KR101171383B1 - Channel Management Method For Cognitive Radio Sensor Networks - Google Patents

Abstract

An object of the present invention is to provide a channel management method in an cognitive wireless sensor network for improving energy efficiency and effectively protecting a user (PU). In the cognitive radio sensor network of the present invention, a channel management method includes an operation channel sensing mode, a backup channel sensing mode, an operation channel change mode, a backup channel change mode, and data in a cognitive radio sensor network including a cluster head and a plurality of cluster members. Classifying a mode into an operation mode of the cognitive wireless sensor network; And a second step of adaptively determining the operation mode of the first step by using the Markov determination method based on the history of the channel sensing result and the last operation mode at the time before the start of each operation mode. It is done. According to the present invention, by determining the optimal operation mode based on the channel sensing history, it is possible to prevent unnecessary channel sensing and channel change, thereby improving energy efficiency. In addition, since channel sensing is performed at an appropriate time, first, there is an effect of better protecting the user (PU).

Description

Channel Management Method For Cognitive Radio Sensor Networks

The present invention relates to a channel management method in a cognitive wireless sensor network. More particularly, the present invention relates to a channel management method in a cognitive wireless sensor network in which an operation mode is adaptively determined to improve energy efficiency and protect a user (PU). It is about.

Wireless sensor networks are ultra-lightweight, low-power wireless sensors with low cost and low data rates. Most wireless sensor networks use unlicensed bands. However, because the band is already used by many communication systems, there is a problem that the spectrum for the wireless sensor network is insufficient.

In order to solve this problem, radio recognition (CR) technology may be considered. The wireless sensor network (hereinafter, referred to as 'CRSN') is a method of using a channel of a specific time and a specific area that is first authorized by a user (hereinafter, referred to as a 'PU') but is not used by the PU.

That is, since the PU has the privilege to use the channel, the CRSN should not interfere with the transmission of the PU. To this end, the CRSN periodically senses the operating channel to detect the activity of the PU, and if the CRSN detects the presence of the PU in the channel, the CRSN can quickly switch the operating channel to another empty channel (not used by the PU). Channel).

In addition, the CRSN can prepare a candidate channel (or candidate list) of a new operating channel (another empty channel) in advance, and can quickly change a channel by reducing the time for searching for another empty channel. The candidate channel is called a backup channel.

As mentioned above, the CRSN must perform CR specific operations such as channel sensing and channel switching as well as the operation of conventional sensor networks such as data transmission / reception, and additional CR specific operations require that the CRSN be more than conventional sensor networks. There was a problem of consuming energy.

In addition, this energy consumption has a problem of shortening the life of the CRSN. In addition, it is also important to protect the PU while reducing energy consumption in the CRSN.

In the past, many studies have been conducted to improve energy efficiency in the CRSN, and a new sensing structure for saving energy in the CRSN has been proposed, and a structure for reducing energy consumption by controlling power and data transmission rate in an operating channel is proposed. Suggested.

However, the proposed structure focuses on energy efficiency only in a specific operation mode (for example, channel sensing or data transmission), and does not consider the entire operation mode, which does not effectively reduce energy consumption.

In addition, there was a problem that does not effectively protect the PU.

The present invention has been made to solve the above problems,

An object of the present invention is to provide a channel management method in a cognitive wireless sensor network for improving energy efficiency.

In addition, an object of the present invention is to provide a channel management method in a cognitive wireless sensor network to effectively protect a user (PU).

In addition, an object of the present invention is to provide a channel management method in a cognitive radio sensor network to more accurately predict the presence of a user.

In the cognitive wireless sensor network, a channel management method includes an operation channel sensing mode, a backup channel sensing mode, an operation channel changing mode, a backup channel changing mode, and a data mode in a cognitive wireless sensor network including a cluster head and a plurality of cluster members. Categorizing into an operation mode of the cognitive wireless sensor network; And a second step of adaptively determining the operation mode of the first step by using the Markov determination method based on the history of the channel sensing result and the last operation mode at the time before the start of each operation mode. It is done.

Further, when the next operation mode is determined as the data mode at the determination time, the cluster head notifies the cluster member of the transmission schedule information, the cluster member transmits data to the cluster head according to the transmission schedule, The cluster member is in a dormant state immediately after the data transmission, and the cluster head is in a dormant state when receiving data from all cluster members, and when the data mode ends, the cluster head is in an active state and thus the history and the last operation mode of the channel sensing result. It is characterized in that to adaptively determine the next operation mode based on.

In addition, when the next operation mode is determined as the operation channel sensing mode at the determination time, the cluster head instructs all the cluster members to sense the operation channel, the cluster head and the cluster member sense the operation channel, The cluster member reports the sensing result of the operation channel to the cluster head, and the cluster head predicts the state of the operation channel based on the sensing result of the operation channel, and predicts the operation channel and the backup channel state and the last operation. It is characterized by adaptively determining the next operation mode based on the mode.

In addition, when the next operation mode is determined as the backup channel sensing mode at the determination time, the cluster head instructs all the cluster members to sense the backup channel, the cluster head and the cluster member sense the backup channel, The cluster member reports the sensing result of the backup channel to the cluster head, and the cluster head predicts the status of the backup channel based on the backup channel sensing result, and predicts the operation channel and the backup channel state and the last operation mode. It is characterized in that to adaptively determine the next operation mode based on.

In addition, when the next operation mode is determined as the operation channel change mode at the determination time, the cluster head randomly selects one channel as a new backup channel, and the cluster head detects the new operation channel and the new backup channel. Instructs all cluster members, the cluster head and the cluster member sense a new operating channel and a new backup channel that were backup channels, the cluster head sends additional beacons for synchronization in the new operating channel, and the cluster member A connection message for sensing and synchronization of a new operation channel and a new backup channel is transmitted to the cluster head, the cluster head sending a connection response message to the cluster member in response to the connection message, and the cluster head new Based on the sensing result of work channel and new backup channel, it predicts the status of new operation channel and new backup channel, and adaptively determines the next operation mode based on the predicted new operation channel, new backup channel status and last operation mode. Characterized in that.

In addition, when the next operation mode is determined as the backup channel change mode at the determination time, the cluster head randomly selects one channel as a new backup channel, and the cluster head selects all the cluster members to sense the new backup channel. And the cluster head and the cluster member sense the new backup channel, the cluster member reports the sensing result of the new backup channel to the cluster head, and the cluster head detects the sensing result of the new backup channel. It predicts the state of the new backup channel on the basis of, and adaptively determine the next operation mode based on the predicted operation channel, the new backup channel state and the last operation mode.

In addition, when it is determined that the operation channel sensing mode, the backup channel sensing mode, the operation channel changing mode, or the backup channel changing mode is determined at the determination time, the user may first quantize the statistical values for the channels measured through the channel sensing to a plurality of levels. And the statistical value is a value obtained by dividing the sum of the energy received from the cluster head and the cluster member by the noise spectral density.

The cluster head may select one of all operation modes at the time of determination, and notify the cluster member of the selected operation mode.

The determination time may be calculated based on a statistical value obtained through channel sensing, calculating a probability that a user exists first in each channel, and calculating an information vector based on a probability that the user exists first. And calculating the updated information vector based on the observation process.

In addition, the determination time is characterized in that for calculating the optimum value function to adaptively determine the operation mode based on the updated information vector.

In addition, the decision timing may be derived from a policy function based on the optimum value function, and adaptively determines an operation mode using the policy function.

According to the present invention, by determining the optimal operation mode based on the channel sensing history, it is possible to prevent unnecessary channel sensing and channel change, thereby improving energy efficiency.

In addition, since channel sensing is performed at an appropriate time, first, there is an effect of better protecting the user (PU).

In addition, by considering all the operation modes when determining the optimal operation mode, the energy efficiency is improved and the user (PU) can be better protected first.

In addition, since the presence of the user PU is first determined by quantizing the channel sensing result to a plurality of levels, the presence of the user PU on the channel can be more accurately predicted.

1 illustrates an embodiment of the present invention.
2 is a graph showing the disturbance rate of the preferred user (PU) according to the present invention.
Figure 3 is a graph showing the energy consumption rate according to the present invention.

Hereinafter, a preferred embodiment of a channel management method in a cognitive wireless sensor network according to the present invention will be described in detail with reference to the accompanying drawings. In the following description, well-known functions or constructions are not described in detail since they would obscure the invention in unnecessary detail.

In the cognitive radio sensor network of the present invention, a channel management method includes an operation channel sensing mode, a backup channel sensing mode, an operation channel change mode, a backup channel change mode, and data in a cognitive radio sensor network including a cluster head and a plurality of cluster members. Classifying a mode into an operation mode of the cognitive wireless sensor network; And a second step of adaptively determining the operation mode of the first step by using the Markov determination method based on the history of the channel sensing result and the last operation mode at the time before the start of each operation mode. It is done.

In the present invention, the cognitive wireless sensor network (hereinafter referred to as 'CRSN') is composed of one cluster head (hereinafter referred to as 'CH') and (N-1) cluster members as in the IEEE 802.15.4 shaping topology network. It is composed. There are M channels, which have the same bandwidth B.

All channels have the same priority user (hereinafter, referred to as 'PU'), and the active state that the PU transmits and the inactive state that the PU does not transmit alternately appear in the channel. The active (inactive) duration of the PU in the channel is distributed exponentially with an average of 1 / λ (1 / μ). Therefore, the probability P ₂ that the randomly selected channel is empty is μ / (λ + μ).

The channel where the PU is not detected is called an operation channel, and the CRSN operates in the operation channel. In addition, another channel for which no PU is detected is called a backup channel, and the CRSN manages the backup channel.

In addition, all operation modes of the CRSN are classified into five modes. The operation mode includes a data mode (DATA) for transmitting / receiving data, an operation channel sensing mode (SO) for sensing an operation channel, and a backup channel. The backup channel sensing mode SB, the operation channel change mode CO for changing the operation channel to the backup channel, and the backup channel change mode CB for changing the backup channel to another backup channel are included. The length of the operation mode is indicated as L _DATA , L _SO , L _SB , L _CO and L _CB .

Channel time is divided into frames, each frame starting with a beacon. At the start of each frame, i.e. at the time of decision, the CH determines one of the five modes of operation and notifies the cluster member via the beacon of the determined mode of operation. The frame length is equal to the length of the selected operation mode. Thus, the frame length is variable depending on the selected operation mode.

Hereinafter, the operation performed in each operation mode of the present invention will be described in detail.

At each decision time, CH is selected from five operation modes (data mode (DATA), operation channel sensing mode (SO), backup channel sensing mode (SB), operation channel change mode (CO) and backup channel change mode (CB)). One is determined and the cluster member is notified of the determined operation mode through a beacon.

When the next operation mode is determined as the operation channel sensing mode SO at the determination time, in the operation channel sensing mode SO, the CH instructs all cluster members to sense the operation channel, and the CH and all cluster members are operating channels. Sensing. The cluster member reports the sensing result to the CH. The sensing is to detect the presence of the PU in the operation channel.

Thereafter, the CH predicts the operation channel state based on the reported operation channel sensing result, and adaptively determines the next operation mode based on the operation channel sensing result, the predicted operation channel state, and the last operation mode.

At this time, the next operation mode is adaptively determined using a partially observable Markov determination method (hereinafter, referred to as 'POMDP').

If the next operation mode is determined as the backup channel sensing mode (SB) at the determination time, in the backup channel sensing mode (SB), the CH instructs all cluster members to sense a backup channel, and the CH and all cluster members are backup channels. Sensing. The cluster member reports the sensing result to the CH. The sensing is to detect the presence of a PU in the backup channel.

Thereafter, the CH predicts the state of the backup channel based on the reported backup channel sensing result, and adaptively determines the next operation mode based on the backup channel sensing result, the predicted backup channel state, and the last operation mode. At this time, the next operation mode is adaptively determined using the POMDP.

If the next operation mode is determined as the backup channel change mode (CB) at the determination time, in the backup channel change mode (CB), the CH randomly selects one channel as a new backup channel and senses the new channel. Command.

The CH and the cluster member sense the new backup channel, and the cluster member reports the sensing result to the CH. The sensing is to detect the presence of a PU in a new backup channel.

Thereafter, the CH predicts the state of the backup channel based on the reported backup channel sensing result, and adaptively determines the next operation mode based on the new backup channel sensing result, the predicted new backup channel state and the last operation mode. do. At this time, the next operation mode is adaptively determined using the POMDP.

When the next operation mode is determined to be the data mode DATA at the determination time, in the data mode DATA, the CH notifies the cluster member of the transmission schedule information through a beacon, and the cluster member transmits to the transmission schedule. Accordingly, data is transmitted to the CH. The cluster member is in a dormant state immediately after data transmission, and the CH is in a dormant state when receiving data from all cluster members.

At the end of the data mode DATA, the cluster head enters the wake up state to adaptively determine a next operation mode based on the history of the channel sensing result and the last operation mode. At this time, the next operation mode is adaptively determined using the POMDP.

When the next operation mode is determined as the operation channel change mode CO at the determination time, in the operation channel change mode CO, the CH and the cluster member are the new operation channel which was the backup channel and a new backup channel randomly selected by the CH. Sensing them all.

In addition, when the CRSN operates on a new channel, the CH and all cluster members must be newly connected. That is, the CH sends an additional Beacon to synchronize on the new operating channel, the cluster member sends a connection message to the CH to notify of successful connection, while the cluster member sends a new operating channel and a new backup The sensing result of the channel is transmitted to the CH.

The CH completes synchronization by sending a connection response message to the cluster member in response to the connection message to confirm the connection with the cluster member.

After synchronization, the CH adaptively determines the next operation mode based on the sensing results of the new operation channel and the new backup channel, the state of the predicted new operation channel and the new backup channel, and the last operation mode. If the next operation mode is not the operation channel change mode CO, it operates in the new operation channel. At this time, the next operation mode is determined using POMDP.

Hereinafter, the operation state of this invention on a channel is demonstrated in detail based on [FIG. 1].

Before transmitting and receiving data in an operation channel, a backup channel is selected. That is, in the backup channel sensing mode SB, the CH and the cluster member sense the backup channel, and the CH determines the next operation mode based on the sensing result. In FIG. 1, since the PU is detected in the previous backup channel from the sensing result, the PU determines the backup channel change mode CO for changing the backup channel to the next operation mode at the determination time.

In the backup channel change mode (CO), the CH instructs all the cluster members to sense the backup channel, the cluster member sends the sensing result of the backup channel to the CH after sensing the backup channel, and the CH sends the backup. A new backup channel is selected by predicting the status of the backup channel based on the channel sensing result.

As shown in FIG. 1, a new backup channel is selected in the backup channel change mode CO, and based on the backup channel sensing result, the predicted backup channel state and the last operation mode after the backup channel is selected. Adaptively determine the next mode of operation. In FIG. 1, the data mode DATA is determined as the next operation mode.

In the data mode DATA, the CH notifies the cluster member of transmission schedule information, the cluster member transmits data to the CH according to the transmission schedule, and the cluster member is idle immediately after the data transmission. The CH enters a dormant state when receiving data from all cluster members, and upon termination of the data mode, the cluster head is activated to adaptively determine the next operation mode based on the history of the channel sensing result and the last operation mode. In FIG. 1, the operation channel sensing mode SO is determined as the next operation mode.

In the operation channel sensing mode (SO), the CH instructs all the cluster members to sense the operation channel, the cluster member reports the sensing result of the operation channel to the CH after sensing the operation channel, and the CH The state of the operating channel is predicted based on the sensing result of the operating channel. In FIG. 1, since it is detected that the PU exists in the operation channel, it is determined as the operation channel change mode CO as the next operation mode.

Therefore, the channel that was the backup channel becomes a new operation channel, and in the operation channel change mode CO, the cluster head and the cluster member sense the new operation channel and the new backup channel that were the backup channel, and the cluster head receives the new operation channel. Transmits additional beacons for synchronization, the cluster member transmits a sensing result of a new operation channel and a new backup channel and a connection message for synchronization to the cluster head, and the cluster head responds to the connection message in response to the connection message. A message is sent to the cluster member, and the cluster head predicts a state of a new operation channel and a new backup channel based on the sensing result of the new operation channel and the new backup channel.

That is, when it is detected that the PU exists in the new operation channel, the operation channel change mode CO is determined as the next operation mode, and when it is detected that the PU does not exist, operation except the operation channel change mode CO as the next operation mode. The mode is determined.

In addition, a backup channel is determined according to whether a PU exists in the new backup channel. Therefore, a new operation channel, a new backup channel, and a next operation state are determined according to the result of sensing in the operation channel change mode CO.

1 illustrates an embodiment of the present invention, and the present invention can be implemented in various embodiments according to the state of a channel.

Hereinafter, a method of detecting a PU in a channel will be described in detail.

In the operation channel sensing mode SO, the backup channel sensing mode SB, the operation channel change mode CO, and the backup channel change mode CB, the statistical values of the channels measured through channel sensing are quantized to a plurality of levels. First, the existence of the user PU is determined, and the statistical value is a value obtained by dividing the sum of the energy received from the cluster head and the cluster member by the noise spectral density.

That is, when the CH selects the operation channel sensing mode (SO), the backup channel sensing mode (SB), the operation channel change mode (CO), the backup channel change mode (CB) as the operation mode, the CH through the beacon (Beacon) Notify the cluster member of the sensing start time and sensing period. The CH and cluster members then measure the energy received on the target channel (operation channel or backup channel) for a given sensing time and the cluster members report the measurement result (ie, the received energy) to the CH.

Since the measurement result is used to determine the operation mode of the next frame, the measurement result should be transmitted within the frame in which the measurement is performed. To this end, the CH allocates a transmission time to each cluster member so that it can report immediately after searching within the same frame.

The CH divides the sum of the energy received from all cluster members by the noise spectral density to produce a statistical value. Here, s _t ^(m) is referred to as a test statistic value for the channel m measured during the sensing period T _s in the (t-1) th frame. Under the assumption that channel m is empty (H ₀ ), the probability density function (pdf) of s _t ^(m) is a chi-square distribution with a degree of freedom of 2BT _s N. Under the assumption (H ₁ ) that the PU is in channel m, s _t ^(m) is the sum of the freedom of 2BT _s N and the non-central parameters E / N ₀ , E is the sum of the PU signal energy received by the cluster members. , N ₀ is the noise spectral density).

In addition, in order to more easily process the continuous random variable s _t ^(m) in the POMDP structure, the value of s _t ^(m) is quantized to k levels. To do this, we introduce (k + 1) thresholds, γ ₀ , ..., γ _k , where γ ₀ <γ ₁ <... <gamma _k , gamma ₀ = -∞, gamma _k = + ∞.

Where o _t ^(m) is the quantized value of s _t ^(m) , and for k = 1, 2, ..., k, if γ _k-1 <s _t ^(m) <γ _k , o _t ^(m) = k. In addition, the following conditional probabilities may be limited.

ν ₀ (k) and υ ₁ (k) can be easily calculated with pdf of s _t ^(m) .

Hereinafter, a method of adaptively determining an operation mode using the POMDP will be described in detail.

The present invention uses a POMDP structure that allows for uncertainty of information when adaptively determining the operation mode. This is because the CRSN cannot accurately determine the state of the operation channel and the backup channel, and can only speculate whether the PU exists in the operation channel or the backup channel based on the result of channel sensing.

First, the structure of the POMDP will be described. The state of the core process X _t indicates the state of the operation channel and the backup channel at the determination time t. Here, the operation channel is displayed as '1' and the backup channel is displayed as '2'. Thus, X _t = (x _t ⁽¹⁾ , x _t ⁽²⁾ ), where x _t ^(m) = 0 means that channel (m) is empty, and x _t ^(m) = 1 It indicates that a PU exists in channel m.

Action A _t represents the operation mode determined by CH at decision time t, and A represents the set of all operable modes of operation. That is, A = {DATA, SO, SB, CO, CB} and A _t = A. (DATA is the data mode, SO is the operating channel sensing mode, SB is the backup channel sensing mode, CO is the operating channel changing mode, and CB is the backup channel changing mode.)

The transition probability P of the core process becomes P (a) = [p _{(i, j) (i ', j')} (a)] when the action (a) is taken, Where p _{(i, j) (i ', j')} (a) = Pr {X _{t + 1} = (i ', j') | X _t = (i, j), A _t = a}. The transition probability p _{(i, j) (i ', j')} (a) can be calculated using the following two probabilities. That is, when action (a) is taken, u _{i, i '} (a) is the probability that the channel transitions from state (i) to state (i'), and w _i (a) is a channel randomly determined at decision time. The probability of being in state (i) immediately before this next decision.

Then u _0,0 (a) = e ^-λLa , u _0,1 (a) = 1-e ^-λLa , u _1,0 (a) = 1-e ^-μLa , u _1,1 (a) = e- ^μLa .

And w ₀ (a) = P _e ^-λLa + (1-P _e ) (1-e ^-μLa ), w ₁ (a) = (1-P _e ) e ^-μLa + P _e (1-e ^-λLa ).

For a '{DATA, SO, SB}, p _{(i, j) (i', j ')} (a) = u _{i, i'} (a) x u _{j, j '} (a). If an action CO is taken, p _{(i, j) (i ', j')} (CO) = u _{j, i '} (CO since the backup channel becomes a new operating channel and the new backup channel is randomly selected. ) X w _{j '} (CO). Furthermore, since the new backup channel can be randomly selected under the action CB, p _{(i, j) (i ', j')} (CB) = u _{i, i '} (CB) x w _j' (CB) do.

At decision time t, the state of the observation process is represented by o _t = (o _t ⁽¹⁾ , o _t ⁽²⁾ ), where o _t ^(m) (m = 1,2) is o _t ^(m) ∈ { The quantized statistical values obtained by sensing the channel m during the 1,2, ..., k} and (t-1) th frames. When the CRSN does not sense channel m and does not have a statistic for channel m, o _t ^(m) is set to '0'. Thus, for A _t-1 ∈ {DATA, SB, CB}, o _t ⁽¹⁾ = 0, and for A _t-1 ∈ {DATA, SO}, o _t ⁽²⁾ = 0.

The probabilistic relationship between the core process state and the observed process state is called Q (a) = [q _{(i, j) (k, l)} (a)]. That is, q _{(i, j) (k, l)} (a) = Pr {O _t = (k, l) | X _t = (i, j), A _t-1 = a}. If n (k) is the probability that o _t ^(m) = k for the unsensed channel ^m, then n (0) = 1, regardless of channel state x _t ^(m) , for all k≥1 n (k) = 0. Then, q _{(i, j) (k, l)} (a) may be defined through ν _i (k) (Equation 1, Equation 2) and n (k).

Since the CRSN does not sense a channel under the action DATA, q _{(i, j) (k, l)} (DATA) = n (k) × n (l). Considering that only the operation channel is sensed in the action SO, q _{(i, j) (k, l)} (SO) = υ _i (k) × n (l). Also, under the action CB or SB where only the backup channel is sensed _, q _{(i, j) (k, l)} (CB) = n (k) × υ _j (l) and q _{(i, j) (k, l)} (SB) = n (k) × υ _j (l). Furthermore, since the CRSN senses both the operation channel and the backup channel under action CO, q _{(i, j) (k, l)} (CO) = υ _i (k) × υ _j (l).

r _{(i, j) (i ', j')} (a) taking action (a) from state (i, j) and causing a transition from state (i ', j') to the instantaneous compensation that the CRSN receives ( Immediate Reward). Then, the total expected reward obtained by performing action (a) in state (i, j) is

to be. The compensation should be determined to reduce energy consumption while adequately protecting the PU. Therefore, the value of compensation is set in consideration of the probability of disturbing the PU and the energy consumed.

To determine the operation mode, the CRSN manages an information vector ∏ (t) that summarizes all the information needed to make a decision at decision time t. The information vector ∏ (t) is represented by [π _(0,0) (t), π _(0,1) (t), π _(1,0) (t), π _(1,1) (t)]. Where π _{(i, j)} (t) is the probability that there is a core process (X _t ) in state (i, j). Since the CRSN has no information on the channel state at network initialization, it is assumed that the initial information vector ∏ (0) is a stable channel state probability vector.

이다.therefore,

to be.

According to the characteristics of the POMDP, the information vector 추 (t + 1) can be inferred from the action A _t , the information vector ∏ (t) and the observation O _{t + 1} . The conversion from t (t) to ∏ (t + 1) can be specified by Bayes' formula as follows.

At decision time t + 1, CH calculates the updated information vector 1 (t + 1) using Equation (4). In order to select an action with the updated information vector, the CH depends on the policy δ, which is a mapping function that maps from the information vector to the action. There may be many viable policies δ. Of all possible policies δ, we need to find the optimal policy δ ^* that maximizes the expected discounted infinite horizon compensation with the discount coefficient β (0 ≦ β ≦ 1).

In order to find the optimal policy δ ^* (∏), we define the optimal value function V ^* (∏) as follows.

here,

이다.

to be.

Then, the optimal policy δ ^* (k) is derived using the optimum value function V ^* (k).

The optimal value function V ^* (i) and the optimal policy δ ^* (i) can be calculated according to a value iteration based on dynamic programming. Furthermore, the optimal policy can be derived offline before starting the CRSN.

By the above-described method, at each decision time, the CH updates the information vector based on the last sensing result and the last action. The CH then determines the optimal mode of operation a ^* using the calculated optimal policy δ ^* that maps the information vector to a ^* .

Hereinafter, a simulation result using the channel management method in the cognitive wireless sensor network of the present invention will be described in detail.

The CRSN consists of eight sensor nodes (one CH and seven cluster members) with the basic performance of IEEE 802.15.4 WSN. As in IEEE 802.15.4, the bandwidth B is set to 2 MHz and the transmission rate of the CRSN is fixed at 250 kbps. The lengths of the five modes of operation are L _DATA = 50 ms, L _SO = L _SB = L _CB = 5.8 ms, and L _CO = 15.1 ms.

In the simulation, the size of the beacon, the sensing result report packet, the sensing result report / connection request packet, the connection response packet and the data packet are set to 40, 22, 31, 27 and 131 bytes, respectively. It is assumed that the cluster members send the sensing result report packet or data packet to the CH without collision according to the time division method. In any mode of operation, the sensor node is idle when it does not transmit or receive traffic.

Since the coverage of the CRSN is much smaller than the transmission range of the PU, it is assumed that the PU signal power received at each sensor node is the same. According to the FCC, the CR system detects a PU when the power received from the PU is higher than -114 dBm in a bandwidth of 6 MHz. However, since the bandwidth of the CRSN is 2 MHz, it is assumed that only one third of the PU signal power is collected at each sensor node. The active period (1 / λ) and the inactive period (1 / μ) of the PU in the channel are set to 30s and 10s, respectively, T _s = 2ms and N ₀ = -163 dBm / Hz.

이 된다. 여기서, γ₀=-∞이고, γ_k=+∞이다.
The discount coefficient β = 0.95 and the quantization level k = 20. The quantization thresholds γ ₁ , γ _k-1 are determined to satisfy ν ₀ (0) = 0.001 and ν ₀ (k-1) = 0.001. Then, for 1 <k <k-1,

. Where γ ₀ = −∞ and γ _k = + ∞.

In addition, the value of the instantaneous compensation used in the simulation is investigated. Considering that data transmission on the channel where the PU exists (i.e., the presence channel of the PU) should be avoided for PU detection but transmission on an empty channel needs to be facilitated, r _{(1,?) (? ,?) Set} (DATA) =-10 and r _{(0,?) (?,?)} (DATA) = 10.

On the other hand, sensing an empty channel should be avoided in order to prevent energy waste due to unnecessary sensing. The compensation for sensing an empty channel is determined based on the energy consumed in the operation mode. If E _a is the total energy consumption of the CRSN in operation mode (a), then r _{(0,?) (?,?)} (SO) =-E _SO / E _SB = -1 and r _{(?, 0) (? (?)} =-E _SB / E _SB = -1

이고, r_(?,0)(?,0)(CB)=r_(?,1)(?,1)(CB)=-E_CB/E_SB=-1이 된다.Likewise, changing an empty channel (the channel on which the PU exists) to another empty channel (the channel on which the PU exists) results in unnecessary energy consumption. therefore,

R _{(?, 0) (?, 0)} (CB) = r _{(?, 1) (?, 1)} (CB) =-E _CB / E _SB = -1.

Furthermore, since changing an empty channel to a channel in which the PU exists is much worse than moving between channels in the same state, we have additional penalties for changing the empty channel to a channel in which the PU exists. (indicated by ξ _c ). Therefore, r _{(0,?) (1,?)} (CO) = ξ _c × r _(0,?) ( _0,?) (CO), and r _{(?, 0) (?, 1)} (CB) = ξ _c × r _{(?, 0) (?, 0)} (CB). In the simulation, ξ _c = 2. Therefore, r _{(0,?) (1,?)} (CO) =-5, and r _{(?, 0) (?, 1)} (CB) =-2.

On the other hand, sensing the channel in which the PU exists is desirable for fast PU detection. Furthermore, since the operating channel needs to be treated as more important than the backup channel, the weighting channel w ₀ is assigned to the operating channel. That is, r _{(1,?) (?,?)} (SO) = w ₀ x r _{(?, 1) (?,?)} (SB). Also, changing a channel in which the PU exists to an empty channel may be a more urgent operation than simply searching for a channel in which the PU exists for PU detection. To reflect this urgency, another weight w _c is introduced for channel change. Then r _{(1,?) (0,?)} (CO) = w _c × r _{(1,?) (?,?)} (SO), and r _{(?, 1) (?, 0)} (CB) = w _c xr _{(?, 1) (?,?)} (SB) In the simulation w ₀ = 5, w _c = 2 and r _{(?, 1) (?,?)} (SB) = 1. Then r _{(1,?) (?,?)} (SO) = 5, r _(1,?) ( _0,?) (CO) = 10 and r _{(?, 1) (?, 0)} (CB) = 2

The present invention is compared with the fixed sensing period structure in which the operating channel and the backup channel are sensed in each fixed period. In the simulation, the CRSN having a fixed sensing period structure senses the operation (backup) channel after performing the DATA mode ζ (2ζ) times. In channel sensing, the sensor node determines that the PU exists in the corresponding channel if the sum of the signal energies received in the sensing channel is greater than the detection threshold and immediately instructs the cluster members to move to the backup channel. . For the received PU power, the detection threshold is determined such that the PU false alarm probability is equal to the PU miss detection probability.

2 illustrates a PU disturbance ratio defined as a time portion in which a channel in which a PU exists for transmission of traffic of a CRSN during a time in which a PU exists in a channel is used. It can be seen from FIG. 2 that the PU is more properly detected in the structure according to the present invention than in the fixed sensing period structure. That is, since the present invention manages the information vector tracking the channel state with the history of the sensing result, it is possible to predict the state of the operation channel and the backup channel more accurately.

Furthermore, the present invention always determines the mode of operation to optimize the compensation while meeting the design goals of low energy consumption and high PU protection. As a result, based on the more accurately predicted operating channel and backup channel status, the CRSN of the present invention senses the channel where the PU is likely to be active more frequently, and the CRSN moves to an empty backup channel more quickly when the PU appears in the active channel. do. Therefore, in the present invention, the PU disturbance ratio is made low.

3 shows the ratio of energy consumed for channel sensing and change to total energy consumed. As shown in FIG. 2, the present invention can more accurately predict the channel state by tracking the history of the sensing result. Furthermore, the present invention reduces negative channel searching and switching by granting negative compensation to sensing and changing the empty channel.

As a result, the present invention consumes less energy in CR specific operation than a fixed sensing period structure in which the channel is periodically sensed and channel switching unconditionally performs channel switching for alarm detection from one sensing.

In addition, it can be observed from FIGS. 2 and 3 that the present invention improves energy efficiency of about 50% with significant PU protection performance compared to a fixed sensing period structure with ζ = 30.

As described above, although an embodiment of the present invention has been described, the technical idea of the present invention is not limited to the above embodiment, and a channel management method may be implemented in various cognitive wireless sensor networks in a range that does not depart from the technical idea of the present invention. .

Claims (11)

In a cognitive wireless sensor network comprising a cluster head and a plurality of cluster members,
Classifying an operation channel sensing mode, a backup channel sensing mode, an operation channel change mode, a backup channel change mode, and a data mode into an operation mode of the cognitive wireless sensor network; And
And a second step of adaptively determining the operation mode of the first step by using the Markov determination method based on the history of the channel sensing result and the last operation mode at the time before the start of each operation mode. How to manage channel in cognitive wireless sensor network. The method of claim 1,
The second step comprises:
If the next operation mode is determined as the data mode at the determination time,
The cluster head notifies the cluster member of transmission schedule information,
The cluster member transmits data to the cluster head according to the transmission schedule,
The cluster member is in a dormant state immediately after the data transmission,
The cluster head is idle when receiving data from all cluster members,
And the cluster head is activated when the data mode ends, and adaptively determines a next operation mode based on a history and a last operation mode of a channel sensing result. The method of claim 1,
The second step comprises:
If the next operation mode is determined as the operation channel sensing mode at the determination time,
The cluster head instructs all the cluster members to sense an operating channel,
The cluster head and the cluster member sense the operating channel;
The cluster member reports the sensing result of the operation channel to the cluster head,
The cluster head predicts a state of an operating channel based on a sensing result of the operating channel,
A channel management method in a cognitive radio sensor network, characterized in that adaptively determines the next operation mode based on the predicted operation channel, the backup channel state, and the last operation mode. The method of claim 1,
The second step comprises:
If the next operation mode is determined as the backup channel sensing mode at the determination time,
The cluster head instructs all the cluster members to sense a backup channel,
The cluster head and the cluster member sense the backup channel,
The cluster member reports the sensing result of the backup channel to the cluster head,
The cluster head predicts the state of the backup channel based on the backup channel sensing result,
A channel management method in a cognitive radio sensor network, characterized in that adaptively determines the next operation mode based on the predicted operation channel, the backup channel state, and the last operation mode. The method of claim 1,
The second step comprises:
If the next operation mode is determined as the operation channel change mode at the determination time,
The cluster head randomly selects one channel as a new backup channel,
The cluster head instructs all the cluster members to sense a new operating channel and a new backup channel,
The cluster head and the cluster member sense a new operation channel and a new backup channel that were backup channels,
The cluster head transmits additional beacons for synchronization in a new operating channel,
The cluster member transmits a connection message for sensing and synchronization of a new operation channel and a new backup channel to the cluster head,
The cluster head transmits a connection response message to the cluster member in response to the connection message;
The cluster head predicts the state of the new operation channel and the new backup channel based on the sensing results of the new operation channel and the new backup channel.
A channel management method in a cognitive wireless sensor network, characterized in that adaptively determines the next operation mode based on the predicted new operation channel, the new backup channel state, and the last operation mode. The method of claim 1,
The second step comprises:
If the next operation mode is determined as the backup channel change mode at the determination time,
The cluster head randomly selects one channel as a new backup channel,
The cluster head instructs all the cluster members to sense the new backup channel,
The cluster head and the cluster member sense the new backup channel,
The cluster member reports the sensing result of the new backup channel to the cluster head,
The cluster head predicts the state of the new backup channel based on the sensing result of the new backup channel,
A channel management method in a cognitive radio sensor network, characterized in that adaptively determines the next operation mode based on the predicted operation channel, the new backup channel state, and the last operation mode. The method of claim 1,
The second step comprises:
If it is determined that the operation channel sensing mode, the backup channel sensing mode, the operation channel change mode or the backup channel change mode at the determination time,
Characterizing the presence of the user (PU) by first quantizing the statistical value for the channel measured through the channel sensing to a plurality of levels,
The statistical value is a sum of the energy received from the cluster head and the cluster member divided by the noise spectral density, channel management method in a cognitive radio sensor network. 8. The method according to any one of claims 1 to 7,
The cluster head selects one of all operation modes at the time of determination, and notifies the cluster member of the selected operation mode. The method of claim 8,
The decision time is
Based on the statistics obtained through channel sensing, the probability that a user exists first in each channel is calculated.
Calculating an information vector based on a probability that the user exists first,
And calculating an updated information vector based on the information vector, an operation mode, and an observation process. 10. The method of claim 9,
The decision time is
And calculating an optimum value function to adaptively determine an operation mode based on the updated information vector. The method of claim 10,
The decision time is
Deriving a policy function based on the optimum value function, and adaptively determine the operation mode using the policy function, Channel management method in a cognitive radio sensor network.

Images