JP4793198B2 - Multiple access communication method - Google Patents

Description

  The present invention relates to an adaptive back-off type multiple access communication method used for asynchronous transmission by a plurality of terminals in a media sharing type transmission system.

  In the media sharing type transmission method, an access method for resolving access contention between terminals is required. In particular, aloha (ALOHA) and CSMA (Carrier Sense Multiple Access) are well known as distributed access communication methods that do not require a central control device such as a base station. A collision avoidance type carrier sense multiple access communication method called CSMA / CA (Carrier Sense Multiple Access with Collision Avoidance) is used in a transmission method such as wireless or power line communication where signal attenuation is large and collision detection is difficult.

A typical example is an access method called DCF (Distributed Coordination Function) defined by IEEE802.11. In this method, as shown in FIG. 6A, two frame intervals of SIFS (Short Inter Frame Space) and DIFS (Distributed access Inter Frame Space) are used, and each terminal receives an ACK frame that informs that the DATA frame has been received correctly. Transmission is performed after SIFS has passed, and when a DATA frame is sent, it is detected whether another terminal is sending on the communication medium before the DIFS time has passed and the packet is sent after the idle time. Perform carrier sense, wait for a predetermined time (DIFS) to elapse after the communication medium becomes empty, and then randomly select from the range of 0 to the contention window value (hereinafter abbreviated as CW <Contention Window>). Select a value for the backoff time, with the slot time as the smallest unit, the value of the backoff time x slot time To start transmission after a lapse. Even when two or more terminals attempt to transmit simultaneously, collision can be avoided by selecting a back-off value randomly.

  That is, the terminal that has selected the smallest back-off value starts transmission first, and when the other terminals detect that the terminal has started transmission by carrier sense, the back-off process is interrupted and the communication media is again transmitted. The back-off process is resumed after it becomes free.

  The backoff value is selected at random, but if the same value is accidentally selected, a collision occurs. Each time a collision occurs, an exponential back-off is performed that doubles the CW and expands the range of random values to reduce the collision probability.

  If the number of terminals (competing terminals) to be transmitted simultaneously increases, the probability of selecting the same back-off value increases, collisions occur frequently, and throughput decreases. In IEEE802.11a, the initial value of CW is defined as 15, and is increased exponentially to 31, 63, 127,.

  Increasing the initial value of CW reduces the collision probability and suppresses the decrease in throughput even when there are many competing terminals. On the other hand, if there are few competing terminals, the wasteful waiting time increases and the throughput decreases. (FIG. 6B shows the relationship between the number of terminals and the throughput when the minimum value of CW is changed to 3, 7,..., 127, 255).

  In order to solve these problems, there is provided a method for measuring the use status of communication media and dynamically changing the random range of the backoff value accordingly (for example, Patent Document 1).

In order to reduce the collision probability, there is provided a method for adding a different offset value for each terminal to a random back-off value to avoid matching the back-off values. (For example, patent document 2). According to these methods, it is possible to increase the throughput by selecting an appropriate backoff value according to the number of competing terminals.
Japanese Patent Laying-Open No. 2005-12275 (paragraph 0023) Japanese Patent Laying-Open No. 2005-64795 (paragraphs 0025 to 0028)

  By the way, since the access method of IEEE802.11 selects a random back-off time value, the average back-off time of each terminal is CW / 2 when considered over a long period of time, and a transmission opportunity can be obtained fairly. When considered in a short time, the transmission opportunities of terminals that continuously select a small back-off time value increase, and the fairness decreases.

  In addition, while waiting time increases exponentially in the terminal that has caused a collision, a terminal that can transmit without a collision happens to change the random backoff time value from the range of the initial value of CW next time. Since it is selected, the waiting time is small, and the probability that a terminal that has succeeded in transmission continuously transmits is not low. FIG. 7A shows the result of calculating the number of successful transmission packets for each terminal by computer simulation when the number of competing terminals is 16 in the IEEE 802.11 access method, and is 50 packets / second. It varies in the range of ~ 250 packets / second, and a difference of up to 5 times can be seen.

  In the methods disclosed in Patent Document 1 and Patent Document 2, the purpose is to improve transmission efficiency (total throughput), and fairness is not improved. For example, when the number of competing terminals is 16 in FIG. 6B, the transmission efficiency is improved if the CW = 63 with respect to the default CW = 15 of IEEE802.11 (the total throughput increases). However, in this case, the number of successfully transmitted packets is as shown in FIG. 7B, and although the width of variation is slightly smaller than that in FIG. 7A, it is 70 packets / second to 210 packets / second. There is still about 3 times the variation.

  The present invention has been made in view of the above-described points. The purpose of the present invention is to obtain a transmission opportunity earlier for a packet having a higher priority class with respect to the content of the packet, regardless of the number of competing terminals. Therefore, it is an object of the present invention to provide a multiple access communication method capable of ensuring high throughput and ensuring fairness in which communication of the same priority class can be obtained in all terminals at the same level.

  In order to achieve the above object, in the invention according to the multiple access communication method of claim 1, a random back-off time for avoiding a collision after each terminal detects a free state of communication media by carrier sense. A multiple access communication method for starting transmission after a lapse of time, classifying a class having a plurality of priorities according to a type of a transmission packet in advance, with respect to a predetermined class, from 0 to the predetermined number A random value is uniformly selected from integers up to the contention window value corresponding to the class, and a value obtained by adding the selected value and the offset value is used as a backoff time, and the random value and the offset value are selected. Is calculated for each priority class, and each terminal observes the traffic load status of each class on the communication medium with a predetermined index, and the predetermined class And wherein the dynamically changing the offset value of the plant staple th classes according to the traffic load status.

  In the invention according to the multiple access communication method of claim 2, the multiple access is started in which each terminal starts transmission after a random backoff time for collision avoidance has elapsed after detecting a free state of communication media by carrier sense. A communication method, which is classified into classes having a plurality of priorities in advance according to the type of transmission packet, and contention window values corresponding to the predetermined class from 0 to the predetermined class From the integer up to, select a uniformly random value, the value obtained by adding the selected value and the offset value as a backoff time, the random value and the offset value are calculated for each priority class, At each terminal, the traffic load status on the communication media is observed with a predetermined index, and depending on the traffic load status, all classes Wherein the dynamically changing collectively serial offset value.

    The invention according to claim 3 is characterized in that, in the invention according to claim 1, the traffic load situation is observed using the number of competing terminals estimated based on idle time as a predetermined index.

  According to a fourth aspect of the present invention, there is provided the multiple access communication method according to the first aspect, wherein the traffic load state is observed using the idle time as a predetermined index.

  The invention according to a fifth aspect of the present invention relates to the multiple access communication method according to the first aspect, wherein the traffic load state is observed using the number of times the back-off is interrupted as a predetermined index.

  The invention according to a sixth aspect of the present invention relates to the multiple access communication method according to the first aspect of the invention, wherein the traffic load state is observed using the number of terminals that have transmitted within a predetermined time as a predetermined index.

In the invention according to any one of claims 1 to 6, a field indicating the current offset value is provided in a header of a transmission packet, and each terminal that has received the transmission packet observes a traffic load situation. In addition, it is desirable to calculate a predetermined index using the offset value in the received header.

Further, in the invention according to any one of claims 1 to 6, a field indicating the current offset value is provided in a header of a transmission packet, and each terminal receiving the transmission packet observes a traffic load situation In this case, it is desirable to perform weighting according to the offset value in the received header.

  According to the multiple access communication method of the present invention, a packet having a higher priority class with respect to the packet content can obtain a transmission opportunity earlier and realize so-called QoS (Quality of Service), and at the same time, regardless of the number of competing terminals. In addition, high throughput can be ensured, and in addition, in communication with the same priority class, it is possible to ensure fairness with which all terminals can obtain the same transmission opportunity.

(Embodiment 1)
The method of this embodiment is adopted in a communication system that performs communication by using, for example, a power line as a transmission line and superimposing a power line carrier signal on the power line, and FIG. 1A is used in the communication system. The circuit structure of the terminal for power line carrier communications is shown.

  This terminal transmits a transmission packet created by a signal processing unit (not shown) to the modulation unit 3 via the transmission buffer 1 and the transmission control unit 2 and modulates the transmission packet, and then transmits from the transmission AFE (analog front end) unit 4. A transmission function for transmitting the power line carrier signal on the power line L via the coupler 5 and a power line carrier signal on the power line L via the coupler 5 into the receiving AFE unit 6 and a demodulator from the fetched power line carrier signal 7 is provided with a receiving function for demodulating the received packet and fetching it into a signal processing unit (not shown) through the reception control unit 8 when this received packet is addressed to itself. In addition, a carrier detection unit 9 is provided that outputs to the transmission control unit 2 a signal indicating that the power line carrier signal on the power line L is no longer detected through the reception AFE unit 3.

  As shown in FIG. 1 (b), the transmission control unit 2 measures the idle time from when the signal indicating the vacant state is output from the carrier detection unit 9 until the start of transmission or reception of a packet after the passage of DIFS. The idle time observation unit 21 that calculates the number of competing terminals M as an index corresponding to the traffic load situation from the measured value, and the offset value as described later using the number of competing terminals M calculated by the idle time observation unit 21. The offset value setting unit 22 for obtaining the value, and selecting a random value uniformly from integers from 0 to CW, adding the value and the above-described offset value to obtain a backoff time, and the backoff time has elapsed And a packet transmission processing unit 23 for performing control to start transmission of the transmission packet described above. FIG. 2A shows an observation state of idle time with respect to packet transmission / reception timing.

  Also, the reception control unit 8 sends an ACK signal to the transmission buffer 1 when SIFS has elapsed when it correctly receives a reception packet addressed to itself, and transmits the transmission control unit 2, the modulation unit 3, the transmission AFE unit 4, and the combiner 5 The function of transmitting as a power line carrier signal is provided.

Here, if the above-mentioned backoff time is Backoff, the selected random value is rand (0, CW), and the offset value is CWoffset, the backoff time is
Backof = rand (0, CW) + CWoffsetf
It becomes.

  Here, assuming that CW is 15 and the number of competing terminals is 2 to 16, the offset value (CWoffset) is changed from the IEEE802.11 standard to each slot of 3, 7, 15, 31, 63, 127. When the total throughput was calculated by computer simulation, a result as shown in FIG. 2B was obtained.

  As a result, when the number M of competing terminals is small, the smaller the offset value (CWoffset) is, the lower the waiting time is and the transmission efficiency is high (throughput is high). When the number M of competing terminals was 16, the highest efficiency was obtained when CWoffset = 63. This is because the collision is reduced by increasing the back-off time.

  In the method of the present embodiment, the idle time observation unit 21 described above observes the idle time, and the offset value setting unit 22 appropriately sets the CW and the offset value (CWoffset) based on the idle time, as will be described later. It is characterized by reducing collisions while ensuring high throughput.

  First, when CWoffset is 0, the backoff time is a random value from 0 to CW. Therefore, the probability P of transmission in a certain slot is obtained by the following equation (1).

  Further, when the number of competing terminals is M, the probability q that one or more devices transmit in a certain slot is obtained by the following equation (2).

  Further, the average idle time is obtained by the following equation (3).

  Further, if CWoffset is taken into consideration, the idle time becomes longer by an average of CWoffset / M. Therefore, the average idle time in this case is obtained by the following equation (4).

  By the way, in the method of this embodiment, for example, four priority classes are provided according to the contents of the packet to be transmitted, and the values of CW and CWoffset are set for each class according to the above-described idle time. . The correspondence between the transmission packet and the priority class is, for example, a CoS (Class of Service) field in a VLAN (Virtual LAN) tag defined in IEEE802.1Q, ToS (Type of Service) in an IP (Internet Protocol) header, or A value such as DSCP (Differentiated Services Code Point) may be used, or may be explicitly specified from an application.

  Here, class numbers are set to 0 to 3, and CW and CWoffset are defined as shown in Table 1 so that priority is given to a class having a larger number. Here, as in the case of IEEE 802.11e, the CW value is decreased for higher priority classes.

  Then, a field indicating a class number is provided in the MAC header, or a class is determined by reading a VLAN-CoS or IP-ToS field, and the idle time observation unit 21 observes the idle time for each class. For example, when a packet received after the lapse of DIFS is class 3, the class 3 idle time is averaged with the past value.

  Then, a threshold value of idle time is determined for each class, and when it is equal to or greater than the threshold value, the value of CWoffsetA in Table 1 is adopted as the offset value, and when it falls below the threshold value, the value of CWoffsetB is adopted. CWoffsetA uses a value of CW + CWoffsetA + 1 of the one higher class so that the higher class can get an opportunity to transmit first.

  In Table 1, the class 3 offset value (CWoffset) is set to 0 for the following reason. That is, since class 3 is used for voice, each terminal transmits a fixed-length packet at regular time intervals, and no bursty traffic occurs. Therefore, even if the number M of competing terminals increases, fairness is not inherently lost, so there is no need to increase the offset value (CWoffset). Rather, since it is important that the voice packet has a low delay, CWoffsetA and CWoffsetB of class 3 are always set to 0.

  On the other hand, the use of class 2 is video, but in the case of variable rate, the amount of data transmitted by each terminal is not exactly constant and varies somewhat. Accordingly, since there are some bursty elements, CWoffsetB is set to a value about twice that of CW.

  Since burst traffic occurs for classes 1 and 0, CWoffsetB is set to a value about twice that of CW in order to ensure fairness when the number of competing terminals M is large (= the idle time is short).

  Next, the processing operation of the terminal using the method of this embodiment will be described in detail with reference to the flowchart of FIG.

  First, the carrier detection unit 9 constantly monitors whether or not the transmission line (communication medium) becomes empty depending on whether or not the power line carrier signal is detected on the power line L through the reception AFB unit 6 (step S1). When the free state is entered, a signal indicating that the free state is entered is output to the packet transmission processing unit 23 and the idle time observation unit 21.

  The packet transmission processing unit 23 counts the time until the DIFS elapses after becoming empty (step S2), and checks whether there is a transmission packet in the transmission buffer 1 when the DIFS elapses (step S3). If there is a transmission packet, it is checked whether or not the back-off time is not set (step S4). If it is not set, the offset value setting unit 22 determines, for example, that the priority is a predetermined class (k). A back-off time is set by adding a random value <rand (0, CW (k))> from 0 to CW (k) in FIG. 4 and an offset value (CWoffset (k)) calculated as described later ( Step S5).

  If it is determined in step S3 that there is no transmission packet, or if it is determined in step S4 that the back-off time is set, or if step S5 is completed, step S6 is executed. Based on the detection of the carrier detection unit 9, a check is made as to whether or not a packet from another terminal has been received. If not, a check is made again as to whether or not the back-off time has not been set. Performed in S7. If it is not set here, the process returns to step S1. If it is set, the packet transmission processing unit 23 counts the time until the time of one slot elapses (step S8). After the time has elapsed, the back-off time value is counted down (step S9), and whether or not the back-off time value becomes “0” by this count-down is checked in the next step S10. Returning to Step 6, the processes from S6 to S10 are repeated. If it is determined in step S10 that the back-off time value is “0”, transmission processing for the transmission packet in the transmission buffer 1 is performed (step S11).

  When it is determined that there is packet transmission in step S11 or packet reception from another terminal in step S6 described above, the idle time observation unit 21 performs, for each priority class, under the packet transmission processing unit 23. The idle time from the time when DIFS has elapsed to the start of packet reception or transmission is measured (step S12). That is, the measured idle time is used as the idle time related to the priority class k of the packet received or transmitted this time, and the subsequent processing is performed.

  That is, in step S13, the above equation (4) is calculated for the class of the packet just received or transmitted, and the offset value (CWoffset) is calculated using the past value of the class and the average number of competing terminals M. .

  In the next step S14, the offset value setting unit 22 compares the calculated and averaged number of competing terminals M with the threshold value of the class k among the threshold values preset for each class, and exceeds the threshold value. The offset value (CWoffsetA) or (CWoffsetB) determined in advance for each class as shown in Table 1 above is selected and set as the offset value CWoffset (k) for the class k). For example, when a priority class 1 packet is received from another terminal in step S6, an average value of the measured idle time and 10 past idle time measurement values of class 1 is calculated, and the calculated value is class 1 If the threshold value is exceeded, the value 31 of CWoffsetB related to class 1 in Table 1 is selected, and CWoffset (3) = 31 is used as an offset value at the time of subsequent transmission. Conversely, if the calculated value is less than or equal to the class 1 threshold, the value 12 of CWoffsetA is selected and CWoffset (3) = 12.

  As described above, in the method of the present embodiment, a priority class is provided according to the content of the transmission packet, and is selected corresponding to CW (k) determined for each class and an idle time that is an indicator of traffic load status. By adding the offset value (CWoffse (k)) and setting the backoff time, the same class can ensure fairness and high throughput regardless of the number of competing terminals. In addition, it is possible to realize QoS that a transmission opportunity is more easily obtained as a higher priority class.

  Although the method of the present embodiment is applied to power line carrier communication, it can also be applied to wireless communication and is not limited to power line carrier communication.

  Further, the idle time may not be observed for each class, and all classes may be observed collectively without distinguishing the classes. In this case, in step S13 described above, the number M of competing terminals is calculated by the inverse function of Equation (4), the number M of competing terminals is averaged, and the offset value (CWoffset is used using the number M of competing terminals. ) Value.

When the number M of competing terminals observed without class distinction exceeds the threshold, CWoffsetA in Table 1 is adopted for all classes, and when the number is below the threshold, CWoffsetB is also adopted for all classes. For example, if the number of class 2 is small and traffic is not congested, but class 0 is also congested, changing class 2 backoff to a larger value will further hinder class 0 communication. The possibility is reduced. That is, despite the fact that class 2 is not crowded (the number of units is small), a large offset value is used and the throughput of class 2 is reduced, but the increase in communication delay of class 0 is prevented, and voice communication is performed. It becomes possible to carry out with high quality.
(Embodiment 2)
In the present embodiment, by calculating the offset value (CWoffset) from the number M of competing terminals, the offset value (CWoffsetA, CWoffsetB) is more finely adjusted than the offset value (CWoffsetA, CWoffsetB) as in the first embodiment. CWoffset) can be set.

  Table 2 shows an example of obtaining an offset value (CWoffset) for each class. In class 3, the offset value (CWoffset) is 0, but in class 2, 4 is added to the number M of competing terminals, Class 1 doubles the number of competing terminals M and adds 8; class 0 also quadruples the number of competing terminals Mn and adds 16 to calculate the offset value (CWoffset) for each class. It has become. As the number of competing terminals M increases, the offset value (CWoffset) becomes larger. For example, for M = 16 units, the offset values (CWoffset) are 0, 20, 40, and 80 for classes 3, 2, 1, and 0, respectively. . In other words, since the value is slightly more than twice the CW, fairness can be secured.

  In Table 2, when calculating the offset value (CWoffset) from the number M of competing terminals, 4, 8 and 16 are added, but this is done by adding a value larger than one higher class CW, This is because higher class traffic is preferentially transmitted.

(Embodiment 3)
In the first embodiment, the traffic load state is observed using the number M of competing terminals calculated based on the observed idle time as a predetermined index. However, the idle time becomes shorter as the number M of competing terminals increases. The method of this embodiment uses the idle time as a predetermined index without calculating the number of competing terminals M, and sets an offset value (CWoffset) based on the idle time. Calculation is unnecessary.

  However, even with the same number of competing terminals, the idle time varies greatly depending on the offset value (CWoffset).

  FIG. 4 is a result of obtaining the idle time by simulation when CWoffset = 0, when CWoffset = 31, and when the number of terminals is 1 to 16, and when they are mixed. CW = 15, and the idle time exceeding 15 is regarded as 15. Even for the same number of competing terminals, the observed idle time varies greatly depending on the value of the offset value (CWoffset). On the other hand, even if a certain idle time is observed, the actual number of competing terminals cannot be specified, so there is a possibility that the optimum offset value (CWoffset) cannot always be set.

However, since the fairness is not so bad when the number of competing terminals M is about eight, and extremely deteriorates when the number is 16 or more, in the method of this embodiment, for example, when idle time = 4 is set as a threshold, When CWoffset = 0 or more, CWoffset = 31 and CWoffset is switched between 0 and two values of about CW × 2.
(Embodiment 4)
As described in the third embodiment, the idle time observed in the case of the same number of competing machines varies greatly depending on the offset value (CWoffset) of the transmitting terminal. For example, when a terminal having a large offset value (CWoffset) transmits, the idle time tends to be long, and when a terminal having a small offset value (CWoffset) transmits, the idle time tends to be short.

  Therefore, in this embodiment, an offset value (CWoffset) is stored in the header of the transmission packet as shown in FIG. 5A, and the observed idle time is weighted by the offset value (CWoffset) in the header.

  For example, when CWoffset = 0 and CWoffset = 31 in the third embodiment are used in two stages, the observation idle time in the case of CWoffset = 0 is doubled, for example. b) (not shown in the case of mixing). In this embodiment, idle time is determined according to the offset value (CWoffset) by weighting the variation width of the idle time by the offset value (CWoffset) by performing idle time finding corresponding to the offset value (CWoffset). Is less than the result of FIG. 4, and the estimation accuracy of the number of competing terminals is improved.

  Thus, according to the present embodiment, it is possible to set the offset value (CWoffset) more in accordance with the actual traffic load situation by providing more than two stages of offset value (CWoffset) classifications. Become.

  As a method of estimating the number M of competing terminals other than observing the idle time as in the third and fourth embodiments, the other terminals A method of observing the number of times the backoff was interrupted because transmission was started earlier, a method of estimating the degree of traffic congestion by counting the number of packets received during a predetermined period, and a period of time It is also effective to observe how many terminals have transmitted by observing the source address in the header of the received packet.

(A) is a block diagram of the terminal used for Embodiment 1, (b) is a detailed block diagram of the transmission control part of a terminal. (A) is a time chart for explaining the observation of idle time in the first embodiment, and (b) is an explanatory diagram of the result of the total throughput obtained by changing the offset value by simulation. 3 is a flowchart for explaining processing operations according to the first embodiment. It is an idle time observation result figure by Embodiment 3. (A) is a format of the MAC header used in Embodiment 4, and (b) is an idle time observation result diagram. (A) is explanatory drawing of the access system of a prior art example, (b) is change explanatory drawing of the total throughput at the time of changing CW. (A) is an explanatory diagram of the result of obtaining the number of successful transmission packets by computer simulation in the conventional access method, and (b) is the number of transmitted packets when the default value of CW is changed in the IEEE 802.11 access method. It is explanatory drawing which shows a change.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 Transmission buffer 2 Transmission control part 21 Idle time observation part 22 Offset value setting part 23 Packet transmission process part 3 Modulation part 4 Transmission AFE part 5 Combiner 6 Reception AFE part 7 Demodulation part 8 Reception control part 9 Carrier detection part

Claims (6)

A multiple access communication method in which each terminal starts transmission after a random backoff time for collision avoidance has elapsed after detecting a free state of communication media by carrier sense,
A class having a plurality of priorities is classified in advance according to the type of transmission packet, and for a predetermined class, an integer from 0 to a contention window value corresponding to the predetermined class is set to one. In this way, a random value is selected, and the value obtained by adding the selected value and the offset value is used as a back-off time, and the random value and the offset value are calculated for each priority class,
Each terminal observes the traffic load status of each class on the communication medium with a predetermined index, and dynamically changes the offset value of the predetermined class according to the traffic load status of the predetermined class. A multiple access communication method characterized by the above. A multiple access communication method in which each terminal starts transmission after a random backoff time for collision avoidance has elapsed after detecting a free state of communication media by carrier sense,
A class having a plurality of priorities is classified in advance according to the type of transmission packet, and for a predetermined class, an integer from 0 to a contention window value corresponding to the predetermined class is set to one. In this way, a random value is selected, and the value obtained by adding the selected value and the offset value is used as a back-off time, and the random value and the offset value are calculated for each priority class,
Each terminal is characterized by observing a traffic load situation on a communication medium with a predetermined index, and dynamically changing the offset values of all classes collectively regardless of the class according to the traffic load situation. Multiple access communication method. 3. The multiple access communication method according to claim 1, wherein the traffic load state is observed using the number of competing terminals estimated based on idle time as a predetermined index. 3. The multiple access communication method according to claim 1, wherein the traffic load state is observed using idle time as a predetermined index. 3. The multiple access communication method according to claim 1, wherein the traffic load state is observed using the number of times the back-off is interrupted as a predetermined index. Multiple access communication how according to claim 1, wherein observing the traffic load situation, as a predetermined index number of terminals has transmitted within a predetermined time.

Images