JP2008227737A - Wireless unit, and wireless communication network having the same - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a wireless unit improved in communication efficiency of a cognitive wireless network. <P>SOLUTION: A terminal unit includes a wireless system conforming to IEEE 802.11j, and a wireless system conforming to IEEE 802.11g. The link cost of aggregated links having a wireless link ML1 conforming to IEEE 802.11j aggregated with a wireless link ML2 conforming to IEEE 802.11g becomes a function of convex downward relative to a packet distribution ratio to the wireless link ML1 (refer to curve k1). The terminal unit determines the packet distribution ratio so as to obtain the minimum link cost of the aggregated links by the descending method (refer to ARW1-ARW4), and transmits packets to the destination by distributing the packets to the two wireless links ML1, ML2, according to the determined packet distribution ratio. <P>COPYRIGHT: (C)2008,JPO&INPIT

Description

  The present invention relates to a wireless device and a wireless communication network including the wireless device, and more particularly to a wireless device that performs wireless communication by selecting a wireless system suitable for a wireless communication environment from a plurality of different wireless systems, and a wireless communication network including the wireless device. It is about.

  In recent years, the use of various wireless systems such as mobile phones, PHS (Personal Handyphone System), IEEE802 wireless LAN (Local Area Network), and Bluetooth has been expanded. In ubiquitous communication, a sensor network is configured, and the use of short-range wireless systems such as ZigBee is also expected.

  Such wireless systems are rapidly expanding and diversifying their use, becoming a wireless communication environment in which various wireless systems having different frequency bands and different communication methods are mixed, and various applications are expected to be used. ing.

  On the other hand, since radio resources are limited, there is a concern that radio resources will be depleted as the use of radio systems expands and diversifies. As a technique for solving this problem, a cognitive radio technique has been proposed (Non-Patent Document 1).

  The cognitive radio technology uses a plurality of radio systems in a network of a base station equipped with a plurality of different radio systems and a terminal device equipped with a plurality of different radio systems in accordance with radio communication conditions and user requests. Is a technique that uses them properly or uses them simultaneously.

On the other hand, there has been proposed a control method for improving the communication throughput between nodes equipped with a plurality of wireless systems having the same performance and the same control procedure (Non-Patent Document 2).
Harada, “Basic study on communication system using cognitive radio”, IEICE Technical Report, SR2005-17, pp. 117-124, 2005. Alex C. Snoeren, "Adaptive Inverse Multiplexing for Wide-Area-Wireless Networks", Proceedings of IEEE GlobCom, 1999. J. Little, “A Proof of the Quenching Formula L = λW”, Opre Res J. 18: 172-174, 1961. ST. Sheu and J. Chen, “A Novel Delay-Oriented Shortest Path Routing Protocol for Mobile Ad Hoc Networks”, in Proc. IEEE ICC '01, 2001. RL Rivert, “Network Control by Bayessian Broadcast”, Report MIT / LCS / TM-285, 1985. WA Rosenkrantz, “Some Theorems on the Instability of the Exponential Back-off Protocol”, Performance '84, Elservier, pp. 199-205, 1984. IEEEStd 802.16-2004, “Part 16: Air Interface for Fixed Broadband Wireless Access Systems”, 2004.

  However, since the conventional control method is a communication control method between nodes equipped with a plurality of wireless systems having the same performance and the same control procedure, each terminal apparatus can be controlled by a different communication method using the conventional control method. There is a problem that it is difficult to improve the communication efficiency of a cognitive radio network that performs radio communication.

  Accordingly, the present invention has been made to solve such a problem, and an object of the present invention is to provide a radio apparatus capable of improving the communication efficiency of a cognitive radio network.

  Another object of the present invention is to provide a wireless communication network including a wireless device capable of improving the communication efficiency of a cognitive wireless network.

  According to this invention, the wireless device includes a plurality of wireless modules and communication control means. The plurality of wireless modules perform wireless communication using a plurality of different wireless systems. The communication control means performs wireless communication in the plurality of wireless modules so that the communication cost indicating the entire communication load in the plurality of wireless links established between the plurality of wireless modules with the transmission destination adjacent to the wireless device is minimized. To control.

  Preferably, the communication control means distributes the packet to the plurality of wireless modules so that the communication cost is minimized. Each of the plurality of wireless modules transmits the distributed packet.

  Preferably, the communication control means calculates a link cost indicating a communication load in one radio link for a plurality of radio links, selects a maximum link cost and a minimum link cost from the calculated plurality of link costs, Packets are distributed to a plurality of wireless modules by repeatedly performing packet distribution processing that distributes packets distributed to wireless links having a link cost to wireless links having the minimum link cost.

  Preferably, in the communication control means, the first radio link having the maximum link cost in the nth (n is an integer of 2 or more) packet distribution process has the maximum link cost in the (n-1) th packet distribution process. When the second wireless link is the same as the second wireless link, the same number of packets as the number in the (n-1) th packet distribution process are distributed to the wireless link having the minimum link cost, and the first wireless link is the second wireless link. When they are different, the number of packets smaller than the number in the (n-1) th packet distribution process is distributed to the radio link having the minimum link cost.

  Preferably, when the first wireless link is different from the second wireless link, the communication control unit gradually decreases the number of packets distributed to the wireless link having the minimum link cost as the number of packet distribution processes increases.

  Preferably, the plurality of wireless modules include first and second wireless modules. The first wireless module performs wireless communication according to the first protocol. The second wireless module performs wireless communication according to a second protocol different from the first protocol.

  According to the present invention, the wireless communication network includes a base station and a terminal device. The base station performs wireless communication using a plurality of different wireless systems. The terminal device performs wireless communication with the base station using a plurality of wireless systems. And each of a base station and a terminal device consists of a radio | wireless apparatus of any one of Claims 1-5.

  Preferably, the plurality of radio systems include first and second radio systems. The first wireless system has a first communication range. The second wireless system has a second communication range that is wider than the first communication range. The first and second communication ranges are arranged substantially concentrically. The base station is disposed within the first communication range, the first terminal device disposed within the first communication range, and is outside the first communication range and within the second communication range. Wireless communication is performed with the second terminal device arranged in a certain range.

  Preferably, the plurality of radio systems include first and second radio systems. The first wireless system has a first communication range. The second wireless system has a second communication range that is wider than the first communication range. The base station includes first and second base stations. The first base station includes a first terminal device disposed within the first communication range, and a second terminal disposed within a range outside the first communication range and within the second communication range. Wireless communication with other terminal devices. The second base station performs radio communication only with the first terminal device.

  According to the present invention, wireless communication is performed so that the overall communication load on the plurality of wireless links established between the plurality of wireless modules and the transmission destination is minimized.

  Therefore, according to the present invention, the communication efficiency in the cognitive radio network can be improved.

  Embodiments of the present invention will be described in detail with reference to the drawings. In the drawings, the same or corresponding parts are denoted by the same reference numerals and description thereof will not be repeated.

[Embodiment 1]
FIG. 1 is a schematic diagram of a radio communication network according to Embodiment 1 of the present invention. A wireless communication network 100 according to Embodiment 1 of the present invention includes terminal devices 1 to 4 and a base station 10.

  The terminal apparatuses 1 and 2 and the base station 10 exist in the IEEE 802.11j cell 30, and the terminal apparatuses 3 and 4 exist outside the cell 30 and in an area within the IEEE 802.11g cell 40. To do. The communication range of the cell 40 is wider than that of the cell 30, and the cells 30 and 40 are arranged concentrically. For example, the radius of the cell 30 is 50 m, and the radius of the cell 40 is 100 m.

  The terminal devices 1 to 4 and the base station 10 perform wireless communication with each other by IEEE802.11j and IEEE802.11g.

  The base station 10 is connected to a network 20 such as the Internet by a wired cable 31. The terminal devices 21 to 24 include a personal computer and a server, and are connected to the network 20.

  As will be described later, each of the terminal apparatuses 1 to 4 is equipped with a plurality of radio systems, and accesses the base station 10 by using the equipped radio systems appropriately or simultaneously.

  Each of the terminal devices 1 to 4 and the base station 10 includes a wireless link ML1 that performs wireless communication using IEEE802.11j and a wireless link ML2 that performs wireless communication using IEEE802.11g. Then, each of the terminal devices 1 to 4 and the base station 10 distributes packets to the radio links ML1 and ML2 so that the communication cost between itself and the transmission destination is minimized by a method described later.

  FIG. 2 is a schematic block diagram showing the configuration of the terminal device 1 shown in FIG. The terminal device 1 includes an antenna 11, radio modules 12 and 13, a switching unit 14, a search module 15, an application processor 16, and a bus BS.

  The switching means 14, the search module 15 and the application processor 16 are connected to each other by a bus BS. The radio modules 12 and 13 are connected to the switching means 14.

  The antenna 11 receives data from the base station 10 via the wireless communication space, outputs the received data to at least one of the wireless modules 12 and 13, and data from at least one of the wireless modules 12 and 13. Is transmitted to the base station 10 via the wireless communication space.

  Each of the wireless modules 12 and 13 is equipped with a different wireless system. More specifically, the wireless module 12 is equipped with a wireless system based on IEEE 802.11j, and the wireless module 13 is equipped with a wireless system based on IEEE 802.11g.

  The wireless module 12 performs wireless communication with the base station 10 via the antenna 11 using a wireless system based on IEEE 802.11j. In this case, the radio module 12 uses a 2.4 GHz frequency, a transmission rate of 54 Mbps, an OFDM (Orthogonal Frequency Division Multiplexing) modulation scheme, and a CSMA / CA (Carrier Sense Multiple Access Collision Access Access scheme).

  The wireless module 13 performs wireless communication with the base station 10 via the antenna 11 using a wireless system based on IEEE 802.11g. In this case, the wireless module 13 uses a frequency of 2.4 GHz, a transmission rate of 54 Mbps, an OFDM modulation scheme, and a CSMA / CA access scheme.

  In this way, the wireless modules 12 and 13 are equipped with different wireless systems, and transmit the packets distributed by the switching means 14 singly or simultaneously using the equipped wireless systems.

  The switching means 14 receives the packet distribution rate from the search module 15 via the bus BS, and distributes the packet to the wireless modules 12 and 13 according to the received packet distribution rate.

  The search module 15 uses a method to be described later so that the overall communication cost of the radio links ML1 and ML2 of the terminal device 1 (= the cost of the aggregate link that aggregates the radio links ML1 and ML2) is minimized. The packet distribution rate of the packet distributed to ML2 is determined, and the determined packet distribution rate is output to switching means 14 via bus BS.

  The application processor 16 generates a packet and outputs the generated packet to the switching unit 14 via the bus BS.

  Each of terminal apparatuses 2 to 4 and base station 10 shown in FIG. 1 has the same configuration as that of terminal apparatus 1 shown in FIG.

  In the wireless communication network 100 shown in FIG. 1, the terminal devices 1 to 4 and the base station 10 transmit and receive packets to and from each other by the wireless modules 12 and 13. And in order to maintain the throughput of the radio | wireless communication between the terminal devices 1-4 and the base station 10, and to reduce the delay time of the radio | wireless communication, each of the terminal devices 1-4 and the base station 10 It is effective to distribute packets to the radio links ML1 and ML2 so that the overall communication cost of the two radio links ML1 and ML2 existing between them is minimized. The effectiveness will be described.

[Mathematical model]
(A) Cost of Single Link FIG. 3 is a conceptual diagram for explaining a single link and an aggregated link. The average packet arrival rate at the terminal device i is F _i (packets / sec), and the communication capacity of the link between the terminal device i and the base station 10 (the reciprocal of the average packet processing time, that is, the packet processing rate) is C Assuming _i (packets / sec), the cost d _i in this link is expressed by the following equation based on the M / M / 1 queuing system (see FIG. 3A).

Equation (1) is the average number of packets waiting in the link (including the packet being processed), and is a value obtained by multiplying the average delay time 1 / (C _i −F _i ) by the packet arrival rate F _i . Therefore, in the present invention, the cost d _i means “the number of standby traffic in the link”. Here, the average delay time is an average of time until a packet arrives, waits in a queue, and transmission of the packet is completed. That is, the average delay time is an average of the sum of the waiting time and the processing time in the queue of each packet.

Thus, when the cost d _i is high, the link has more packets and the load is high. On the other hand, when the cost d _i is low, the link has fewer packets and its load is low.

(B) Cost of aggregated link An aggregated link consists of more links than one link. Then, the arrived packet is distributed to a plurality of links of the aggregated link (see (b) of FIG. 3). Therefore, the cost D _{i of the} aggregated link is the sum of the costs of each link and is expressed by the following equation.

The packet arrival rate F _i in the aggregated link is expressed by the following equation.

In Equations (2) and (3), R _i is the number of wireless systems attached to the terminal device i, and C _i ^r is the link capacity of the wireless system r attached to the terminal device i. , F _i ^r is the packet arrival rate of the wireless system r attached to the terminal device i.

(C) Link capacity In a macro period, it is assumed that the average usage rate of radio resources competing in each link is equal. Therefore, each link capacity is estimated as a capacity when the maximum transmission rate of the wireless system is equally distributed according to the number of terminal devices that can communicate with the base station. Each link capacity is independent of the packet arrival rate. That is, each link capacity is limited by the M / M / 1 queuing system and is expressed by the following equation.

In Equation (4), ^Br is the maximum transmission rate, and _Nr is the number of terminal devices that can communicate with the base station using the radio system r.

(D) Network Cost The total link cost of the network (referred to as “network cost G”) is the number of packets in the network indicating the load on the network. That is, the network cost G is expressed as the following equation as the sum of the costs of the aggregated links in the network.

  In Expression (5), N is the number of terminal devices in the network.

[Dependence of cost on packet distribution ratio]
The characteristics of the packet distribution rate using the dependency of the cost on the packet distribution rate in the following configuration will be described.

(A) A base station and a terminal device are each configured with one IEEE802.11j system and one
One IEEE 802.11g system is installed.

(B) The cell of the base station is composed of an IEEE802.11j cell and an IEEE802.11g cell arranged concentrically.

(C) The terminal device is randomly arranged in the cell of the base station. Based on the cell radius of each wireless system, a quarter of the terminal devices are IEEE 802.1
Present in 1j cell.

(D) The packet arrival rate of each terminal device existing in both the IEEE802.11j cell and the IEEE802.11g cell is F _{g ∩j} .
(E) The packet arrival rate of each terminal apparatus existing only in the IEEE _802.11g cell is F _{g / j} .

  (F) The number of terminal devices in the network is N.

(G) IEEE 802.11j and IEEE802.11g maximum transmission rate of, respectively, a ^{B g} and ^{B j.}

Note that _{/ j} in F _{g ∩ / j} indicates that the terminal device does not exist in the IEEE802.11j cell.

  In the configurations shown in (A) to (G) above, the network cost is expressed by the following equations (6) to (8).

In Equation (6), F ^g _g∩j indicates the packet arrival rate to the radio link by IEEE802.11g in a terminal device existing in both the IEEE802.11j cell and the IEEE802.11g cell. F ^j _g∩j indicates the packet arrival rate to the radio link by IEEE802.11j in the terminal apparatus existing in both the IEEE802.11j cell and the IEEE802.11g cell. Therefore, F _g∩j = F ^g _g∩j + F ^j _g∩j .

FIG. 4 is a diagram showing the relationship between the packet distribution ratio to the wireless link and the network cost according to IEEE802.11j. In this case, the terminal device exists in both the IEEE _802.11j cell and the IEEE _802.11g cell, the terminal device number N is 40, F ^g _g∩j is 1 packet / sec, F ^j _g∩j is 5 packets / sec, each of B ^g and B ^j is 54 Mbps, and the average packet size is 1 Mbits.

  In FIG. 4, the vertical axis represents the network cost, and the horizontal axis represents the packet distribution ratio to the wireless link according to IEEE802.11j.

The network cost decreases as F ^j _{g ∩j} increases, and F ^j _{g ∩j} is about 4.5 (corresponding to a packet distribution rate to the radio link by IEEE _802.11j of 0.9). Beyond, it begins to increase. Therefore, the network cost has an optimal solution for the packet distribution rate.

In order to verify this, the cost of the radio link d ^g _{g∩ / j} by the IEEE _{802.11g of} the terminal device that exists only in the IEEE _802.11g cell, the IEEE _802.11g cell, and the IEEE _802.11j cell exists. The wireless link cost d ^g _{g∩j according} to IEEE802.11g of the terminal device and the wireless link cost d ^j _g∩j of the terminal device existing in the IEEE802.11j cell ).

  Then, using the first derivative and the second derivative of each link cost based on the packet arrival rate of the packet distributed to each wireless link, the dependency of each link cost on the packet arrival rate is expressed by the following equation (9 ) To (15).

Equations (10) and (11) are convex monotonically increasing functions with respect to the packet arrival rate of the wireless link by IEEE802.11g under the condition that the link cost d ^g _{g∩ / j} is F _{g∩ / j} <C ^g. Indicates that there is.

Further, Expression (12) and Expression (13) indicate that a convex monotonically increasing function with respect to the packet arrival rate of the wireless link according to IEEE802.11g under the condition that the link cost d ^{g g} _∩j is F ^{g g} _∩j <C ^g Indicates that

Further, Expression (14) and Expression (15) are expressed as a convex monotone increasing function with respect to the packet arrival rate of the wireless link according to IEEE802.11j under the condition that the link cost d ^j _{g ∩j} is F ^j _{g ∩j} <C ^j. Indicates that

Further, since F _g∩j = F ^g _g∩j + F ^j _g∩j , dF ^g _g∩j / dF ^j _g∩j <0 and d ² (F ^g _g∩j ) / d (F ^j _g∩j ) ² = 0.

Therefore, first derivative and second derivative by _{^F j} ^g∩j cost _{^d g} ^g∩j are respectively expressed by the following equation (16) and (17).

As a result, the link cost d ^g _g∩j becomes a convex monotone decreasing function with respect to the packet distribution rate to the wireless link by IEEE802.11j, and the network cost is the cost of the wireless link by IEEE802.11j which is a convex monotonically increasing function. And the cost of the wireless link according to IEEE 802.11g, which is a convex monotonously decreasing function. That is, the network cost has a completely optimal solution for the packet distribution rate.

  FIG. 5 is a diagram illustrating the relationship between the cost and the packet distribution rate. In FIG. 5, the vertical axis represents the cost, and the horizontal axis represents the packet distribution ratio to the radio link according to IEEE802.11j. A curve k1 indicates the cost of the radio link according to IEEE802.11g, a curve k2 indicates the cost of the radio link according to IEEE802.11j, and a curve k3 indicates the network cost G.

Note that the cost of the wireless link according to IEEE802.11g is expressed by 3Nd ^{g g} _{/ j} / 4 + Nd ^{g g} _j / 4 from Equation (9), and the cost of the wireless link according to IEEE _802.11j is from Equation (9). Nd ^j _gＮj / 4.

  As is apparent from FIG. 5, the network cost G has a minimum value (= optimum solution) with respect to the packet distribution ratio to the radio link according to IEEE802.11j.

The first and second derivatives of the equation (1) according to the packet arrival rate F _i are expressed by the following equations (18) and (19), respectively.

Therefore, the link cost itself is a convex monotonically increasing function with respect to the packet distribution rate under the condition of F _i <C _i . In addition, the cost of the aggregate link, which is the sum of the cost of the wireless link based on IEEE 802.11j and the cost of the wireless link based on IEEE 802.11g, is a convex function with respect to the packet distribution rate, like the network cost.

  As a result, the cost of each aggregated link gives an optimal solution for the packet distribution rate, and the optimal solution for each aggregated link constitutes the optimal solution for the network cost. That is, the cost of the aggregate link between each of the terminal devices 1 to 4 in FIG. 1 and the base station 10 (= the sum of the cost of the radio link according to IEEE802.11j and the cost of the radio link according to IEEE802.11g) is optimized ( = Minimum), the network cost G, which is the cost of the entire wireless communication network 100, is optimized.

[Delay time and throughput]
FIG. 6 is a diagram illustrating the relationship between the sum of delay times and throughput. In FIG. 6, the vertical axis represents the sum of delay times, and the horizontal axis represents throughput. Curves k4 to k7 indicate the relationship between the sum of the delay times and the throughput when the packet distribution ratios to the wireless links according to IEEE 802.11j are 0.5, 0.7, 0.9, and 1.0, respectively. Indicates.

  The sum of delay times is the sum of the time intervals from the arrival of packets at the terminal device to the arrival of packets at the base station in unit time, and the throughput is the number of packets that arrive at the unit time.

When the packet distribution rate is 0.9, the delay time decreases and the throughput increases with respect to other packet distribution rates. A packet distribution rate of 0.9 gives a solution close to the optimal solution for network cost. That is, the optimal solution for network cost corresponds to the minimum delay time and maximum throughput,
Therefore, by optimizing (minimizing) the network cost, the minimum delay time and the maximum throughput can be obtained.

[Packet distribution characteristics of links according to IEEE 802.11]
(Link cost)
Since transmission and reception of packets depend on each other in the link according to IEEE 802.11, and the packet service time of the link according to IEEE 802.11 is an exponential distribution, the above-described equations (1) and (2) are changed to IEEE802. It cannot be applied to the link by 11.

However, since the Little theorem (Non-Patent Document 3) is established in a link having an exponential packet arrival interval, the link cost d _i and the aggregate link cost D _i according to IEEE 802.11 are respectively expressed by the following equations: It is represented by (20) and formula (21).

In Equation (20), T _i is an average delay time in the terminal device i, and in Equation (21), T ^r _i is a delay time of the wireless system r attached to the terminal device i.

(Dependence on packet distribution of cost)
As described above, the link cost includes a delay time and a packet arrival rate that is a packet distribution rate. The delay time consists of a waiting time in the queue and a packet service time.

  Packet distribution characteristics are shown for packet service time, which is the time consumed to successfully transmit one packet.

  FIG. 7 is a diagram illustrating a model for analyzing the packet service time. Packet service time is divided into e-slots. One e-slot is defined as a unit of event according to IEEE 802.11, and is different from the slot time of IEEE 802.11.

  The packet service time includes a back-off period and a transmission trial period. The back-off period is a waiting time until transmission is attempted, and within this period, an idle e-slot, a congested e-slot, and a collision e-slot occur as events. An idle e-slot means waiting for transmission, a congested e-slot means that any terminal device is transmitting, and a collision e-slot means a packet collision.

  The transmission trial period is a period spent for a transmission attempt of a packet after the back-off period, and consists of one e-slot. During this period, successful e-slots or failed e-slots occur as events. A successful e-slot means successful transmission of a packet, and a failed e-slot means transmission due to a collision.

If a failed e-slot occurs, the next back-off period for packet retransmission starts. If the transmission attempt rate of a packet in one e-slot in the cell c is G _i , G _i includes a retransmission packet. Then, packets transmitted in the cell c are, because they are located well randomly by many terminals in the cell c, G _i is assumed to have a Poisson distribution.

G _i is independent for each e-slot, and F ^c _i and F _i are the packet arrival rate in cell c and the packet arrival rate in terminal device i, respectively. Then, when j is the number of terminal devices in the cell c, F _i is F ^c _i / j. Each packet arrival process is a Poisson process.

The dependency of the packet service time S _{i on} the terminal device i on the packet arrival rate F ^c _i will be described. First, the dependency of the packet service time S _{i on} the transmission attempt rate G _i is shown, and then the dependency of the transmission attempt rate G _{i on} the packet arrival rate F ^c _i is shown. Finally, these dependencies are shown. , The dependence of the packet service time S _{i on} the packet arrival rate F ^c _i is shown.

  The back-off period is based on a binary exponent back-off process and corresponds to a time until the back-off counter becomes “0”. And the back-off counter is decremented by one idle e-slot and not decremented by a congested e-slot or a collision e-slot. That is, the backoff counter is not decremented until an idle e-slot occurs (see FIG. 7).

When the probability that a congested e-slot or collision e-slot will occur in the backoff period is P _bc , the probability that a congested e-slot or collision e-slot will occur m times in the backoff period is the geographical When the distribution is used, P ^m _bc (1-P _bc ) is obtained.

Also, assuming that the time spent in the congested e-slot or the collision e-slot and the time spent in the idle e-slot are T _bc and T _id , respectively, the time consumed because the back-off counter is decremented by “1”. M (z), which is a probability generation function (PGF), is expressed by the following equations (22) to (28).

Here, T _s and T _c are time standardized by the slot time of IEEE802.11 and time spent for packet transmission and collision, respectively. Similarly, T _id , which is the time spent for the same idle e-slot as one slot time, is “1” when normalized by the IEEE 802.11 slot time.

P _bc and P _s are the probability of more than one transmission attempt and the probability of one transmission attempt, respectively. Using M (z), the average time / M spent in the back-off counter decremented by “1” is expressed by the following equations (29) and (30).

Here, N _bc is an average number of times that a congested e-slot or a collision e-slot continuously occurs in the back-off period.

The back-off period is repeated until the packet transmission is successful. When a failed e-slot (ie, packet collision) occurs n times consecutively before a successful e-slot (ie, successful packet transmission) within the transmission attempt period, the backoff period is repeated n + 1 times. The average time K _n spent in the (n + 1) total back-off periods is expressed by the following equations (31) and (32).

Here, W is an IEEE 802.11 contention window size. Failure before a successful e-slot, where P _xc is the probability of a failed e-slot occurring within the transmission attempt period and P _tx is the probability that one or more packets will be transmitted within the transmission attempt period The probability that an e-slot occurs n times consecutively is P _tx P ⁿ _xc (1−P _xc ).

  As a result, the average time / B spent in the entire back-off period until the packet transmission is successful is expressed by the following equations (33) to (35).

  The transmission trial period starts when the back-off counter reaches “0” (= when the back-off period ends), and is repeated until the packet transmission is successfully performed in the same manner as the back-off period.

  Under the condition that one or more packets are transmitted within the transmission attempt period, A (z), which is a probability generation function (PGF) of the time spent for the entire transmission attempt period, is expressed by the following equation. The

  When A (z) is used, the average time / A spent in the total transmission trial period is expressed by the following equations (37) and (38).

Here, N _xc is the average number of packet collisions within the packet transmission attempt period.

When / A and / B are used, the packet service time S _i is expressed by the following equation.

Furthermore, the first and second derivatives of the packet service time S _{i according to} the transmission trial rate G _i are expressed by the following equations (40) and (41), respectively.

Accordingly, the packet service time S _i is a convex monotonically increasing function of the packet transmission trial rate G _i .

  FIG. 8 is a diagram illustrating the relationship between the packet service time and the packet transmission trial rate. In FIG. 8, the vertical axis represents packet service time, and the horizontal axis represents packet transmission trial rate.

  Curves k8 to k11 indicate the relationship between the packet service time and the packet transmission trial rate when the packet size is 0.1 Kbyte, 0.5 Kbyte, 1 Kbyte, and 2 Kbyte, respectively.

As is apparent from the curves k8 to k11 shown in FIG. 8, the packet service time S _i is a convex monotonically increasing function of the packet transmission trial rate G _i .

In order to obtain the dependence of the packet service time S _{i on} the packet arrival rate F ^c _i , the dependence of the packet transmission trial rate G _{i on} the packet arrival rate F ^c _i was examined. The number of e-slots N _e within the packet service time is introduced by the same method as the packet service time S _i and is expressed by the following equation.

When N _e is used, the number of packet transmission attempts within the packet service time is expressed as N _e G _i .

On the other hand, the number of transmission attempts per packet is (N _xc +1), and the arrival number of packets within the packet service time is F ^c _i S _i . Here, F ^c _i is the packet arrival rate within the slot time.

As a result, the number of packet transmission attempts within the packet service time is expressed as F ^c _i S _i (N _xc +1). From the results described above, F ^c _i is expressed as follows.

Further, the first and second derivatives of the packet arrival rate F ^c _{i by} the transmission trial rate G _i are expressed by the following equations (44) and (45), respectively.

Accordingly, the packet arrival rate F ^c _i is a convex monotonically increasing function of the transmission trial rate G _i .

  FIG. 9 is a diagram illustrating the relationship between the packet arrival rate and the packet transmission trial rate. In FIG. 9, the vertical axis represents the packet arrival rate, and the horizontal axis represents the packet transmission trial rate.

  Curves k12 to k15 indicate the relationship between the packet arrival rate and the packet transmission trial rate when the packet size is 0.1 Kbyte, 0.5 Kbyte, 1 Kbyte, and 2 Kbyte, respectively.

As is apparent from the curves k12 to k15 shown in FIG. 9, the packet arrival rate F ^c _i is a convex monotonically increasing function of the packet transmission trial rate G _i .

Based on equation (44) and equation (45), the first and second derivatives of the packet transmission trial rate G _{i by} the packet arrival rate F ^c _i are respectively expressed by the following equations (46) and ( 47).

Also, based on the model assumption, F _i is F ^c _i / j, and the first and second derivatives of the packet transmission trial rate G _{i according to} the packet arrival rate F _i are respectively It represents with Formula (48) and Formula (49).

As a result, first derivative and second derivative with the packet arrival rate F _i of the service time S _i of the packet, respectively, are expressed by the following equation (50) and (51).

From the results described above, the packet service time S _i is a convex monotonically increasing function of the packet arrival rate F _i .

If the dependency of the packet service time S _{i on} the packet arrival rate F _i is given, the dependency of the waiting time W _{i on} the queue on the packet arrival rate F _i can be examined.

The waiting time W _i is defined as the average time that packets arriving at the terminal device i stay in the queue. The number N _{Q of} packets remaining in the queue (not including currently serviced packets) is F _i × W _i using the Little theorem (Non-patent Document 3). The waiting time W _i is comprised of the time spent processing N _Q packets and the total remaining service time for processing the packets when each packet arrives. As a result, the waiting time W _i is expressed by the following equation.

Here, R indicates the total remaining service time for processing a packet when each packet arrives. The total remaining service time for processing a packet when one packet arrives is / S ² _i / 2S _i . Here, / S ² _i is the second moment of S _i .

If the above equation (52) is used, the standby time W _i is expressed by the following equation.

In order to obtain the dependence of the waiting time W _{i on} the packet arrival rate F _i , the dependence of / S ² _i on S _i was examined. Using M (z), the variance V (M) of the time consumed for the back-off counter decreasing by “1” is expressed by the following equation.

  Applying equation (54) to equations (31) and (33) instead of / M, the variance V (M) of the time consumed for the total backoff period is expressed by the following equation: .

As a result, the second moment / B ² of the time consumed for the total back-off period is expressed by the following equation.

  Similarly, using A (z), the variance V (A) of the time consumed for the total transmission trial period is expressed by the following equation.

Then, the second moment / A ² of the time consumed for the total transmission trial period is expressed by the following equation.

Using Equation (56) and Equation (58), / S ² _i is expressed by the following equation.

Furthermore, first derivative and second derivative by transmission attempts rate G _i of the second moment / S ² that is the time consumed for transmitting trial period the total, respectively, the following equation (60) and ( 61).

  FIG. 10 is a diagram illustrating the relationship between the second moment of the packet service time and the packet transmission trial rate. In FIG. 10, the vertical axis represents the second moment of the packet service time, and the horizontal axis represents the packet transmission trial rate. Curves k16 to k19 indicate the relationship between the second moment of the packet service time and the packet transmission trial rate when the packet size is 0.1 Kbyte, 0.5 Kbyte, 1 Kbyte, and 2 Kbyte, respectively.

  Curves k16 to k19 shown in FIG. 10 show the same characteristics as those shown by Expression (59).

Using equations (46) through (49), the first and second derivatives of the second moment of time consumed for the total transmission trial period / the packet arrival rate F _i of S ² _i are: Are represented by the following equations (62) and (63), respectively.

From the results described above, the first and second derivatives of the waiting time W _{i according to} the packet arrival rate F _i are expressed by the following equations (64) and (65), respectively.

F _i S _i indicates the link usage factor of the terminal device i and does not exceed “1”. Therefore, the waiting time W _{i is} also a convex monotonically increasing function of the packet arrival rate F _i .

Finally, the dependency of the average delay time T _i and the link cost d _{i on} the packet arrival rate F _i will be described.

Since the average delay time T _i is the sum of the waiting time W _i and the packet service time S _i , the average delay time T _i is also a convex monotonically increasing function of the packet arrival rate F _i . Then, the first and second derivatives of the link cost d _{i by} the packet arrival rate F _i are expressed by the following equations (66) and (67), respectively.

  Equations (66) and (67) show that the cost of the wireless link according to IEEE 802.11 is also a convex monotonically increasing function of the packet distribution rate. As a result, the packet distribution characteristic of the wireless link according to IEEE 802.11 is the same as the packet distribution characteristic of M / M / 1, and the network cost of the wireless network according to IEEE 802.11 gives an optimal solution for the packet distribution rate. .

[Traffic control method]
(Link cost)
A traffic control method will be described based on the packet distribution characteristics described above. The link cost d _i is calculated by the Little's theorem using the packet arrival rate F _i and the average delay time T _i measured in the period [t, t + Δt] as follows:

Note that the average delay time T _i (t, t + Δt) is an average interval in a period [t, t + Δt] from when the packet arrives at the terminal device until it receives an acknowledgment (ACK: Acknowledge).

Depending on Δt, there is a probability that T _i (t, t + Δt) will not respond to packets that arrive in the period [t, t + Δt]. Therefore, Δt is assumed to be a sufficiently long period with respect to T _i (Δt).

(Search for optimal solution in network)
As described above, the network cost is a convex function of the packet distribution rate, and gives an optimal solution for the packet distribution rate. Therefore, a method for searching for an optimal solution of the network cost with respect to the packet distribution rate will be described.

  The aggregated link cost is calculated based on the link cost of the wireless link based on IEEE802.11g, which is a convex monotonously decreasing function, and the packet distribution rate on the wireless link based on IEEE802.11j. The aggregate link cost is also a convex function of the packet distribution rate because it is the sum of the link cost of the wireless link according to IEEE802.11j, which is a convex monotonically increasing function.

  Therefore, each aggregate link cost also has an optimal solution for the packet distribution rate.

  From the results described above, searching for the optimal solution for the aggregate link cost in each terminal device results in searching for the optimal solution for the network cost.

The search for the optimal solution of the convex function is performed using the descent method. The descent method starts with an initial value D (p ^r (0)), determines α so as to satisfy the inequality sign of the following equation, and finds an optimal solution by repeating the determination of α.

Here, p ^r (k) indicates the number of packets distributed to the link using the radio system r in the terminal device, α indicates the step size, and D (p ^r (t)) indicates the terminal The aggregated link cost when the number of packets distributed to the link using the wireless system r in the apparatus is p ^r (t) is shown.

  The descent method is used for searching for an optimum solution of the aggregate link cost in each terminal apparatus by the following method.

MTH1) For the first time only, packets are rounded to each link of the aggregated link.
Evenly distributed using bins. Therefore, the packet distribution rate of each link
Are equal at first.

MTH2) At the start of one period, the packet arrival rate F _i and the average delay time
T _i is measured.

MTH3) At the end of one period, the measured packet arrival rate F _i and the average
Using the delay time T _i , the link cost d _i and the aggregate link cost D _i
Is calculated.

MTH4) At the end of the initial period (k = 0), the highest cost among the aggregated links
The link with is selected and the lowest cost from that selected link
At the end of MTH5) normal cycle packet is distributed in the initial number p ^r to link with ^{(0) (k ≠ 0)} , the aggregate link cost is reduced
The link with the highest cost is selected from the aggregated link costs.
Previously moved from the selected link to the link with the lowest cost
Packet number ^p r (k-1) the same number as the packet ^p r a (k)
Distribute (p (k) = p (k−1)).

MTH6) At the end of the normal period (k ≠ 0), the aggregate link cost has increased.
Select the link with the highest cost other than the previously selected link
Smaller than the number of previously moved packets p ^r (k−1)
Packet p ^r (k) with the lowest cost from the selected link
Distribute to the link (p (k) = α · p (k−1), 0 <α <1).

  MTH7) The above-described process is repeatedly executed for each period.

  FIG. 11 is a diagram for explaining a packet distribution method. In FIG. 11, the vertical axis represents the link cost, and the horizontal axis represents the packet distribution ratio to the radio link according to IEEE 802.11j. The black squares indicate the relationship between the packet distribution rate to the wireless link ML1 by IEEE802.11j and the link cost of the wireless link ML1, and the black triangles indicate the packet distribution rate to the wireless link ML1 by IEEE802.11j and IEEE802.11. The relationship with the link cost of the radio link ML2 by 11g is shown.

  As described above, there is an optimal solution for the network cost in the packet distribution rate, and the network cost is a downward convex function with respect to the packet distribution rate (see curve k20 in FIG. 11).

  Accordingly, the minimum solution is obtained by repeatedly applying the equation (69) to the high-cost link using the above-described descent method.

  Some of the packets (determined by α · p (t)) are moved from the high cost link to the low cost link (see arrow ARW1 in FIG. 11). Then, after moving some packets, if the high-cost link is the same as the previous link, the same amount of packets (p (t) = p (t−1)) as the previous link is transferred to the low-cost link. Moving. On the other hand, when a high-cost link is switched, a lower amount of packets (p (k) = α · p (k−1), 0 <α <1) than the previous time from a new high-cost link has a low cost. Move to the link (see arrow ARW2 in FIG. 11). Then, the above-described operation is repeated to gradually approach the vicinity of the minimum solution while reducing the packet distribution amount (see ARW3 and ARW4 in FIG. 11). This provides an optimal solution for the aggregate link cost.

  Accordingly, in each of the terminal devices 1 to 4 and the base station 10 illustrated in FIG. 1, the search module 15 has a cost of an aggregate link including the radio link ML1 included in the radio module 12 and the radio link ML2 included in the radio module 13. The optimum solution (= packet distribution rate) when it becomes the minimum is searched by the above-described descent method, and the switching means 14 is controlled to distribute the packets to the wireless modules 12 and 13 with the searched packet distribution rate.

  FIG. 12 is a conceptual diagram of packet distribution when an optimal solution is obtained using the descent method. In addition, in FIG. 12, the concept of packet distribution when obtaining an optimal solution using the descent method will be described by taking as an example the case where the terminal device 1 transmits a packet to the base station 10.

  The terminal device 1 includes a radio link ML1 by the radio module 12 and a radio link ML2 by the radio module 13 between the base station 10. Then, the terminal device 1 transmits the packets PKT1 to PKT8 to the base station 10 using the radio link ML1, and transmits the packets PKT9 to PKT11 to the base station 10 using the radio link ML2 (see (a) of FIG. 12). ).

The search module 15 of the terminal device 1 transmits the packet arrival rate and the average delay time for each of the radio links ML1 and ML2 in a state where the packets PKT1 to PKT8 and the packets PKT9 to PKT11 are transmitted by the radio link ML1 and the radio link ML2, respectively. And the link costs d _ML1 and d _ML2 of the radio links ML1 and _ML2 are calculated based on the measured packet arrival rate and average delay time.

  Since the packet arrival rate is a rate at which the application processor 16 arrives at the radio link ML1 or the radio link ML2 via the switching unit 14, the search module 15 can measure the packet arrival rate. Since the average delay time is the sum of the waiting time in the queue and the packet processing time, the search module 15 measures the time from when the packet is distributed to the wireless module 12 or 13 until it is transmitted. By doing so, the average delay time can be measured.

When the search module 15 of the terminal device 1 calculates the link costs d _ML1 and d _ML2 of the radio links ML1 and ML2, the search unit 15 calculates the maximum link cost d _ML1 and the minimum link cost d _ML1 and d _ML2 from the calculated link costs d _ML1 and d _ML2 . The link cost d _ML2 is detected.

Then, the search module 15 of the terminal device 1 determines whether the radio link ML1 having the maximum link cost d _ML1 is the same as the radio link previously selected as the link having the maximum link cost, and the maximum If the radio link ML1 having the link cost d _ML1 of the same is the same as the previously selected link, the same number of packets PKT4 to PKT8 are moved to the radio link ML2 having the minimum link cost d _ML2 (FIG. 12). (See (b)). Then, the terminal device 1 transmits packets PK1 to PKT3 via the radio link ML1, and transmits packets PKT9 to PKT11 and PKT4 to PKT8 via the radio link ML2.

On the other hand, when the search module 15 of the terminal device 1 determines that the radio link ML1 having the maximum link cost d _ML1 is different from the radio link selected as the link having the maximum link cost last time, the search module 15 was moved last time. moving the number of packets PKT7, PKT8 less than packet PKT4~PKT8 to the radio link ML2 having the smallest link cost _{d ML2} (see (c) of FIG. 12). Then, the terminal device 1 transmits the packets PK1 to PKT6 through the radio link ML1, and transmits the packets PKT9 through PKT11, PKT7, and PKT8 through the radio link ML2.

Thereafter, in the state shown in (b) or (c) of FIG. 12, the terminal device 1 transmits packets to the base station 10 through the radio links ML1 and ML2, and the packet arrival rate and average delay of each of the radio links ML1 and ML2 The link costs d _ML1 and d _ML2 of the radio links ML1 and ML2 are calculated by measuring the time, and based on the calculated link costs d _ML1 and d _ML2 , the aggregated link costs (= link costs d _ML1 and d _ML2 The packet is moved between the radio links ML1 and ML2 so that (sum) is minimized.

Therefore, when the optimum solution is searched by the descent method, the packet goes back and forth between the radio links ML1 and ML2. If it is determined that the radio link ML1 having the maximum link cost d _ML1 is different from the radio link previously selected as the link having the maximum link cost, the packet going back and forth between the radio links ML1 and ML2 The number gradually decreases as the number of searches for the optimal solution increases.

  FIG. 13 is a flowchart for explaining the operation for obtaining the optimal solution using the descent method. When a series of operations is started, the search module 15 of the terminal device 1 sets n = 1 (step S1), and determines whether n = 1 (step S2).

  If it is determined in step S2 that n = 1, the search module 15 of the terminal device 1 controls the switching unit 14 so that the packets generated by the application processor 16 are evenly distributed to the wireless modules 12 and 13. Then, the switching unit 14 evenly distributes the packet received from the application processor 16 to the wireless modules 12 and 3 (step S3).

  Then, the search module 15 of the terminal device 1 measures the packet arrival rate and average delay time for each wireless link (step S4), and uses the measured packet arrival rate and average delay time to determine the link cost of each link. Calculation is performed (step S5).

  Then, the search module 15 of the terminal device 1 detects the link having the maximum link cost and the link having the minimum link cost based on the calculated link cost (step S6), and determines the maximum link cost. The packet is moved from the link having the link to the link having the minimum link cost (step S7). In this case, the number of packets to be moved is set in the terminal device 1 as an initial value.

  Thereafter, the search module 15 of the terminal device 1 sets n = n + 1 (step S8). Then, the series of operations returns to step S2.

  On the other hand, when it is determined in step S2 that n = 1 is not true, the search module 15 of the terminal device 1 measures the packet arrival rate and the average delay time for each link (step S9), and links for each link. The cost is calculated (step S10).

  And the search module 15 of the terminal device 1 calculates the sum total of the calculated link cost, calculates an aggregate link cost (step S11), and determines whether an aggregate link cost is the minimum (step S12).

  When it is determined in step S12 that the aggregate link cost is the minimum, the series of operations ends.

  On the other hand, when it is determined in step S12 that the aggregate link cost is not the minimum, the search module 15 of the terminal device 1 determines the link having the maximum link cost and the minimum link cost based on the calculated link cost. A link is detected (step S13). Then, the search module 15 of the terminal device 1 determines whether or not the link having the maximum link cost is the same as the previous one (step S14), and if the link having the maximum cost is the same as the previous time, The same number of packets are moved from the link having the maximum link cost to the link having the minimum link cost (step S15).

  On the other hand, when the link having the maximum link cost is not the same as the previous link, the search module 15 of the terminal device 1 links the number of packets smaller than the previous number from the link having the maximum link cost to the link having the minimum link cost. (Step S16).

  Then, after step S15 or step S16, the series of operations returns to step S8.

  Thereafter, steps S2 and S9 to S16 described above are repeatedly executed until it is determined in step S12 that the aggregate link cost is minimum. If it is determined in step S12 that the aggregate link cost is the minimum, the series of operations ends.

  When the search module 15 of the terminal device 1 searches for the optimum solution according to the flowchart shown in FIG. 13, the search module 15 detects the packet distribution rate when the aggregated link cost is minimum, and sends the detected packet distribution rate to the switching unit 14. Output. Then, the switching unit 14 distributes the packet received from the application processor 16 to the wireless modules 12 and 13 according to the packet distribution rate received from the search module 15, and the wireless modules 12 and 13 transmit the distributed packet.

  As a result, the terminal device 1 transmits a packet to the base station 10 with a minimum aggregate link cost.

  Further, in the flowchart shown in FIG. 13, it has been described that the optimal solution is searched by the descent method until it is determined that the aggregate link cost is the minimum. However, the present invention is not limited to this, and the optimal solution is searched. When the number of searches reaches a predetermined number, the search for the optimum solution may be terminated.

  Furthermore, each of the terminal devices 2 to 4 and the base station 10 also searches for an optimal solution (= packet distribution rate when the aggregated link cost is minimized) according to the flowchart shown in FIG.

[Embodiment 2]
FIG. 14 is a schematic diagram of a wireless communication network according to the second embodiment. The wireless communication network 100A includes terminal devices 101 to 106 and base stations 110, 120, and 130.

  The terminal apparatuses 101 to 103 and the base stations 110, 120, and 130 exist in the IEEE 802.11g cell 40, and the terminal apparatuses 104 to 106 are outside the cell 40 and in the IEEE 802.16 cell 50. Exists in a certain area. The communication range of the cell 50 is wider than the communication range of the cell 40.

  The terminal device 101 performs wireless communication with the base stations 110 and 130 through IEEE802.11g and IEEE802.16, and the terminal device 102 performs wireless communication with the base station 110 through IEEE802.11g and IEEE802.16, The terminal device 103 performs wireless communication with the base stations 110 and 120 through IEEE802.11g and IEEE802.16, and each of the terminal devices 104 to 106 wirelessly communicates with the base station 110 through IEEE802.11g and IEEE802.16. Communicate.

  The base station 110 is connected to the network 20 such as the Internet by a wired cable 31, the base station 120 is connected to the network 20 by a wired cable 32, and the base station 130 is connected to the network 20 by a wired cable 33.

  As will be described later, each of the terminal apparatuses 101 to 106 is equipped with a plurality of radio systems, and uses the equipped radio systems appropriately or simultaneously to access the base stations 110, 120, and 130.

  Each of the terminal devices 101 to 106 and the base stations 110, 120, and 130 includes a wireless link ML2 that performs wireless communication by IEEE802.11g and a wireless link ML3 that performs wireless communication by IEEE802.16. Then, each of the terminal apparatuses 101 to 106 and the base stations 110, 120, and 130 distributes the packets to the radio links ML2 and ML3 so that the communication cost between itself and the transmission destination is minimized by a method described later. .

  FIG. 15 is a schematic block diagram showing the configuration of the terminal device 101 shown in FIG. The terminal device 101 is the same as the terminal device 1 except that the wireless module 12 of the terminal device 1 shown in FIG.

  The wireless module 12A performs wireless communication with the base stations 110 and 130 by IEEE 802.16.

  Each of terminal apparatuses 102 to 106 and base stations 110, 120, and 130 shown in FIG. 14 has the same configuration as that of terminal apparatus 101 shown in FIG.

  In the wireless communication network 100A illustrated in FIG. 14, the terminal devices 101 to 106 and the base stations 110, 120, and 130 transmit and receive packets to and from each other using the wireless modules 12A and 13. In order to maintain the wireless communication throughput between the terminal apparatuses 101 to 106 and the base stations 110, 120, and 130 and to reduce the delay time of the wireless communication, each of the terminal apparatuses 101 to 106 and the base station It is effective to distribute packets to the radio links ML2 and ML3 so that the overall communication cost of the two radio links ML2 and ML3 existing between the stations 110, 120 and 130 is minimized. Will be described.

[Mathematical model]
(Link cost)
Also in the second embodiment, the link cost d _i and the aggregated link cost D _i are expressed by the above-described equations (20) and (21), respectively. Further, the network cost G of the wireless communication network 100A is expressed by the above-described equation (5).

(Link characteristics according to IEEE 802.11)
In the second embodiment, a packet distribution characteristic of a radio link by CSMA / CA will be described as a link characteristic by IEEE 802.11.

  As described above, the delay time includes the waiting time in the queue and the packet service time. First, the packet processing time will be described. In Non-Patent Document 4, it is assumed that the packet arrival process is a Poisson process, and the packet processing time is obtained as follows.

Where P _n (t) is the probability of arrival of n packets in time interval t, λ is the packet arrival rate of terminal device i, and P _o (t) is the packet arrival rate in time interval t. Is the probability of not arriving (that is, the probability of successful transmission), | Adj (i) | is the number of neighboring terminal devices of the terminal device i, and P ⁱ _idle (t) is the time interval t of the terminal device i. This is the probability of detecting that the channel is idle, slot is the unit time of backoff, DIFS is the frame interval for distributed control (DCF Inter Frame Space), and SIFS is the short frame interval (Short) an Inter Frame Space), bf is the average packet off interval, EA _i is the expected value of the MAC delay time in trial conditions, EB Is the expected value of the MAC delay time in back-off state, RTS is the transmission time of the transmission request frame, CTS is the transmission time of the reception preparation completion frame, DATA is the data frame transmission time, ACK is the transmission time of the acknowledgment frame, packet_len is the average packet size, tx_rate is the transmission rate, β _i ^MAC is the MAC delay time, β _i ^tx is the transmission time, S _i is the packet processing time in the terminal device i.

The packet arrival process is also assumed to be a Poisson process, and the above-described equations (70) to (79) can be applied. In the following, λ and ^ λ (^ λ indicates that “^” is placed on λ) are replaced with packet arrival rates F _i and F ^c _i , respectively. In the second embodiment, F ^c _i indicates the packet arrival rate of the shared channel by the terminal device i.

The dependence of the packet service time S _{i on} the packet arrival rate F ^c _i will be described. Since the packet service time S _i includes a MAC delay time β _i ^MAC and a transmission time β _i ^tx , the dependency of the MAC delay time β _i ^{MAC on} the packet arrival rate F ^c _i will be described first.

When the collision probability (1-P ⁱ _DIFS ) is used, the average number of collisions U _i in the packet is expressed by the following equation.

In addition, the first and second derivatives of the average number of collisions U _{i by} the packet arrival rate F ^c _i are expressed by the following equations (81) and (82), respectively.

When F ^c _i → ∞, U> 0 and U → ∞. Based on Equation (81) and Equation (82), the average number of collisions U _i is a convex monotonically increasing function of the packet arrival rate F ^c _i .

Therefore, when the packet arrival rate F ^c _i increases, also increases the number of packets to migrate to the back-off state, increases also significant differences in the increase. If EA _i and EB _i are compared, EB _i clearly takes a long time. As a result, when the packet arrival rate ^F _{c i} increases, MAC delay beta _i ^MAC is increased, also increased its significance.

Based on the above, the first and second derivatives of the MAC delay time β _i ^MAC are represented by the following equations (83) and (84), respectively.

Next to the MAC delay time β _i ^MAC , the dependency of the transmission time β _i ^{tx on} the packet arrival rate F ^c _i will be described. Based on Equation (81) and Equation (82), as the packet arrival rate F ^c _i increases, the average number of collisions U _i and its significant difference increase. In this case, the IEEE 802.11 multi-rate control reduces the transmission rate and increases the chance of the average number of collisions U _i increasing. As a result, the first and second derivatives of the transmission rate tx_rate are expressed by the following equations (85) and (86), respectively.

Using Expression (85) and Expression (86), the first and second derivatives of the transmission time β _i ^tx are represented by the following Expression (87) and Expression (88), respectively.

Based on the dependency of the MAC delay time β _i ^MAC and the transmission time β _i ^{tx on} the packet arrival rate F ^c _i described above, the dependency of the packet service time S _{i on} the packet arrival rate F ^c _i is as follows: It is represented by Formula (89) and Formula (90).

In addition, when the increment of the packet arrival rate F ^c _i is δ and the number of terminal devices adjacent to the terminal device i is m, the expected value of the increase of the packet arrival rate F _i is δ / m. .

As a result, dF _i / dF ^c _i > 0, d ² F _i / d (F ^c _i ) ² = 0, the first derivative by the packet arrival rate F _i of the packet service time S _i , and 2 The second derivatives are represented by the following equations (91) and (92), respectively.

Following the dependency of the packet service time S _{i on} the packet arrival rate F _i , the dependency of the waiting time W _i in the queue on the packet arrival rate F _i will be described. The expected interval from the end point of the (i-1) th packet service to the end point of the i-th packet service is expressed as EW _i , and the waiting time W _i is the same as that of the pseudo Bayes algorithm (Non-Patent Document 5). It is approximated using a method. That is, all packets including newly arrived packets are treated as retransmission packets (packets that have failed to be transmitted), and the packet transmission probability in each period is n retransmission packets in the shared channel, and the transmission probability is q. In this case, G = n · q. The transmission probability in each period is assumed to be an exponential distribution with G as an average. Based on the above, EW is shown in the following equation.

  Further, when EW is obtained from the above formula, it is as follows.

The number M of waiting packets is obtained as F _i × W _i by applying the waiting time W _i in the queue and the packet arrival rate F _i of the terminal device _i to the Little theorem. The standby time W _i is M × EW + R. Here, R indicates the sum of the remaining service times of the current service packet at each packet arrival.

In the packet service period, the arrival number of packets is F _i S _i , so the expected remaining service time for each packet arrival is / S ² _i / S _i . / S ² _i is the second moment of the packet service time S _i .

Solving W _i using the above relationship, the following equation.

In order to obtain the dependence of the waiting time W _{i on} the packet arrival rate F ^c _i , the following equations (96) to (99) are obtained by differentiating the equations (94) and (95) with respect to the packet arrival rate F ^c _i. ) Is obtained.

Here, S _i ′ = dS _i / dF ^c _i , S _i ″ = d ² S _i / d (F ^c _i ) ² , G ′ = dG / dF ^c _i , G ″ = d ² G / d (F ^c _i ) ²

As described above, G = n · q (Non-Patent Document 5) indicates that the number n _{τ of} backoff packets that collide τ times is larger than F ^c _i 2 ^r of binary exponential backoff.

Based on the number _nτ , suppose _nτ is F ^c _i . Non-Patent Document 6 shows that the transmission probability q ^r backoff packet n _tau is 2 ^-r binary exponential backoff. As a result, the first and second derivatives of G and its packet arrival rate F ^c _i are expressed by the following equations (100), (101), and (102), respectively.

  Here, U is the average number of collisions.

  Based on the formulas (91), (92) and the formulas (100) to (102), the formulas (97) and (98) become the following formulas (103) and (104), respectively.

Further, dependence on the packet arrival rate ^F _{c i} of the second moment ^{/ S} _{2 i} service time is the same as that of the service time _{S i,} second moment ^{/ S} _{2 i} packet arrival rate _{F i} of The first derivative and the second derivative with respect to are represented by the following equations (105) and (106), respectively.

According to the equations (103) to (106), when F _i <1 / EW _i , the first derivative and the second derivative with respect to the packet arrival rate F _i of the waiting time W _i are It is represented by Expression (107) and Expression (108).

When F _i <1 / EW _i , W _i is positive, and when F _i → 1 / EW _i , W _i → ∞. 1 / EW _i is equivalent to the link capacity. Therefore, according to the equations (107) and (108), the waiting time W _i is a convex monotonically increasing function of the packet arrival rate F ^c _i .

Finally, the dependency of the delay time T _i and the link cost d _{i on} the packet arrival rate F _i will be described. Since the delay time T _i is the sum of the waiting time W _i and the service time S _i , it is a convex monotonically increasing function of the packet arrival rate F ^c _i .

Also, based on Little's theorem, d _i = F _i × T _i , so the first derivative and second derivative of the packet arrival rate F _i of the link cost d _i are respectively expressed by the following equations (109 ) And formula (110).

  Equations (109) and (110) show that the cost of the radio link according to CSMA / CA is a convex monotonically increasing function with respect to the packet distribution rate.

[Link characteristics according to IEEE 802.16]
In IEEE 802.16, bandwidth reservation is supplied based on QoS (Quality of Service) (Non-patent Document 7). When bandwidth reservation is provided, packet arrival rate and packet service time are independent of each other.

  The distribution of the packet service time is not an exponential distribution but a normal distribution due to media control delay, retransmission, and adaptive modulation and coding (AMC).

  Therefore, the wireless link according to IEEE 802.16 can be assumed to be an M / G / 1 link. The link cost of M / G / 1 is expressed by the following equation.

Where C _i indicates the reserved bandwidth, / X ² indicates the second moment of packet service time, and both C _i and / X ² are independent of the packet arrival rate F _i . is there.

The first and second derivatives of the link cost packet arrival rate F _i are expressed by the following equations (112) and (113), respectively.

When F _i <C _i , the link cost d _i becomes positive, and when F _i → C _i , the link cost d _i approaches ∞.

  As a result, the cost of the wireless link according to IEEE 802.16 is a convex monotonically increasing function of the packet distribution rate.

  As described above, the cost of the radio link ML2 according to CSMA / CA (IEEE802.11g) and the cost of the radio link ML3 according to IEEE802.16 are convex monotonically increasing functions with respect to the packet distribution rate. The aggregate link cost obtained by consolidating the link ML3 is a downward convex function with respect to the packet distribution rate to the radio link ML2 by CSMA / CA (IEEE802.11g).

  FIG. 16 is another diagram for explaining a packet distribution method. In FIG. 16, the vertical axis represents the link cost, and the horizontal axis represents the packet distribution rate to the radio link ML2 by IEEE802.11g. Further, the black triangle indicates the relationship between the packet distribution rate to the radio link ML2 by IEEE802.11g and the link cost of the radio link ML2, and the black square indicates the packet distribution rate to the radio link ML2 by IEEE802.11g and IEEE802.11g. 16 shows the relationship with the link cost of the radio link ML3 according to FIG.

  As described above, there is an optimal solution for the aggregate link cost with respect to the packet distribution rate, and the aggregate link cost is a downward convex function with respect to the packet distribution rate (see curve k21 in FIG. 16).

  Accordingly, the above-described formula (69) is repeatedly applied to a high-cost link using the above-described descent method to obtain an optimal solution.

  Some of the packets (determined by α · p (t)) are moved from the high cost link to the low cost link (see arrow ARW5 in FIG. 16). Then, after moving some packets, if the high-cost link is the same as the previous link, the same amount of packets (p (t) = p (t−1)) as the previous link is transferred to the low-cost link. Moving. On the other hand, when a high-cost link is switched, a lower amount of packets (p (k) = α · p (k−1), 0 <α <1) than the previous time from a new high-cost link has a low cost. Move to the link (see arrow ARW6 in FIG. 16). Then, the above operation is repeated to gradually approach the vicinity of the optimum solution while reducing the packet distribution amount (see ARW7 and ARW8 in FIG. 16). This provides an optimal solution for the aggregate link cost.

  Accordingly, in each of the terminal apparatuses 101 to 106 and the base stations 110, 120, and 130 shown in FIG. 14, the search module 15 is an aggregation composed of the radio link ML3 that the radio module 12A has and the radio link ML2 that the radio module 13 has. An optimum solution (= packet distribution rate) when the link cost of the link is minimized is searched by the above-described descent method, and the switching means 14 is distributed so as to distribute packets to the wireless modules 12A and 13 with the searched packet distribution rate. Control.

  The search for the optimal solution in the second embodiment is performed according to the flowchart shown in FIG.

  As described above, according to Embodiments 1 and 2, the search modules 15 of the terminal devices 1 to 4, 101 to 106 and the base stations 10, 110, 120, and 130 are the terminal devices 1 to 4, 101 to 106. The two radio links ML1, ML2 / ML2, and ML3 are arranged so that the link cost of the aggregate link consisting of the two radio links ML1, ML2 / ML2, and ML3 between the base stations 10, 110, 120, and 130 is minimized. The packet distribution ratio when distributing the packet is determined, and the switching means 14 distributes the packet to the two wireless links ML1, ML2 / ML2, ML3 based on the determined packet distribution ratio, and the two wireless modules 12 , 13 / 12A, 13 transmit the distributed packet. That is, the packet is distributed and transmitted to the two radio links ML1, ML2 / ML2, and ML3 so as to minimize the communication load. As a result, the delay time is minimized and the throughput is maximized.

  Therefore, according to the present invention, the communication efficiency of the cognitive radio network can be improved.

  In the above description, it has been described that each of the terminal devices 1 to 4, 101 to 16 and the base stations 10, 110, 120, and 130 includes two wireless modules that perform wireless communication using two different wireless systems. However, in the present invention, the terminal devices 1 to 4, 101 to 16, and the base stations 10, 110, 120, and 130 are not limited to this, and perform wireless communication using three or more different wireless systems. More than one wireless module may be provided, and in general, a plurality of wireless modules that perform wireless communication by a plurality of different wireless systems may be provided.

  In this case, the link cost of one radio link becomes a convex monotonically increasing function with respect to the packet distribution ratio of the plurality of radio links included in the plurality of radio modules to one radio link. The link cost of the wireless link is a convex monotonously decreasing function, so the link cost of the aggregated link that aggregates a plurality of wireless links is a downward convex function with respect to the packet distribution rate to one wireless link. As a result, the optimum solution (packet distribution rate that minimizes the link cost of the aggregated link) can be searched by the above-described descent method. Therefore, the communication efficiency of the cognitive radio network can be improved.

  In the present invention, each of the terminal devices 1 to 4, 101 to 106 and the base stations 10, 110, 120 and 130 constitutes a “wireless device”, and the switching means 14 and the search module 15 are “communication control means”. Configure.

  In addition, the process for obtaining the optimum solution according to the flowchart shown in FIG.

  Further, the wireless system based on IEEE 802.11j constitutes a “first wireless system”, and the wireless system based on IEEE 802.11g constitutes a “second wireless system”.

  Further, the wireless system based on IEEE802.11g constitutes a “first wireless system”, and the wireless system based on IEEE802.16 constitutes a “second wireless system”.

  Further, the wireless module 12 constitutes a “first wireless module”, and the wireless module 13 constitutes a “second wireless module”.

  Further, the wireless module 12A constitutes a “first wireless module”.

  Furthermore, the base station 110 constitutes a “first base station”, and each of the base stations 120 and 130 constitutes a “second base station”.

  The embodiment disclosed this time should be considered as illustrative in all points and not restrictive. The scope of the present invention is shown not by the above description of the embodiment but by the scope of claims for patent, and is intended to include all modifications within the meaning and scope equivalent to the scope of claims for patent.

  The present invention is applied to a radio apparatus capable of improving the communication efficiency of a cognitive radio network. The present invention is also applied to a wireless communication network including a wireless device capable of improving the communication efficiency of a cognitive wireless network.

BRIEF DESCRIPTION OF THE DRAWINGS It is the schematic of the radio | wireless communication network by Embodiment 1 of this invention. It is a schematic block diagram which shows the structure of the terminal device shown in FIG. It is a conceptual diagram for demonstrating a single link and an aggregation link. It is a figure which shows the relationship between the packet distribution rate to the wireless link by IEEE802.11j, and network cost. It is a figure which shows the relationship between cost and a packet distribution rate. It is a figure which shows the relationship between the sum of delay time, and a throughput. It is a figure which shows the model for analyzing packet service time. It is a figure which shows the relationship between the service time of a packet, and the transmission trial rate of a packet. It is a figure which shows the relationship between a packet arrival rate and the transmission trial rate of a packet. It is a figure which shows the relationship between the secondary moment of the service time of a packet, and the transmission trial rate of a packet. It is a figure for demonstrating the distribution method of a packet. It is a conceptual diagram of packet distribution when calculating | requiring an optimal solution using a descent method. It is a flowchart for demonstrating the operation | movement which calculates | requires an optimal solution using a descent method. FIG. 6 is a schematic diagram of a wireless communication network according to a second embodiment. It is a schematic block diagram which shows the structure of the terminal device shown in FIG. It is another figure for demonstrating the distribution method of a packet.

Explanation of symbols

1-4, 21-24, 101-106 Terminal device 10, 110, 120, 130 Base station, 11 Antenna, 12, 12A, 13 Wireless module, 14 Switching means m15 Search module, 16 Application processor, 20 Network, 30 , 40, 50 cells, 31-33 wired cable, 100, 100A wireless communication network.

Claims (6)

A plurality of wireless modules that perform wireless communication using a plurality of different wireless systems;
The wireless communication in the plurality of wireless modules is controlled so that the communication cost indicating the total communication load in the plurality of wireless links established between the plurality of wireless modules with the transmission destination adjacent to the wireless device is minimized. A wireless device comprising communication control means. The communication control means distributes packets to the plurality of wireless modules so that the communication cost is minimized,
The wireless device according to claim 1, wherein each of the plurality of wireless modules transmits the distributed packet.   The communication control means calculates a link cost indicating a communication load in one radio link for the plurality of radio links, selects a maximum link cost and a minimum link cost from the calculated plurality of link costs, and selects the maximum The packet is distributed to the plurality of wireless modules by repeatedly performing a packet distribution process for distributing a packet distributed to a wireless link having a link cost to a wireless link having the minimum link cost. Wireless device.   The communication control means determines that the first radio link having the maximum link cost in the n-th (n is an integer of 2 or more) packet distribution processing is the maximum link cost in the n-1th packet distribution processing. , The same number of packets as the number in the (n−1) th packet distribution process are distributed to the wireless link having the minimum link cost, and the first wireless link is the first wireless link. 4. The wireless device according to claim 3, wherein when different from the two wireless links, the wireless device distributes a smaller number of packets to the wireless link having the minimum link cost than the number in the n−1th packet distribution processing.   When the first wireless link is different from the second wireless link, the communication control means gradually decreases the number of packets distributed to the wireless link having the minimum link cost as the number of packet distribution processes increases. The wireless device according to claim 4. A base station that performs wireless communication using a plurality of different wireless systems;
A terminal device that performs wireless communication with the base station using the plurality of wireless systems,
Each of the said base station and the said terminal device is a radio | wireless communication network which consists of a radio | wireless apparatus of any one of Claims 1-5.

Images