RU2419997C2 - Secure group transfer with linear independent coding of data packages - Google Patents

Abstract

FIELD: information technologies.

SUBSTANCE: in joint planning and coding procedure a composite data package is formed as a suspended linear combination of regular data packages. According weight ratios of coding are adapted on the basis of feedback information from the receiving devices on the accepted data packages so that a composite package represents a new linearly independent coding of regular data packages, different from any data package transferred in a group transfer mode, which has been earlier received by means of each receiving device in a selected set of the receiving devices during a session of group transfer. Besides, using a vector of weight ratios, at least with two different non-zero weight ratios of coding adds an additional extent of freedom and guarantees the possibility to always form a composite data package, which represents a new linearly independent coding for transfer.

EFFECT: strategy efficient in throughput capacity for a secure group transfer, especially when number of users is high.

29 cl, 7 dwg, 3 tbl

Description

FIELD OF THE INVENTION

The present invention relates generally to multicast systems, and more particularly, to a new strategy for scheduling and encoding packets for multicast.

State of the art

There is an ever-growing need to improve the performance of communication systems, in general, and wireless networks, in particular. Multicast transmission is an approach to increase throughput by efficiently transmitting the same information to multiple users / receivers. When the channel is unreliable, it may not be possible to deliver data to all receivers in a reliable manner. Therefore, multicast over untrusted channels typically requires the use of forward error correction (FEC) methods or reverse error correction methods, such as automatic retransmission requests (ARQ).

The so-called fountain coding is an interesting approach for reliable multicast transmission [1-2]. Font encoding is an ideal data encoding scheme without regard to speed, an FEC form that generates new parity information until all users have decoded the sent data file. The approximation of ideal fountain codes are the so-called tornado codes and LT codes. Other fountain codes include so-called tornado codes and operational codes. The basic philosophy in LT codes (and, in general, in most fountain codes), in which new parity information is generated, is based on a random choice of the degree of coding, i.e. the number of packets to be coded together according to a predetermined power distribution and the subsequent bitwise XOR operation (modulo 2 addition) of the packets. Although not internally part of fountain coding, acknowledgment is typically sent when the receiver has received a sufficient number of parity packets that enable decoding, and not earlier. Therefore, only one acknowledgment is required per FC file for each user. The FC approximation data provides efficient multicast transmission in erased channels where more redundancy than data should be received by approximately ε.

ARQ is an effective error control strategy for transmitting data in which the receiver detects transmission errors in messages and automatically requests retransmission from the transmitter. Typically, feedback messages from receivers to a transmitter are used to enable the transmitter to determine which packets were received correctly by the given receivers and which were not. Packets that are not received correctly must be retransmitted.

ARQ, as proposed, should be used as a standard for communication over the air in cellular wireless communication systems and can also be used in multi-span systems. Typically, information is divided into smaller packets, called protocol data units (PDUs), prior to transmission. Reliable transmission is ensured by encoding packets with an error detection code so that the receiver can detect erroneous or lost packets and thereby request retransmission. The integrity of the data sequence is usually ensured by sequentially numbering the packets and applying certain transmission rules.

It may be useful to first review the basic principles of ARQ based on unicast transmission before explaining the details regarding multicast ARQ, bearing in mind that the basic principles of unicast ARQ transmission are mainly carried over to multicast ARQ.

In the simplest form of ARQ, commonly referred to as stop and wait ARQ, the transmitting data device stores each sent data packet and waits for acknowledgment from the receiver of the correctly received data packet through an acknowledgment message (ACK). When the ACK is received, the transmitter discards the stored packet and transmits the next packet. The process is typically supplemented by timers and the use of negative acknowledgment messages (NACK). The transmitting entity uses a timer that starts when a data packet is transmitted, and if an ACK (or NACK) is not received before the timer expires, the data packet is retransmitted. If the receiving device detects errors in the packet, it may transmit the NACK to the transmitting device. After receiving the NACK, the transmitter retransmits the data packet without waiting for the timer to expire. If the ACK or NACK message is lost, the timer ultimately expires and the transmitting device retransmits the data packet.

From a simple stop-and-wait transmission, more elaborate traditional ARQ schemes have been created, for example, the method of returning to N and the method of selective rejection (or selective repetition), which provide higher throughput.

In the other direction of ARQ development, coding redundancy is used in various ways to increase communication performance (generally, measured as throughput). These schemes are referred to as hybrid ARQ schemes. The combination of coding and ARQ, a hybrid ARQ scheme, allows you to provide some adaptation to changes in the radio environment, for example, attenuation. In hybrid ARQ 1, FEC is combined with ARQ. In hybrid ARQ 2, PDUs are sent to a greater or lesser extent encoded by FEC, but with the attached cyclic redundancy check (CRC) code to check for bit errors after decoding, and if retransmission is requested, parity bits (also known as redundancy bits) generated by the FEC encoder, systematic bits, or a combination of both.

If efficiency is not the primary goal, then several parallel ARQ unicast processes can be used to provide reliable "multicast". This may be a standard approach when the size of a multicast group, i.e. the number of preset receivers is small. However, for large groups, this communication method is ineffective. Therefore, in a typical embodiment, it is considered more efficient to combine parallel unicast processes into one so-called multicast ARQ process, in which one information is transmitted to several users [3].

Reference [4] relates to performance optimization for reliable multiple unicast streams.

References [5-7] describe schemes in which one data is transmitted in a group mode to several users, and effective retransmission is provided by combining several packets negatively acknowledged (NACK transmission) by several receiving devices into one block.

The problem with traditional multicast ARQ, in general, is that performance is extremely low when the number of users is large. In fact, it should be noted, as well as easy to understand, that throughput tends to zero when the number of users K tends to infinity.

Although fountain coding mitigates performance problems in traditional multicast ARQ, its performance is approximately T = p / (1 + ε), where T is the bandwidth, ε depends on the amount of data sent, and p is the reception probability. In [2], it was approximately established that 1 + ε is greater than 1.06 with a probability of 1/10, and not more than 1.10 for 10,000 tests of the so-called implementation of the Z-code tornado. In addition, it was found that tornado codes do not have sufficiently clear limits on decoding inefficiency for commercial applications and that for current commercial implementations of digital fountain LT codes, decoding inefficiency is more than 1.05 with a probability of less than 10 for almost any size of the source file.

Fountain coding has several drawbacks. FC is based on decoding statistics by providing redundancy information and is not based on determinable decodability. First, in a typical embodiment, large amounts of information are required for a small ε value, and performance is typically low when only a small number of data packets are transmitted. The time to decode the delivered data is quite significant, since all redundancy must be obtained before decodability can be guaranteed. LT codes, a form of fountain codes, are known to have an unsatisfactory lower error limit, which is at least partially resolved by other FCs, such as tornado codes.

Although the circuits proposed in references [5-7] represent a step in the right direction for multicast ARQ with deterministic decodability, these circuits still do not provide optimal performance. The reason for this is that the traditional ARQ multicast operation is more static than necessary and does not recognize or use all degrees of freedom.

Thus, there is a general need for an even more efficient bandwidth strategy for reliable multicast.

Disclosure of invention

The present invention overcomes these and other disadvantages of the means of the prior art.

The overall objective of the present invention is to provide a high throughput strategy for reliable multicast.

An object of the invention is to provide a method and means for scheduling and encoding data packets for transmission in a multicast session from a transmitter to at least two receivers in an ARQ multicast system.

The goal is also to provide a transmitter configured to be used in an ARQ multicast system.

The specific goal is to provide a multicast strategy that provides increased throughput, reduced latency and / or reduced power consumption.

It is especially desirable to provide a multicast strategy that provides deterministic decodability of encoded packets.

These and other objectives are achieved by the invention in accordance with the attached claims.

The invention generally relates to scheduling and encoding data packets for transmission in a multicast session from a transmitter to a multicast group from at least two receivers in an ARQ multicast system.

The basic idea of the invention is to perform a procedure in each of a plurality of planning events, including generating at least one composite data packet as a weighted linear combination of conventional data packets based on a corresponding weighting vector containing a set of encoding weights, by using feedback information from receivers indicating received data packets and transmitting the generated composite data packet (s) to riemnye apparatus of this group multicast. In this procedure, the weighting coefficients of the weighting vector coding are adapted for each composite packet or data packets based on the feedback information so that the composite packet is a linear combination that is linearly independent for each receiver, at least from a set of receivers , from any data packet previously received by the receiver during the multicast session in question.

By performing this procedure in connection with each of a plurality of scheduling events, the number of transmissions required to transmit a given amount of information transmitted in the multicast mode will be reduced.

The ability to transmit new linearly independent data packets (composite data packets and, optionally, also regular packets), at least in most transmission events during a multicast session, significantly increases the throughput for reliable multicast, especially when the number of users is large.

In an ideal situation, during the systematic transmission of a certain number of regular and / or composite data packets that are linearly independent of any data packet (regular or composite) previously received by each of the receivers consisting of a set of receivers during a multicast session , just enough to send N such data packets to retrieve all N regular data packets, thus obtaining maximum throughput.

In any case, compared to traditional multicast ARQ, the invention requires that fewer data packets are sent from the transmitting device to the receiving devices in order to ensure that the transmitted data packets are correctly received.

Thus, the invention provides increased throughput for a (fully) reliable multicast, and also provides a deterministically decoded retransmission scheme.

For example, in fountain coding, the number of subpackets that must be received to decode the complete file is a random variable, i.e. is not deterministically limited.

The invention provides a system that implements the aforementioned scheme, as well as a corresponding transmitting device configured to be used in an ARQ multicast system.

The present invention is generally applicable to multicast ARQ in communication systems having unreliable communication lines, such as wireless communication systems. For example, the invention can advantageously be used in a typical cellular system with a base station supporting communication with a plurality of mobile terminals.

The invention may offer the following advantages:

- Reliable multicast.

- Increased (or even maximum) bandwidth.

- Reduced (or even minimal) delay time.

- Efficient use of energy / power.

- Determined limits on the number of packets that must be received for decodability.

Other advantages proposed in accordance with the invention should be taken into account when reading the following description of embodiments of the invention.

Brief Description of the Drawings

The invention, together with its additional objectives and advantages, will be best understood by referring to the following description, taken in conjunction with the accompanying drawings, of which:

Figure 1 illustrates the throughput of a fully reliable multicast ARQ for K = 1, 2, 4, 8, 16, 32, 64, and 128 as a function of the probability of correct reception.

2 is a schematic block diagram illustrating a simple example of a general representation of an ARQ multicast system.

FIG. 3 is a flowchart illustrating an exemplary complete ARQ procedure.

4 is a flowchart of a method for scheduling and encoding conventional data packets into a composite data packet according to a preferred exemplary embodiment of the invention.

5 illustrates a possible embodiment of encoding two conventional data packets into a composite data packet.

6 is a schematic block diagram illustrating a system for multicast data packets from a transmitter to multiple users according to a preferred exemplary embodiment of the invention.

7 is a schematic block diagram of scheduling and coding a data packet according to an exemplary embodiment of the invention.

The implementation of the invention

In the drawings, the same reference numerals are used for corresponding or similar elements.

Further, the term “regular data packet” refers to a standard (non-composite) data packet (i.e., only a simple ordinary data packet), whereas a composite data packet is based on at least two ordinary data packets. A “multicast data packet” may be a regular data packet or a composite data packet, and in its most universal meaning, this data packet may be referred to as a “general data packet”, “multicast data packet” or “general transmitted in multicast mode data packet. "

The problem with traditional ARQ multicast, in general, is that performance is extremely low when the number K of receivers / users is large. Figure 1 illustrates the throughput of a fully reliable multicast ARQ for K = 1, 2, 4, 8, 16, 32, 64, and 128 as a function of the probability of correct reception. The far left curve illustrates the situation for K-1. As you can see, the performance decreases as the number of K nodes increases, and approaches zero when the number of nodes approaches infinity, except in the special ideal case when the probability of correct reception is 1.

Thus, there is a general need for a more bandwidth-efficient strategy for reliable multicast.

It will be helpful to start with a brief overview of the ARQ multicast system. In the ARQ communication system to be used with the present invention, the transmitter 100 is generally involved in a multicast session with multiple receivers 200, as schematically illustrated in FIG. 2. The transmitting device 100 is usually represented by the transmitting node, such as a base station, in the transmission operation, and the receiving devices 200-1 to 200-K are usually represented by the receiving nodes, for example, implemented by mobile terminals, in the receiving operation. It should be noted that the mobile terminal, for example, can also act as a transmitting node, and the base station as a receiving node. In a multicast transmission, the same set of information is usually transmitted to all receivers that are part of a multicast group. A multicast group generally includes at least two receivers, sometimes called users. To transmit a certain number of data packets in a multicast session, this generally means that each individual data packet is a multicast data packet intended for all receivers in question. It should be noted that if communication is intended for all users within, for example, a cell or system, it is often referred to as a broadcast transmission rather than a multicast. Therefore, broadcast transmission is an extreme case of multicast, in which all receivers in a cell, etc. receive data.

The present invention can advantageously be used in typical cellular systems with a base station that supports communication with multiple mobile terminals. In general, various scheduling schemes can be used in such systems to increase productivity, and ARQ is used to increase transmission reliability.

The invention is described in a wireless scenario, but with the same success can be used in other communication scenarios with unreliable communication lines.

The main idea of the invention is based on the execution of the procedure in each of the many planning events, including the formation of at least one composite data packet as a weighted linear combination of conventional data packets based on the corresponding weighting vector containing a set of encoding weighting factors, using information feedback from receivers indicating received data packets and transmitting the generated composite data packet (s) to the receivers e devices of the multicast group under consideration. In this procedure, the weighting coefficients of the weighting vector encoding are preferably adaptively selected for each composite packet or data packets based on the feedback information so that the composite packet is a linear combination of conventional data packets linearly independent of any data packet (regular or composite) previously received by each of the receivers in the relevant set of receivers of the multicast group during the multicast session.

The relevant receiver set mentioned above generally includes at least two receivers and may be a subset or a complete set of receivers in the entire multicast group.

By “previously received data packet”, in general, is meant a multicast data packet (regular or composite) that is correctly demodulated and decoded according to the FEC scheme on the receiver side.

By performing this procedure in each of the many (two or more) planning events and by transmitting new linearly independent data packets (composite data packets and, possibly, also ordinary data packets), the number of transmissions required to transmit a given amount of transmitted in multicast mode information is reduced, and throughput for reliable multicast is enhanced.

Specialists in the art should understand that the relevant set of receivers / users, which should be considered when adapting the coding weights, changes as a multicast session is performed, and may differ from one scheduling event to the next. Once the receiver has received all data packets transmitted in multicast mode, this receiver is no longer considered. In general, only receivers that still need to receive one or more multicast data packets are considered as a relevant multicast group set. This means that during the entire multicast session, the relevant set of receivers considered in the planning process (i.e., adaptation of the weights) gradually decreases until all receivers have received all the information transmitted in the multicast mode.

The inventor has identified the importance of adapting code weights based on a previously accepted set of packets for each receiver in a selected set of receivers, in contrast to the selection in the prior art of a subset of receivers for which a static code vector of only one is applied, provided that all the receivers devices accepted all packets except one.

As explained in detail later, the idea of linearly independent coding of a data packet can be described with respect to increasing the rank of a matrix of weighting vectors (containing weighting coding coefficients). More specifically, previously received multicast data packets (regular and / or composite data packets) for each receiver can be described as a linear system of equations based on a matrix of weight vectors. The increase in rank corresponds to the formation of a new linearly independent row in the matrix of weight vectors, which is equivalent to the new linearly independent encoding of the data packet transmitted in the multicast mode.

For a better understanding of the invention, it may be useful to consider an illustrative example. Consider a transmitter with two conventional data packets that need to be transmitted to two receivers / users U ₁ and U ₂ . Table I below illustrates an example of a possible transmission / retransmission scenario according to an exemplary embodiment of the invention. For example, coding set “001” means that the first regular packet is multiplied by a coding weight of 0, the second regular packet is multiplied by a coding weight of 0, and the third regular packet is multiplied by a coding weight of 1, effectively meaning that only the third regular packet is transmitted . In the same way, “001” for the user means that this particular user has received the third regular data packet, while “x” for the user means that the information has not been received correctly by that user.

Table I TX # Set U ₁ U ₂ Comments one 001 001 X U ₁ receives the third regular packet. 2 010 X 010 U ₂ receives the second regular packet. 3 one hundred X one hundred U ₂ receives the first regular packet. four 111 111 X It is possible, for example, that not 110 is sent (since the rank for U ₂ would not increase), but 111, 101 and 011 OK. U ₁ receives a composite packet based on weights 111. 5 011 011 X U ₁ receives a composite packet based on weights 011. 6 001 001 U ₂ receives a composite packet based on weights 001.

Example table I shows two users receiving three packets. It should be noted that, in general, it is not enough to receive only regular packets and one composite packet at first to guarantee decoding of all ordinary data packets. Instead, it is illustrated that the transmitter must adaptively select code weights, and in this particular example, it does this at every step so that the rank of the receiver-specific matrix of each user is increased for all relevant users if they receive a composite packet. Adaptation of weights, as described here, instead of adapting the set of users to whom one composite packet is sent, as described in the documents of the prior art [5-7], dramatically improves productivity. In addition, references [5-7] do not use user-specific matrices for planning and coding. Additionally, it should be noted that in references [5-7], the transmitting device does not send packets so that the receiving device can use several composite packets simultaneously for decoding.

In this example, it can also be noted that as soon as one user has received a sufficient number of data packets, regular and composite data packets, and the rank of the corresponding matrix is complete, it is no longer necessary or even impossible to further increase the matrix rank of this receiving device. Therefore, only matrices of incomplete rank are considered when determining adaptive code weights, which ensures that the rank increases for the remaining receivers of the matrix of incomplete ranks. Assume that the coding weights of the composite data packet in the first scheduling event are adapted so that the composite packet is a linear combination that is linearly independent for each receiver from the first set of receivers, from any data packet previously received by the receiver in multicast session time. In a second subsequent scheduling event, the encoding weights of another composite data packet can thereby be adapted so that the composite packet is a linear combination that is linearly independent for each receiver from the second smaller set of receivers, from any data packet previously received by the receiver during a multicast session.

Naturally, in a more realistic scenario, the number of users and the number of data packets are often larger, but the above simplified example serves to illustrate key features of the present invention. In practical applications, the number of users / receivers K may be any number from two to, possibly, a hundred or more.

Typically, a linear combination of, for example, two conventional data packets, such as packet A and packet B, simply means that the packets are combined with equal weighting, effectively using the same encoding weight of “1”. Those conventional data packets (such as packet C and packet D) that are not considered in linear combination may be considered to have an encoding weight of “0”.

Under certain circumstances, especially when a multicast group includes three or more users, the ability to select a weighting vector having at least two different non-zero encoding weights increases (and in most cases guarantees) the ability to form a composite data packet, which is a new linearly independent coding of ordinary data packets, different from any data packet (regular data packet or composite data packet) previously nyatogo by each relevant receiving device that is part of a selected set of receivers during the multicast session.

Consider a transmitter with three regular data packets that must be transmitted to three receivers / users U ₁ , U ₂ and U ₃ . Table II below illustrates an example of a possible transmission / retransmission scenario according to another exemplary embodiment of the invention. For example, the encoding set “01” means that the first regular packet is multiplied by the encoding weight of 0, and the second regular packet is multiplied by the encoding weight of 1, effectively meaning that only the second regular packet is transmitted. In the same way, “01” for the user means that this particular user has received the second regular data packet, while “x” for the user means that the information has not been received correctly by that user.

Table II TX # Set U ₁ U ₂ U ₃ Comments one 01 01 X X U ₁ receives a second regular packet. 2 10 X 10 X U ₂ receives the first regular packet. 3 eleven X X eleven U ₃ accepts the composite packet. four 12 12 X X It is necessary to go beyond the field {0,1} in order to find a new linearly independent coding: 12. 5 01 X X U ₁ is the full rank, so only U ₂ and U _{3 are} considered. None accept. 6 01 - X 01 U ₃ receives the second regular packet. 7 01 01 U ₁ and U ₃ are the full rank, so only U _{2 is} considered, U ₂ receives the second regular packet.

With reference to table III below, another example transmission scenario is described. In this particular example, consider a transmitter with three regular data packets that must be transmitted to three receivers / users U ₁ , U ₂ and U ₃ .

Table III TX # Set U ₁ U ₂ U ₃ Comments one 001 001 X X U ₁ receives the third regular packet. 2 010 X 010 X U ₂ receives the second regular packet. 3 one hundred X one hundred X U ₂ receives the first regular packet. four 101 101 X X Perhaps not 110 is sent (since the rank for U ₂ would not increase), but 101 and 011 OK. U ₁ receives a composite packet based on weights 101. 5 011 X X 011 U ₃ receives a composite packet based on weights 011. 6 111 X X 111 U ₃ receives the composite packet based on weights 111. 7 211 X 211 X It is necessary to go beyond the scope of the field {0,1} in order to find a new linearly independent coding: 211. U ₂ receives a composite packet based on weighting factors 211. 8 110 110 - X U ₂ is the full rank, so only U ₁ and U _{3 are} considered. 9 001 - - 001 U ₁ and U ₂ are the full rank, so only U _{3 is} considered.

It should be noted from Table II that each user receives only two different data packets, which can be composite data packets, ordinary data packets, or a combination of both. It can also be confirmed that each user can define these two ordinary data packets, since it is envisaged that each user in each transmission is guaranteed to receive a new linear combination of the usual data packets. The same applies to table I and table III.

As noted above, the special ability to adaptively select a weighting vector having at least two different nonzero coding weights increases the ability to always generate a composite data packet, which is a new linearly independent coding of ordinary data packets, different from any data packet ( conventional data packet or composite data packet) previously received by each receiver for all or a relevant subset of receivers in emya multicast session. In certain circumstances, especially when the number of users is large, this can be important, since the ability to transmit new linearly independent data packets (regular and / or composite data packets) for each receiver, at least in most of the transmission events during the session multicast transmission significantly increases throughput for reliable multicast.

In general, by systematically transmitting a certain number of packets transmitted in multicast mode or common data packets (regular and / or composite data packets) that are linearly independent of any multicast data packet (ordinary data packet or composite data packet) previously received by a set of receivers during a multicast session, it’s enough to send just N such data packets in order to retrieve all N ordinary data packets, thus getting maximum throughput. However, if the receiver receives more than N data packets that are based on only N regular data packets, it is still possible to retrieve N regular data packets.

Compared to traditional multicast ARQ, the invention requires that fewer multicast data packets are sent from the transmitter to receivers in order to ensure that the transmitted data packets are correctly received. The invention provides a strategy for reliable multicast with deterministic limits on the number of packets that must be received for decodability.

In an exemplary embodiment of the invention, it is also possible to form several data packets in a given scheduling event and transmit several composite packets in packet transmission. In such a planning event, several composite data packets are usually generated using the same feedback information, so that several composite data packets are dependent on each other (although each of them is still linearly independent from previously received packets by each receiver from a relevant set of receivers devices, as well as in packet transmission of the above transmitted composite packets). If this is done systematically in a plurality of scheduling events, it may be possible to reduce the amount of feedback overhead required to transmit a given amount of multicast information.

For a better understanding of the invention, an exemplary complete ARQ procedure will now be described briefly with reference to the flowchart of FIG. 3.

In step S1, one or more data packets are sent from one or more nodes in a send operation to at least two receiving nodes. Each separate data packet is intended for all considered receiving nodes, i.e. data packets are “multicast data packets”. Data packets can be transmitted simultaneously or sequentially, depending on the transmission technology. In step S2, the receiving nodes store the corresponding data packets that are correctly received (correctly demodulated and decoded according to the FEC scheme). In step S3, the receiving nodes return information of packets that they have received to the transmitting node (s). In step S4, the transmitting node generates, if advantageous, a composite data packet transmitted in multicast mode by using feedback information about received data packets from receiving nodes 110, 120, or alternatively sends a conventional multicast data packet. A composite multicast data packet is typically generated based on several selected conventional multicast data packets that must be (retransmitted). For example, the number of bits in a composite packet transmitted in multicast mode may advantageously be less than the sum of the number of bits of the packet parts that are jointly encoded. The transmitting node may form many different constituent packets transmitted in the multicast mode for transmission, possibly in combination with the transmission of conventional multicast packets of data to optimize throughput. In step S5, the transmitting node transmits a composite data packet transmitted in the multicast mode or a selected conventional multicast data packet transmitted to all receiving nodes. After receiving the composite data packet transmitted in the multicast mode, the receiving nodes decode, if possible, the composite data packet transmitted in the multicast mode and retrieve the corresponding data transmitted in the multicast transmission mode, which were previously unknown, for the corresponding receiving node. In the decoding and extraction process, multicast data packets previously decoded according to the FEC scheme by the receiver are preferably used.

In this context, the invention is specifically intended to formulate (encode) composite packets and determine (schedule) which conventional data packets to use (and how) in each composite data packet. In this sense, the formation of a composite data packet can be considered as a joint planning and coding procedure. This mainly relates to the side of the transmitting device, as illustrated in the flowchart of FIG. 4. In step S11, the transmitting device receives feedback information indicating received data packets from the receiving devices. As previously mentioned, a key characteristic of the invention is to form at least one composite multicast data packet as a weighted linear combination of conventional data packets based on a carefully selected set of encoding weights, as indicated in steps S12 and S13 . The coding weights for at least one composite data packet are adapted in step S12 based on the feedback information of previously received data packets so that the composite packet is a new encoding of conventional data packets linearly independent of any data packet previously received by each receiver in a relevant set of receivers during a multicast session. This means that the encoding weights are adaptively selected in response to previously received packets of each receiver via feedback to provide a new linearly independent encoding of conventional data packets. In step S13, the composite data packet is preferably formed by multiplying the conventional data packets by the respective adapted coding weights. In step S14, the generated composite data packet is transmitted to the receivers. As illustrated in the full ARQ flowchart of FIG. 3, the procedure is preferably repeated until all normal packets in a multicast session are transferred to all relevant multicast group users.

Additionally, several composite data packets can be generated and transmitted in this scheduling event by repeating steps S12-S14, as indicated by the dashed line in FIG. 4. This is described in more detail later.

Feedback from the receiving nodes to the transmitting node may preferably be part of an acknowledgment procedure (ACK). Feedback messages can typically be composed in various ways, for example, depending on the transmission protocols used, ARQ procedures, etc.

Various encoding methods and formats may be used with the invention, as described in more detail below.

It should be understood that with the present invention, nothing prevents the complete procedure from starting in step S1 (see FIG. 3) by transmitting a composite data packet.

It should also be noted that transmissions can be carried out simultaneously if a transmission technology is used that allows parallel transmissions, for example, orthogonal frequency division multiple access (OFDMA) is used, i.e. the first packet and the second packet are sent in parallel, but over disjoint sets of OFDM subcarriers. Retransmission can also be carried out simultaneously.

In a more mathematical approach, in order to explain the invention, previously received multicast data packets (regular and / or composite data packets) for each receiver can be described as a linear system of equations based on a matrix of weight vector vectors. For a previously received composite data packet, the weight vector denoted by ω should include a set of coding weights, while for a previously received conventional data packet, the weight vector should include one non-zero coding weight. In a coding and scheduling event, when generating a new composite data packet for transmission to receivers, the idea is to adaptively select the coding weights of a weighted linear combination of conventional data packets based on the feedback of previously received data packets so that the rank should increase for a linear systems of equations for each receiver in question. The increase in rank corresponds to the formation of a new linearly independent row in a matrix of weight vectors specific for each receiver, which is equivalent to the new linearly independent encoding of a data packet transmitted in the multicast mode.

Now, when each user receives one or more data packets transmitted in multicast mode (conventional and / or composite data packets), they can be considered as a system of linear equations in which data packets transmitted in multicast mode are known, weightings are known, and each the set of receivers seeks to solve its linear system of equations in order to determine and decode ordinary data packets. On the side of the transmitting device, the goal is to form (dispatch and code) the corresponding system of equations, and on the side of the receiving device, the goal is to solve (decode) the linear system of equations and extract ordinary data packets.

If a new linearly independent multicast data packet is transmitted each time so that the rank is sequentially increased for each received data packet (for each user receiving the data packet) or at least most times, the throughput is significantly increased. In the first case, when a new linearly independent multicast data packet is transmitted each time, the decodability of N ordinary data packets can be guaranteed when just N data packets are received, since the matrix of weight vectors includes N linearly independent rows, of which all N regular data packets can be retrieved. As the matrix rank increases for each received data packet, it is sufficient to ensure that all users receive only N data packets in order to decode and retrieve all N ordinary data packets.

In addition, in order to provide that the rank increases for all or a subset of receivers in question, the inventors have found that sometimes it is not enough to use weights that can only allow values 0 and 1. Sometimes, when the number of receiving nodes is large relative to the number of regular packets that are provided for transmission, it will be useful to use a weighting vector that has at least two different non-zero weighting coding coefficients, as explained by it.

Note that although a linear combination, in principle, can be performed using conventional arithmetic, leading to the transfer of bits that must be generated for the composite packet, it is preferable to use an indicator that ensures that the length of the composite packet is almost the same as for regular packages. Therefore, arithmetic based on finite fields is preferable and is assumed further. A possible type of arithmetic is the so-called modulo arithmetic, but the invention is not limited to it. Also note that a linear encoding operation may be performed for a complete data packet, or alternatively, conventional data packets are divided into several more manageable substrings for which one linear encoding operation is performed.

Note that it is further assumed that we operate with any type of arithmetic, which provides the same rules as traditional arithmetic, which enables the formation (planning and coding) and solution (decoding) of a linear system of equations. Therefore, without loss of generality, the equations are further written in the traditional arithmetic notation.

Firstly, a priori information of each user, i.e. previously received common data packets transmitted in multicast mode (regular and / or composite data packets) can be described as a linear system of equations according to the following:

- это значение для m_k-го общего передаваемого в режиме групповой передачи пакета данных для пользователя k,

- это весовой коэффициент для пользователя k и обычного пакета n и принимаемого общего передаваемого в режиме групповой передачи пакета данных m_k пользователем k, M_k - это последний принимаемый пакет данных пользователем k, а N - это общее число обычных пакетов, которые должны быть переданы. Отметим, что каждый пользователь, возможно, принял различное число передаваемых в режиме групповой передачи пакетов данных, и для этого нужен индекс k в переменной M_k . where d _n is the value of the nth regular packet,

is the value for the m _k th total data packet transmitted in multicast mode for user k,

is the weight coefficient for user k and the usual packet n and the received total data packet transmitted in multicast mode m _{k by} user k, M _k is the last received data packet by user k, and N is the total number of ordinary packets that must be transmitted . Note that each user may have received a different number of data packets transmitted in multicast mode, and this requires the index k in the variable M _k .

Now, to simplify the notation system, we can write (1) in matrix form for user k :

C _k = W _k D (2)

The formation (coding) and solution (decoding) of a system of linear equations, possibly, as indicated earlier, is carried out using ordinary arithmetic, "modulo arithmetic" or any other suitable arithmetic.

могут допускать, в типичном варианте ограничиваются по числу. Дополнительно, изобретение не устанавливает специальным образом то, какие значения можно допускать для весовых коэффициентов, и они могут быть выбраны из произвольного набора, к примеру, {0,..., 3} или даже некоторого более мнимого произвольного набора, такого как {0, 1, 2, 3, 5, 7, 11, 13}. Чем больше число значений, которые могут быть допущены, тем большая степень свободы представляется, что помогает находить кодовый вектор, который может быть использован для того, чтобы формировать составной пакет данных. Кроме того, большее число пакетов влечет за собой большую степень свободы, что помогает находить кодовый вектор, который может быть использован для того, чтобы формировать составной пакет данных.Values that weights

may admit, are typically limited in number. Additionally, the invention does not specifically establish what values can be allowed for weights, and they can be selected from an arbitrary set, for example, {0, ..., 3} or even some more imaginary arbitrary set, such as {0 , 1, 2, 3, 5, 7, 11, 13}. The larger the number of values that can be allowed, the greater the degree of freedom that seems to help find a code vector that can be used to form a composite data packet. In addition, a larger number of packets entails a greater degree of freedom, which helps to find a code vector that can be used to form a composite data packet.

However, the alphabet for the weighting coefficients of the code vectors in a typical embodiment should be large enough to obtain a sufficient degree of freedom to enable all the users in question to receive N linearly independent equations. However, the alphabet does not have to be very large. The simulation demonstrates that for K = 20 users, N = 6 packets, where the reception probabilities for users are independent, and each reception probability is assumed to be equal to p = 0.05, the alphabet size is sufficient (or, equivalently, base b) only in 3 elements {0, 1, 2} (or, equivalently, base b = 3).

To realize the idea of the minimum value for the base b (or, equivalently, the minimum number of elements in the alphabet), we know that the number of code words b ^N must be greater than the number that each of K users accepts, in the worst case, N different transmitted in multicast data packet mode. Therefore, the condition is that b ^N > N · K. Although this is not an ideal estimate, it nevertheless indicates that b, N and K are interconnected. There is also usually a dependence on the probability of receiving p, and with a high probability of receiving, there is a high probability that several users receive one data packet.

Also note that it is possible that a string in W _k and a code vector or a weight vector contain only one nonzero element. This is equivalent to sending just one regular packet for a common multicast data packet. The invention preferably uses finite field arithmetic, which allows more elements to be used than just “0” and “1”. When the number of users is large and the number of conventional data packets to be transmitted is small, this ensures the discovery of new linear combinations.

In a systematic approach, when it is desirable to find a new linear combination every time, the basic sequence of operations is as follows:

- Consideration of a group of group transmission of users, where users are indexed by k

- For N ordinary multicast data packets

- For each user k, extracting the weight matrix W _k

k выбор кодового вектора так, чтобы Rank(W'_k)=Rank(W_k)+1, где

для тех k, где Rank(W_k)<N-

k select a code vector so that Rank (W ' _k ) = Rank (W _k ) +1, where

for those k where Rank (W _k ) <N

- Formation and sending of a composite packet based on the code vector ω

- Updating all weight matrices based on reception feedback.

Another equivalent formulation, also resulting in an increase in rank, is as follows:

- Consideration of a group of group transmission of users, where users are indexed by k

- For N ordinary multicast data packets

- For each user k, extracting the weight matrix W _k

- Definition of a null space for W _k and its designation as Ξ _k

k выбор кодового вектора ω так, что

для тех k, где Rank(W_k)<N -

k the choice of the code vector ω so that

for those k where Rank (W _k ) <N

- Formation and sending of a composite packet based on the code vector ω

- Updating all weight matrices based on reception feedback.

As explained previously, another way of formulating this is to claim that the encoding weights are selected such that a new linearly independent encoding of the ordinary data packets is sent each time.

The above operation ensures that any user receiving a general data packet transmitted in multicast mode always increases his rank. When the user has received N multicast data packets, the equation system must have N linearly independent lines, i.e. full rank. Therefore, the user can decode all the usual data packets, receiving only the minimum number of total data packets transmitted in the multicast mode. Since it may take more time for a specific user to receive their N multicast data packets, some other users may receive their required N multicast data packets and then some subsequent data packets that are redundant and may be discarded (or not received) .

A possible approach in order to find a suitable code vector ω is based on the hypothesis of iterative testing of preliminary code vectors ω. For example, we choose an alphabet (i.e., a field) of values that the elements of the code vector ω can admit. Given the alphabet, we order the list with all possible permutations, i.e. a list that has b ^N variants of the code vector (where b is the number of elements in the field, and N is the number of ordinary data packets to be transmitted). We search the list from beginning to end (or as far as required) and, in turn, assign each element in the list as a variant of the code vector ω. For each variant of the code vector ω, we check whether the condition is satisfied according to the coding / scheduling algorithm (i.e., the rank increases or alternatively a nonzero vector projects ω onto zero space) for all users with matrices of weight coefficients of incomplete rank. If the condition is fulfilled, we choose the variant of the code vector as the code vector ω.

In fact, a variant of the code vector can be created by adding the value "1" to the previous variant of the code vector, i.e. just a counter. In order to provide a suitable set of values for weighting factors, for example, from zero to a certain number b-1, a variant of the code vector can use and perform counting with base b, where each digit must correspond to a weighting factor.

As an alternative to counting in the forward direction with base b for the next version of the code vector, you can instead go through all or a set of words with a weight factor of "1", i.e. only one weighting factor is nonzero, and then all or combinations of a set of words with a weighting factor of "2" and so on. The idea here is to keep the number on non-zero elements of the weight coefficient as low as possible in order to minimize the encoding and decoding complexity. For at least one code vector, however, there are at least two non-zero encoding weights in accordance with the invention.

k, где "•" - это скалярное произведение.Alternatively, it may be possible to determine the code vector directly from the null spaces of the weighting matrices. For each user k, consider the weighting matrix W _k and define the associated null space and denote it by Ξ _k . For each user k, now it is necessary to form a code vector ω with such a property that it has a nonzero projection onto at least one of the vectors in Ξ _k . We know that if the code vector ω is chosen as one of the bases in Ξ _k , say, the first vector of the zero-space ξ _k , each must have the required non-zero projection for this user. However, it must be ensured that this is true for all, or at least for a given subset of users. One way to achieve this is to form a superposition over the entire first vector of the null space ξ _k into Ξ _{k of} the user in question. However, it must be prevented that any vector ξ _{k is} canceled by another ξ _k due to superposition. If this is not achieved, one can change the weighting coefficients for the superposition or choose one or more ξ _k as not the first vector in the corresponding zero-space Ξ _k . Therefore, the task is to find the code vector ω so that ξ _k • ω ≠ 0

k, where "•" is a scalar product.

As an alternative or addition, it is also possible to approach the idea from the point of view of optimization, since there are many different variants of the code vector ω that can be selected in each example. The optimization objective should preferably be to ensure that only a small number of encoding and decoding operations are required, i.e., in essence, make W _k as sparse as possible. Such optimization criteria for ω can be heuristic according to, but not only the following:

- Minimization of ω • 1, where 1 is a column vector of units or

- Minimization of the set of nonzero elements in ω, or,

- Minimization of ω • ω ^T , where T is the transposition of the vector matrix.

- это новый W_k, если пакет успешно принят пользователем k. Вычисление ранга нового

для следующего цикла может быть основано на форме Гауссового исключения

из предыдущего цикла с измененным предварительным кодовым вектором ω.On the side of the transmitting device, in order to keep the complexity as low as possible, for example, when determining the rank, the implementation of this algorithm can use the results and processed versions of the weighting matrixes from the previous calculation cycle. For example, when using the result of a Gaussian exception performed for the last cycles,

is the new W _k if the packet was successfully received by user k. New rank calculation

for the next cycle can be based on the form of a Gaussian exception

from the previous cycle with the modified preliminary code vector ω.

Similarly to the side of the transmitting device, the side of the receiving device can use the results and processed versions of the matrices of weighting coefficients from the previous reception of the data packet transmitted in the multicast mode. For example, the result from the previous Gaussian exception (s) may be used.

Encoding methods

An example encoding suitable for the present invention is preferably based on bitwise encoding according to the XOR scheme because of its simplicity. The XOR operation corresponds to coding weights of one, as previously explained.

Many different ways can be suggested in order to jointly XOR-encode and then identify which packets are co-encoded.

An example of possible code frame formats is provided with reference to the illustrations of FIG. 5. In this example, two data packets (A) 1005 and (B) 1010 are jointly encoded to form a composite multicast data packet 1015. Operational data can be XORed directly, but a means of identifying which individual packets are coded together must be provided. In the exemplary coding method of FIG. 5, identifiers (e.g., headers or subsets of relevant information from individual packet headers) of jointly coded packets participating (only two packets are shown here, but the idea can easily extend to more packets) are reported in the header 1020 of the composite package. The identifier, for example, may contain the address of the transmitting device, the multicast address and the sequence number of the packet for each encoded packet. In addition to identifiers, the header of the composite packet preferably also reports the format of the composite packet transmitted in multicast mode, i.e. where the packets are located in the composite multicast packet. For example, if there are two packets and one of the packets contains fewer bits than the other packet, as in FIG. 5, the number of bits contained in the packet with fewer bits, as well as the position of the first bit of the shorter packet is also usually indicated. When the number of bits is different, padding 1025 is used, as in FIG. 5 for packet B 1010. The header format field 1020 of the composite packet may also indicate that two packets B and C are concatenated one after another (not shown) and then encoded, for example, with the third package A or additional packages. The entire composite packet including the header is then preferably encoded according to FEC.

After correctly decoding the FEC and detecting the correct CRC 1030, the illustrated composite packet header provides a simple identification of which packets are encoded together. The receiver may then use this information to select a previously known packet from the decoded data packet repository and retrieve the other packet (s). It should be noted that the header of the composite package may also contain other information. A different version of the packet format (not shown), other than the one illustrated above, should report the header of the composite packet in a general broadcast message, i.e. in a certain type of out-of-band service signals, when encoding the operational data of several packets. Another additional coding method (not shown) may involve a blind identification approach, i.e. testing the hypothesis of encoded messages, where both the headers and the operational data of the packets are encoded together, in comparison with the database of a priori information and using CRC-testing in order to test the validity of testing the hypothesis.

The present invention is not limited to the use of XOR operations in encoding composite data packets. XOR operation is a linear summation over the field F (2) with elements {0, 1}. As mentioned above, the inventor has found that when the number of receivers increases, it becomes important to expand the field with additional elements to ensure that the rank for the matrix of each receiver can increase in each transmission. Therefore, the field can be expanded to 3 elements {0, 1, 2}, four elements {0, 1, 2, 3} and so on. In general, as is known in the general coding theory, fields must have p ^m elements, where p is a prime and m is a positive integer in order to be clearly defined. It is also known that fields of the same size are isomorphic in the sense that it does not matter if the field contains various elements, say, {0, 1, 2} or {2, 3, 7}. In addition to identifiers (for example, headers or subsets of relevant information from individual packet headers) of the jointly coded packets involved, the values of the weighting vectors used when the field is more than two elements, as explained above, are preferably also reported in the composite header 1020 package.

An additional example of an encoding operation suitable for use in the method according to the invention is based on a modulo operator that operates with signal constellations, i.e. after FEC coding and modulation. The coding for the signal constellation symbol is discussed later, and the procedure can be repeated for a plurality of consecutive symbols of the signal constellation. The modulo operation in this example is performed for the real and imaginary parts independently when processing complex numbers and uses the modulo operation definition and the following mathematical observation:

((A + B) mod L - B) mod L = (A) mod L ,

which indicates that signal B in the form of real values can be superimposed on signal A in the form of real values and provides undistorted restoration of the signal (as long as signal A does not exceed the quantization level L), although the amplitude (and therefore the power) is limited in a (non-linearly encoded) composite signal.

{-3,-1,1,3}+i{-3,-1,1,3}. Теперь, поскольку приемное устройство имеет сведения о последовательности данных D₂(n), оно также имеет сведения о соответствующем символе S₂ (для каждого S₁). Затем для действительной части (и в равной степени для мнимой части) совместно кодированный сигнал в передающем устройстве следующий:In practice, this can be used as follows. The transmitter has the symbols S ₁ and S ₂ , which typically have different meanings. For example, in 16 QAM, S _i

{-3, -1,1,3} + i {-3, -1,1,3}. Now, since the receiver has information about the data sequence D ₂ (n), it also has information about the corresponding symbol S ₂ (for each S ₁ ). Then, for the real part (and equally for the imaginary part), the jointly encoded signal in the transmitter is as follows:

,

,

which is then adopted and equalized, i.e. is compensated for losses in the transmission path (provided that one scale is used both for the received signal and for the signal that is subtracted) and the complex phase (provided that the corresponding in-phase axis and the axis of the quadrature phase are aligned with the signal that is subtracted), to result in a received signal:

where N ^(Re) is a member of noise (and interference). The desired signal is then restored using:

Coding can also be carried out by quantization using a more multidimensional lattice than only one-dimensional quantization, as described above. In this case, quantization operates on a vector, not a scalar.

Typically, the receiving node may not have the full information required to immediately fully decode and extract its own data from the composite data packet transmitted in the multicast mode, but must wait to receive more data packets. If this happens, the receiving node can partially decode the composite multicast data packet and store the result, the residual multicast multicast data packet for additional processing when additional information is available. This decoding, for each receiver, can be accomplished by calculating the reduced form of line levels in accordance with the current code matrix and implementing an updated reduced form of line levels for each received new data packet (regular or composite). Alternatively, the composite multicast data packet is stored without trying to decode until the receiving node has received all the required information. The information required to decode the main part of the constituent multicast data packet may, for example, be contained in the header. Also, this decoding can be performed by the approach of the reduced form of line levels. Each receiving device, in essence, is interested in inverting the code matrix (within the field that it uses) and multiplying it by the received data packets. To be precise, this must be done, with respect to the encoding operation, for each substring of bits into which the data packet is divided.

By using conventional and composite data packets and by allowing the transmitting node (via feedback information) and the receiving node to use this information, it is possible to reduce the number of transmissions required to transfer a certain amount of data from one node to another. This increases the aggregate throughput as well as the throughput of an individual user (or equivalently, a node). In addition, end-to-end delay performance is improved. Alternatively, depending on the conditions, a reduced number of transmissions may be used to increase power and energy efficiency in a communication system with a transmitter and multiple receivers.

It should be understood that the transmitting device can use and apply all available feedback information about received data packets. In addition to the feedback information explained above, additional information, for example, Quality of Service (QoS) information and data packet characteristics, can also be used in the formation of composite data packets.

6 is a schematic block diagram illustrating a system for multicast data packets from a transmitter to multiple users according to a preferred exemplary embodiment of the invention.

The modules and blocks described below, depending on the chosen implementation, can be considered as functional parts of the transmitting and / or receiving node in the communication system and not necessarily as physically distinctive objects. Connections, therefore, can be interpreted as communication lines between functional parts and not necessarily physical connections.

The complete system is based on the exchange of data by the transmitting device 100 with a certain number of users from 200-1 to 200-K through more or less error prone communication channels or communication lines.

A transmitting device, such as a transmitting unit 100, mainly comprises a data buffer module 110, an ACK register 120, a multicast scheduler 130, an encoder 140 of multicast packets, and some form of transmission medium that provides the necessary functionality to perform the actual group transmission.

The transmitting device may also comprise means for processing unicast information (not shown).

The data buffer module 110 temporarily stores data packets for subsequent scheduling / coding and transmission to users / receivers 200.

Typically, as previously explained, the transmitting node 100 first transmits one or more conventional data packets. As feedback information becomes available from the receivers, and the constituent packets transmitted in the multicast mode can be generated, the transmitting node can also form and transmit one or more constituent data packets transmitted in the multicast mode.

The multicast scheduler 130 takes into account the data packets that are constantly located in the data buffer module 110, and also takes into account the ACK information from the various receiving nodes stored in the ACK register 120.

Essentially, the multicast scheduler 130 schedules the usual data packets available in the data buffer module 110 by determining a set of encoding weights based on the feedback of the ACK in the ACK register 120. The multicast scheduler 130 informs the encoder 140 of the multicast data packets that generates or encodes a composite multicast data packet as a weighted linear combination of conventional data packets based on the selected weights odirovaniya. This so-called joint scheduling and coding is illustrated in detail in FIG. The multicast scheduler 130 has a module for determining / adapting weights so that a new linearly independent coding can be generated as required and when required. As indicated earlier, this encoding in a particular embodiment may be based on a set of encoding weights (this set of weights is also referred to as an encoding vector), which includes at least two non-zero encoding weights.

The multicast scheduler 130 is typically configured to process regular / traditional multicast ARQ operations.

Supporting information can also be used for planning. Additional information may include, but not limited to: QoS requirements for being able to service QoS scheduling aspects, as well as the status of individual packets, such as the lifetime value.

The receiving device, such as the receiving node 200, mainly comprises reception means that provide the necessary functionality for actual reception, a decoder 210 of multicast data packets with feedback functionality, and a data buffer module 220. A data buffer module is typically configured to store data packets that have not yet been delivered to k receive modules.

When the composite packet is received, the decoder 210 of the receiver 200 typically identifies which packets are encoded together. Using conventional multicast data packets or multicast multicast data packets, or both stored in a buffer module 220, the decoder attempts to decode a multicast multicast packet to retrieve individual ordinary multicast packets. Decoder 210 is also configured to provide and process feedback messages, including ARQ messages.

After the receiver 200 has successfully received the multicast data packet (a regular data packet or a composite data packet), it sends an ACK message to the transmitting apparatus 100 to confirm that it has received the packet immediately or with a slight delay. A delay can be useful so as not to take up resources and not uselessly waste energy due to overhead information of each feedback packet. The update is preferably achieved through the ARQ means of the receiving device and the ARQ means of the transmitting device.

In addition, note that the receiving device 200 can also act as a transmitting device if the data is then sent. The same applies to the transmitter 60, i.e. it can also act as a receiver for other data.

The above described embodiments are merely exemplary, and it should be understood that the present invention is not limited to them. Additional modifications, changes and enhancements that retain the fundamental principles disclosed and claimed herein are within the scope of the invention.

APPENDIX A

A modeling tool has been created on the basis of the obligatory algorithm for increasing the rank of each data packet transmitted in multicast mode for each user, which increases by one step in each iteration cycle.

The following final simulation results are weighting matrices for three users to whom five multicast data packets should be sent via iid channels with packet erasure, where the probability of reception is p = 0.1, and the weighting vector uses the base b = 3. Note that at most a separate line is added in each case to these matrices when they are created, and the matrices shown here represent the case of the first five received data packets transmitted in multicast mode. It is easy to confirm that each weight matrix has a full rank and that all coded data packets can be retrieved for each user.

Additional modeling with a different number of packets and the number of users has experimentally confirmed that a full rank can always be achieved provided that the base b of the weight vector is chosen large enough to serve a sufficient number of combinations that need to be checked.

It should be noted that the performance is approximately T ~ p when p ≠ 1 for all K, regardless of the number of packets. The invention provides optimal throughput for a fully reliable multicast, i.e. T = p when throughput is analyzed for an infinite number of packets.

The scheme of the invention is determinably decoded, i.e. unlike fountain coding (FC ), it is enough for it to receive exactly the same number of encoded packets in order to extract the same number of regular data packets. FC performance, determined by a factor of 1 + ε, depends on the number of packets that are sent. FC requires a large number of packets, of the order of 1000 or more, in order to ensure low ε. There is no such dependence in the proposed invention.

Information sources

1. J. W. Byers, M. Luby, M. Mitzenmacher, and A. Rege "A digital fountain approach to reliable distribution of bulk data. In Proceedings of ACM", SIGCOMM, pp. 56-67, 1998.

2. J. Byers, M. Luby, and M. Mitzenmacher, "A digital fountain approach to asynchronous reliable multicast", IEEE Journal on Selected Areas in Communications, 20 (8), October 2002.

3. J. Sachs, M. Meyer, S. Wager, H. Huschke, M.W. Larsson, P. Larsson, "Method and devices for efficient data transmission link control in mobile multicast communication systems". Patent Application (United States) 2006/0154603.

4. P. Larsson, N. Johansson, "Multi-User ARQ", In conference proceedings of VTC2006spring, Melbourne, May 7-10, 2006.

5. Shen Yong, Lee Bu Sung, “XOR retransmission in multicast error recovery”, (ICON 2000), Proceedings, IEEE International Conference on Networks, pp. 336-340, 2000.

6. M.A. Jolfaei, S.C. Martin, J. Mattfeldt, "A new efficient selective repeat protocol for point-to-multipoint communication", 1993. ICC 93. Geneva. Technical Program, Conference Record, IEEE International Conference on Communications, Volume 2, May 23-26, 1993, pp. 1113-1117, Volume 2.

7. M. A. Jolfaei, U. Quernheim, "Effective Block Recovery Schemes for ARQ Retransmission Strategies" IEEE Int. Symposium on Personal, Indoor and Mobile Radio Communications, 1994. Wireless Networks - Catching the Mobile Future. 5th volume, 3rd edition, September 18-23, 1994, pp. 781-785, volume 3.

Claims (29)

1. A method for scheduling and encoding data packets for transmission in a multicast session from a transmitting device to a multicast group of receiving devices, including at least two receiving devices in a multicast communication system with ARQ (automatic retransmission request) for transmitting a given amount of information transmitted in the multicast mode, in which, in each of the many planning events, the steps are performed in which:
- receive feedback information from the aforementioned at least two receiving devices indicating the received data packets;
- form at least one composite data packet as a weighted linear combination of conventional data packets based on the corresponding vector of weights containing a set of encoding weights;
- adapt for each of the said at least one composite data packet said coding weights of said weight vector based on said feedback information so that the composite packet is a linear combination of ordinary data packets that is linearly independent for each receiving devices from those receivers of said multicast group that still must receive one or more multicast transmissions Aci data packets from any of the data packet previously received by the receiver during said multicast session;
- transmit via said transmitter at least one composite data packet to those receivers of said multicast group, which still must receive one or more multicast data packets transmitted,
- wherein said coding weights in each of said plurality of planning events are adaptively selected so that for each receiving device, at least from the set of said receiving devices, the vector of weighting factors increases the rank of the matrix of weighting vectors specific for the receiving device for previously received packets data if the weight vector is added to the matrix. 2. The method according to claim 1, wherein said multicast session is set to transmit N ordinary data packets, and N linearly independent multicast data packets including said composite data packets are transmitted to said multicast group, according to of at least two receivers during said multicast session to ensure decodability of all N ordinary data packets. 3. The method according to claim 1, wherein said coding weights in a given scheduling event are adaptively selected such that the composite packet is a linear combination of linearly independent conventional data packets for each receiver, from at least a set of said receivers, from any conventional data packet and composite data packet previously correctly received by the receiver during said multicast session. 4. The method of claim 3, wherein the feedback information indicating the received data packets is based on at least one of the received conventional data packet and the received composite data packet. 5. The method of claim 4, wherein the feedback information is based on a history of the received data packets during said multicast session. 6. The method according to claim 1, wherein said step of generating at least one composite data packet as a weighted linear combination of conventional data packets based on a weighting vector containing a set of encoding weights comprises a step of multiplying ordinary packets by corresponding coding weights. 7. The method according to claim 6, in which conventional packets are divided into substrings, and said step of forming at least one composite data packet as a weighted linear combination of conventional data packets is performed at the substring level. 8. The method according to claim 1, wherein said composite data packet is formed from a portion of the operational data of conventional packets. 9. The method of claim 1, wherein each receiver-specific matrix of weight coefficient vectors for previously received data packets comprises coding weights, each coding weight

is a separate weighting coefficient associated with a conventional data packet n for the m _k th data packet transmitted in the multicast mode received by the receiver k, wherein the m _k th data packet transmitted in the multicast mode is a normal data packet or a composite data packet formed from at least two conventional data packets. 10. The method according to claim 1, wherein said weighting coefficients of the weighting vector encoding are adaptively selected by testing in each of said plurality of scheduling events preliminary weighting vectors until a coding weighting vector is found that increases the rank of said vector matrix weighting factors. 11. The method according to claim 1, wherein said weighting coefficients of the encoding of the vector ω of weighting factors are adaptively selected in each of said plurality of scheduling events by first computing a null space for the receiver k

in a rational form corresponding to the matrix specific for the receiving device, and the subsequent selection of any required vector ξ _k lying in zero-space

for which ω is adaptively chosen so that ξ _k · ω ≠ 0∀k, where "·" is the scalar product. 12. The method according to claim 1, wherein said weighting coefficients of the weighting vector encoding are adaptively selected so that the receiver-specific matrixes of weighting vectors are sparse so as to reduce the number of encoding and decoding operations. 13. The method according to claim 1, wherein the encoding weights of the composite data packet in the first scheduling event are adapted so that the composite packet is a linear combination that is linearly independent for each receiver from the first set of said receivers from any packet data previously received by the receiver during the multicast session and coding weights of another composite data packet in the second subsequent plan event The adaptations are adapted so that the composite packet is a linear combination that is linearly independent, for each receiver from the second smaller set of said receivers, from any data packet previously received by the receiver during said multicast session. 14. The method according to claim 1, in which the said at least two receiving devices contain more than two receiving devices, and the vector of weights of at least one composite data packet has at least two different non-zero encoding weights . 15. The method according to claim 1, wherein said step of generating at least one composite data packet comprises the step of generating several composite data packets based on the same feedback information, said several composite data packets being dependent on each other friend. 16. The method according to claim 1, wherein said adaptation step in this planning event comprises a step of adaptively selecting said encoding weights of said weights vector so that the composite packet is a linear combination of ordinary data packets that is linearly independent of any data packet previously received by each receiver in a set of receivers that must still receive at least one multicast th transmit data packet during said multicast session. 17. A means for scheduling and encoding data packets for transmission in a multicast session from a transmitter to a multicast group of receivers, including at least two receivers in an ARQ multicast communication system (automatic retransmission request) for transmitting a given amount of information transmitted in a multicast mode, wherein said means is configured to perform the following sequence of operations in each set of events:
- receiving feedback information from said at least two receiving devices indicating received data packets;
- the formation of at least one composite data packet as a weighted linear combination of conventional data packets based on the corresponding vector of weights, containing a set of encoding weights;
- adaptation for each of the at least one composite data packet, said coding weights of said weight vector based on said feedback information, so that the composite packet is a linear combination of ordinary data packets that is linearly independent for each a receiver of those receivers of said multicast group that still must receive one or more multicast transmissions giving data packets from any data packet previously received by the receiver during said multicast session;
- transmitting said at least one composite data packet to those receivers of said multicast group, which still must receive one or more multicast data packets transmitted,
- wherein said means is adapted to selectively select coding weighting coefficients in each of the plurality of events, so that for each receiving device, at least from the set of said receiving devices, the weighting vector increases the rank of the matrix of weighting vectors specific to the receiving device for previously received data packets, if the weight vector is added to the matrix. 18. The tool according to claim 17, wherein said multicast session is defined for transmitting N conventional data packets, and said means is configured to transmit N linearly independent multicast data packets including said composite data packets to said a multicast group of at least two receivers during said multicast session to ensure decodability of all N ordinary data packets. 19. The tool according to claim 17, wherein said tool is adapted to adaptively select encoding weights in a given scheduling event such that the composite packet is a linear combination of conventional data packets that is linearly independent for each receiver, at least from a set of said receiving devices, from any composite data packet and possibly a regular data packet previously received by the receiving device during said multicast session ni, and feedback information indicating received data packets is based at least on the received composite data packet and, possibly, the received conventional data packet. 20. The tool according to claim 17, wherein said tool is configured to form said at least one composite data packet as a weighted linear combination of conventional data packets by multiplying the conventional packets by respective coding weights. 21. The tool according to claim 17, wherein said tool is configured to generate said at least one composite data packet by dividing the usual packets into substrings and forming said weighted linear combination of the usual data packets at the substring level. 22. The tool according to claim 17, wherein said tool is configured to form said at least one composite data packet from a portion of the operational data of conventional packets. 23. The tool according to claim 17, wherein said tool is configured to test preliminary weighting vectors in each of the plurality of planning events until an encoding weighting vector is found that increases the rank of said weighting matrix of vectors. 24. The tool according to claim 17, wherein said means is adapted to selectively in each of the plurality of planning events of said coding weighting coefficients of the weighting vector ω by first computing a zero space for the receiver k

in a rational form corresponding to the matrix specific for the receiving device and the subsequent selection of any required vector ξ _k lying in zero-space

for which ω is adaptively chosen so that ξ _k · ω ≠ 0∀k, where "·" is the scalar product. 25. The tool according to 17, in which said transmitting device is a transmitting node, and said receiving devices are receiving nodes. 26. The tool according to 17, wherein said tool is implemented in said transmitting device. 27. The tool according to p. 26, in which the aforementioned transmitting device is a transmitting node containing a group transmission scheduling module for adapting said encoding weights, and an encoding module of a composite data packet for generating said at least one composite data packet. 28. The tool of claim 27, wherein said multicast scheduling module is actually configured to determine which conventional data packets to use in said at least one composite data packet. 29. A transmitting device for use in an ARQ multicast communication system (automatic retransmission request), said transmitting device comprising the means of claim 17.

Images