0030] τ A (t3) = ρ BA (τ B (t4)-dY i )+θ BA (9)

[0031]

[0032]

[0033] Among them, X i and Y i is a random delay, assuming and is zero mean and has a variance of σ 2 Gaussian distributed independent and identically distributed random variables, for the convenience of expression, introduce the constant Q, N is the total number of random delays.

[0034] Furthermore, the specific method of step 3 is:

[0035] Step 3.1: The cluster head node j in the main network initiates a message to the cluster head node at the next level. Synchronization request, τj(t) represents the local time of cluster head node j, represents the drift compensation parameter of cluster head node j, represents the deviation compensation parameter of cluster head node j; the node i in the cluster where cluster head node j is located listens to and records the time information broadcast by cluster head node j, and estimates the relative clock drift ρ based on the two consecutive time information received ij , using ρ ij And the low-pass filter updates its drift compensation parameters

[0036] Step 3.2: Modify the low-pass filter coefficients to reduce the impact of packet loss on the algorithm convergence speed;

[0037]

[0038] in, is the drift compensation parameter value of the last update iteration, and its initial value is 1; α is a predefined weight coefficient, satisfying α∈(0,1), τ i (t) and τ i (t-1) is the time when node i receives time information twice in a row, T is the synchronization period, and ρ ij (t) represents the relative clock drift at time t;

[0039] Step 3.3: After the drift compensation converges, the logical clock drift of the nodes in the cluster is consistent with that of the cluster head node, and the logical clock deviation of the nodes in the cluster gradually converges; to reduce the impact of random delay, node i uses the data of node j received in a short period of time to perform linear regression analysis to estimate the relative clock drift ρ ij ; According to the fixed delay d′ during the transmission process, the deviation compensation parameter is obtained

[0040]

[0041]

[0042]

[0043] Among them, β is the weight coefficient, satisfying β∈(0,1), ρ j represents the slope of the clock at node j, θ j represents the clock deviation of node j, represents the kth τ i The value of (t), represents the mean, τ j (t) and τ i (t) are the local hardware clocks of nodes j and i at time t respectively, d′ is a priori value determined based on the measured value of the transmission delay between the two nodes; n is the number of data within the specified time.

[0044] Beneficial effects of the present invention: The present invention takes into account the problems of communication delay and packet loss, compensates for clock drift, and proposes a clustered fast time synchronization algorithm. The algorithm divides the network into two layers through topology generation, namely, the bidirectional exchange synchronization of the primary network and the listening of the secondary network. It can achieve convergence within a limited time and has scalability. When a secondary network node fails, it will not affect the synchronization of other nodes, thereby enhancing the robustness of the algorithm. The present invention effectively reduces communication overhead, accelerates convergence speed, and reduces the impact of random delays. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] Figure 1 This is an overview map of the study area. A forest area was selected as the study area of ​​the present invention. The forest area is located in the eastern Sichuan fold belt with undulating surface.

[0046] Figure 2 CFTS algorithm flow chart. The fast time synchronization algorithm based on clustering structure proposed in the present invention is mainly divided into two stages, namely topology generation stage and time synchronization stage.

[0047] Figure 3 The figure is a schematic diagram of a two-layer network, where green nodes represent cluster head nodes, white nodes represent ordinary nodes, and blue nodes represent gateway nodes. The primary network consists of cluster head nodes, and the secondary network consists of ordinary nodes in the cluster where each cluster head is located.

[0048] Figure 4 Schematic diagram of bidirectional exchange. Nodes in the main network achieve synchronization by estimating the clock deviation and clock offset of the nodes through a bidirectional exchange mechanism.

[0049] Figure 5 This is a comparison diagram of the simulated global network logic clock convergence, where the blue curve represents the ATS algorithm and the yellow curve represents this algorithm. DETAILED DESCRIPTION

[0050] The synchronization process is divided into two stages: topology generation and time synchronization. In the topology generation stage, the entire network is divided into several clusters using the non-uniform clustering method; in the time synchronization stage, the primary network uses a two-way information exchange mechanism to ensure synchronization accuracy, and the secondary network unidirectionally monitors the primary network to exchange information and synchronizes its own local time to the primary network; on this basis, the algorithm uses Gaussian distribution as the delay model in the primary network and introduces joint maximum likelihood estimation parameters; in the secondary network, linear regression estimation parameters are introduced to reduce the impact of random communication delays and even packet loss in practical applications.

[0051] The present invention will be further described below in conjunction with the accompanying drawings.

[0052] (1) Topology Generation

[0053] In order to avoid imbalance in network energy consumption caused by high energy consumption pressure of some nodes in the structural algorithm, the present invention adopts a non-uniform clustering method to generate a topological structure.

[0054] The gateway node broadcasts an information packet, which contains information such as node number and layer number. When other nodes within the single-hop communication range of the gateway node receive the packet, they calculate the distance between themselves and the gateway node based on the RSSI strength, and start the timer and set the timing time to elect the cluster head node based on their own relative remaining energy and distance. After the timer times out, a message packet is broadcast to the surrounding nodes, announcing that they have successfully run for the cluster head node. Nodes that have not successfully run for the election clear their timers and become ordinary nodes. Ordinary nodes select the nearest cluster head node as the parent node based on the cluster head information they record, join the cluster with this node as the cluster head, and establish a cluster network. The final generated topology is as follows: Figure 3 shown.

[0055] (2) Main network time synchronization

[0056] The main network achieves synchronization by estimating the clock deviation and clock offset of the nodes through a two-way exchange mechanism.

[0057] The schematic diagram of bidirectional exchange is as follows Figure 4 As shown in the figure, taking the local time of node A as the reference time, node B estimates the relative clock drift using the synchronization reply message packets received twice in succession. Node B can adjust its own time to be consistent with the local time of node A through the following calculation.

[0058] τ A (t2) = ρ BA (τ B (t1)+d)+θ BA (1)

[0059] τ A (t3) = ρ BA (τ B (t4)-d)+θ BA (2)

[0060] τ A (t3-1) = ρ BA (τ B (t4-1)-d)+θ BA (3)

[0061]

[0062]

[0063] τ A (t) = τ′ B (t) = ρ BA τ B (t)+θ BA (6)

[0064] Among them, ρ BA represents the relative clock drift of node B relative to node A, d represents the fixed delay during transmission, and θ BA is the relative clock deviation of node B relative to node A. τ A (t3-1) represents the time when node A last replied to the message packet, τ B (t4-1) represents the time when node B last received a reply message packet.

[0065] Use the drift compensation parameters replied by node A and deviation compensation parameters Adjust the clock of node B to be consistent with the logical clock of node A, that is, Synchronization is complete.

[0066]

[0067] Taking into account the influence of random delay, the joint maximum likelihood clock drift and deviation estimator corresponding to the Gaussian random delay model is used on this basis. The expression is as follows:

[0068] τ A (t2) = ρ BA (τ B (t1)+d+X i )+θ BA (8)

[0069] τ A (t3) = ρ BA (τ B (t4)-dY i )+θ BA (9)

[0070]

[0071]

[0072] Among them, X i and Y i is a random delay. Assume and is zero mean and has a variance of σ 2 Gaussian distributed independent and identically distributed random variables. d is a fixed delay.

[0073] (3) Secondary network time synchronization

[0074] The secondary network uses a one-way listening method for time synchronization. The cluster head node j in the primary network initiates a message containing Synchronization request, the node i in the cluster where the cluster head node j is located listens to and records the time information broadcast by the cluster head node j, and estimates the relative clock drift ρ based on the two consecutive time information received ij , using ρ ij And the low-pass filter updates its drift compensation parameters In addition, the low-pass filter coefficients are modified to reduce the impact of packet loss on the convergence speed of the algorithm.

[0075]

[0076] in, is the drift compensation parameter value of the last update iteration, and its initial value is 1. α is a predefined weight coefficient, satisfying α∈(0,1). τ i (t) and τ i (t-1) is the time when node i receives time information twice in succession, and T is the synchronization period.

[0077] After the drift compensation converges, the logical clock drift of the nodes in the cluster is consistent with that of the cluster head node, and the logical clock deviation of the nodes in the cluster gradually converges. To reduce the impact of random delays, node i uses the data of node j received in a short period of time to perform linear regression analysis to estimate the relative clock deviation ρ ij Considering the fixed delay d′ during transmission, the deviation compensation parameter can be obtained

[0078]

[0079]

[0080]

[0081] Among them, β is the weight coefficient, satisfying β∈(0,1), τ j (t) and τ i (t) are the local hardware clocks of nodes i and j at time t, respectively, i d′ can be used as a priori value based on the measured value of the transmission delay between the two nodes. n is the number of data within the specified time, is the mean.

[0082] At this time, the nodes in the cluster gradually converge until they are consistent with the cluster head node.

[0083] (4) Algorithm simulation analysis

[0084] The effectiveness of the algorithm is verified by OPNET simulation and compared with the traditional distributed ATS (Average TimeSync, ATS) algorithm. The node initial clock compensation parameters are set as ρ ij (0) = 1. Local hardware clock parameter ρ i and θ i Randomly select from the uniformly distributed intervals [0.9995, 1.0005] and [0, 0.001] respectively. η=0.2,α=β=0.5,T=600s.

[0085] In a static network without random delay, 50 randomly placed nodes are divided into 10 clusters, each with several common nodes. The convergence speed of the primary network is related to the hierarchy. In this simulation, the logical clock drift of the primary network converges to the same value at the third iteration. The convergence speed of the secondary network is related to the convergence time of the cluster head node. The logical clock drift and deviation of each node in the cluster converge to the same value at the eighth iteration. Figure 5 As shown in the figure, compared with the traditional distributed ATS algorithm, the proposed algorithm converges at the eighth iteration, and the ATS algorithm converges at the tenth iteration. The logic clock drift convergence speed of the proposed algorithm is faster. In addition, the convergence speed of the proposed algorithm is only related to the number of layers. In a single synchronization cycle, as the network scale increases, the number of broadcast packets of the proposed algorithm is smaller. For dynamic networks with random delays in the communication process, in the simulation, the mean is set to 0 and the variance is 2.5×10 -7 The present invention uses maximum likelihood estimation in primary network synchronization and linear regression in secondary network synchronization, so that the logic clock synchronization error of the entire network is effectively reduced.

Claims

1. A wireless sensor network clustering fast time synchronization method for forest fire monitoring, the method comprising: Step 1: Topology generation; First, the cluster head node is elected from the gateway node outward, and the timer is set using the node's own remaining energy and distance. The timer is used to determine which node wins the election. If a node fails to win the election, it will mark itself as an ordinary node and select the nearest cluster head node as the parent node based on its own recorded cluster head information, and join the cluster headed by the node. Only when the number of time synchronization cycles reaches the preset number or a node fails, a new round of topology generation will be performed. Step 2: Main network time synchronization; The cluster head nodes form the main network and achieve time synchronization through a two-way exchange mechanism; let node A be the time reference node, node B be the node to be synchronized, and the timestamp τ A (t2) and τ A (t3) is the local time of node A, timestamp τ B (t1) and τ B (t4) is the local time of Node B; At time t1, node B initiates a synchronization request to node A. Node A receives the request at time t2 and replies with a message packet at time t3. Node B receives the message packet at time t4 and records the timestamp τ. B (t4); Taking the local time of node A as the reference time, the timestamp of node B is represented by node A; node B estimates the relative clock drift by using the synchronization reply message packets received twice adjacently, and further estimates the drift compensation parameter; Using the drift compensation parameters and deviation compensation parameters replied by node A, the clock of node B is adjusted to be consistent with the logical clock of node A. In the case of multi-hop in the main network, the first-level cluster head node is within the single-hop range of the sink node, and synchronizes with the sink node using this method. In turn, all nodes in the main network are directly or indirectly synchronized with the sink node, so that the logical clock of the entire main network is consistent with the sink node. Step 3: Secondary network time synchronization; The synchronization of the secondary network adopts a one-way listening mechanism. When a cluster head node in the main network initiates a message containing a synchronization request to the cluster head node of the upper level, it will broadcast the request. At this time, the cluster nodes in the cluster where the cluster head node is located will listen to and record the time information broadcast by the cluster head node; the relative time drift is estimated based on the two consecutive time information received, and the drift compensation parameters are updated in combination with the low-pass filter; the low-pass filter coefficient is further updated to reduce the impact of packet loss on the convergence speed of the algorithm; after the drift compensation converges, the logical clock drift of the nodes in the cluster is consistent with the cluster head node, and the logical clock deviation of the nodes in the cluster gradually converges.

2. A wireless sensor network clustering fast time synchronization method for forest fire monitoring as claimed in claim 1, characterized in that: The specific method of using the timer in step 1 to determine which node wins the election is as follows: The gateway node broadcasts an information packet, which contains information such as the node number and layer number; When other nodes within the single-hop communication range of the gateway node receive the packet, they calculate the distance between themselves and the gateway node based on the RSSI strength, and start the timer and set the timing time to elect the cluster head node based on their own relative remaining energy and distance; after the timer times out, a message packet is broadcast to the surrounding nodes, announcing that they have successfully run for the cluster head node; nodes that have not successfully run for the election clear the timer and become ordinary nodes.

3. A wireless sensor network clustering fast time synchronization method for forest fire monitoring as claimed in claim 1, characterized in that: The specific method of step 2 is: Taking the local time of node A as the reference time, node B estimates the relative clock drift using the synchronization reply message packets received twice in succession; node B adjusts its own time to be consistent with the local time of node A through the following calculation. t A (t2)=ρ BA (t B (t1)+d)+θ BA (1) t A (t3)=ρ BA (t B (t4)-d)+θ BA (2) t A (t3-1)=ρ BA (t B (t4-1)-d)+θ BA (3) t A (t)=τ′ B (t)=ρ BA t B (t)+θ BA (6) Among them, ρ BA represents the relative clock drift of node B relative to node A, d represents the fixed delay during transmission, and θ BA is the relative clock deviation of node B relative to node A, τ A (t3-1) represents the time when node A last replied to the message packet, τ B (t4-1) represents the time when node B last received the reply message packet, τ B (t) represents the local time of node B, τ′ B (t) represents the corrected node B time consistent with the local time of node A; Use the drift compensation parameters replied by node A and deviation compensation parameters Adjust the clock of Node B To the logical clock of node A Consistent, that is Complete synchronization; τ A (t) represents the local time of node A. Taking into account the influence of random delay, the joint maximum likelihood clock drift and deviation estimator corresponding to the Gaussian random delay model is used on this basis; at this time, the expression is as follows: t A (t2)=ρ BA (t B (t1)+d+X i )+θ BA (8) t A (t3)=ρ BA (t B (t4)-dY i )+θ BA (9) Among them, X i and Y i is a random delay, assuming and is zero mean and has a variance of σ 2 Gaussian distributed independent and identically distributed random variables, for the convenience of expression, introduce the constant Q, N is the total number of random delays.

4. A wireless sensor network clustering fast time synchronization method for forest fire monitoring as claimed in claim 1, characterized in that: The specific method of step 3 is: Step 3.1: The cluster head node j in the main network initiates a message to the cluster head node at the next level. Synchronization request, τj(t) represents the local time of cluster head node j, represents the drift compensation parameter of cluster head node j, represents the deviation compensation parameter of cluster head node j; the node i in the cluster where cluster head node j is located listens to and records the time information broadcast by cluster head node j, and estimates the relative clock drift ρ based on the two consecutive time information received ij , using ρ ij And the low-pass filter updates its drift compensation parameters Step 3.2: Modify the low-pass filter coefficients to reduce the impact of packet loss on the algorithm convergence speed; in, is the drift compensation parameter value of the last update iteration, and its initial value is 1; α is a predefined weight coefficient, satisfying α∈(0,1), τ i (t) and τ i (t-1) is the time when node i receives time information twice in a row, T is the synchronization period, and ρ ij (t) represents the relative clock drift at time t; Step 3.3: After the drift compensation converges, the logical clock drift of the nodes in the cluster is consistent with that of the cluster head node, and the logical clock deviation of the nodes in the cluster gradually converges; to reduce the impact of random delay, node i uses the data of node j received in a short period of time to perform linear regression analysis to estimate the relative clock drift ρ ij ; According to the fixed delay d′ during the transmission process, the deviation compensation parameter is obtained Among them, β is the weight coefficient, satisfying β∈(0,1), ρ j represents the slope of the clock at node j, θ j represents the clock deviation of node j, represents the kth τ i The value of (t), represents the mean, τ j (t) and τ i (t) are the local hardware clocks of nodes j and i at time t respectively, d′ is a priori value determined based on the measured value of the transmission delay between the two nodes; n is the number of data within the specified time.

Images