GB2384145A

GB2384145A - Adaptive signalling technique for repeat transmission requests

Info

Publication number: GB2384145A
Application number: GB0200672A
Authority: GB
Inventors: Bernard Kilbane; Marcello Caramma; Diego Giancola
Original assignee: Ubinetics Ltd
Current assignee: Aeroflex Cambridge Ltd
Priority date: 2002-01-12
Filing date: 2002-01-12
Publication date: 2003-07-16
Also published as: GB0200672D0; WO2003058872A1; AU2002358216A1

Abstract

The invention determines which of several possible signalling methods (BITMAP, LIST, RLIST) to use to request retransmission of data received in error, the determination being based on the relative efficiency of the signalling methods for the particular request. The relative efficiency may be based on the bandwidth or amount of processing required by each signalling method. The repeat requests take the form of STATUS reports comprising super-fields (SUFI) three of which are BITMAP, LIST and RLIST, with different super-fields conveying information in different ways. A linear programming technique is used to choose methods to suit different parts of a pattern of received errors (figure 1), thereby minimising the overall overhead. Application is to mobile terminals, base stations and radio network controllers (RNC) in a UMTS system.

Description

SIGNALING SELECTION TECHNIQUE FOR RADIO LINK CONTROLLER Field of the invention

This invention relates to apparatus and methods for signalling of errors to enable , retransmission, for use in a universal mobile telecommunications system (UMTS).

Background It is known to provide automatic retransmission requests (ARQ) in various telecommunication protocols. One such set of protocols, known as UMTS, is the European proposal for a third generation (3G) cellular network of base stations or radio network controllers (RNC). It is inter-operable with the existing GSM network, and is notable for providing high speed packet-switched data transmission. UMTS transmission protocols are defined in a series of standards developed by the Third Generation Protocol Partnership (3GPP). These standards define a number of layers for the transmissions, broadly in line with the well known ISO seven layer definition. There is a physical layer, 1.. 1, a data link layer, L2, and a network layer L3.

Layer 2 of this UMTS protocol stack contains a radio link layer (RLC) entity, which is responsible for ensuring reliable transfer of data over an unreliable medium i. e. the air. It is described in detail in the 3GPP standard 25.322 to which the reader is referred.

The RLC sublayer consists of RLC entities, of which there are three types: Transparent Mode (TM), Unacknowledged Mode (UM), and Acknowledged Mode (AM) RLC entities. Data and control information is transmitted by the RLC in packets that are referred to as protocol data units (PDU). Each PDU has a number associated with it known as its sequence number, which allows the PDU to be uniquely addressed. For duplex systems as in the embodiments described below, there will be transmitting and receiving RLC entities. In this document, references to "transmitted"are equivalent to"submitted to the lower layer" unless otherwise explicitly stated. Each RLC UM, and TM entity uses one logical channel to send or receive data PDUs. An AM RLC entity can be configured to use one or two logical channels to send or receive data and control PDUs.

The services provided to upper layers of the protocol, by the AM RLC entity to support acknowledged data transfer can be summarised as follows : Segmentation and reassembly.

Concatenation.

Padding.

f Transfer of user data.

Error correction.

In-sequence delivery of upper layer PDUs.

Duplicate detection.

Flow Control.

Protocol error detection and recovery.

Ciphering.

Service data unit (SDU) discard.

Of these, the transfer of user data is the service which is concerned with the signalling of errors to enable retransmission. To guarantee reliable transmission of user data, the RLC uses STATUS reports, which include information about what data has been received and what has been detected missing. The above referenced standard sets out that STATUS reports are comprised of super-fields (SUFI) three of which (BITMAP, LIST and RLIST) convey information about what data has not been received. The three SUFis report errors in different manners: LIST This SUFI is suited to reporting burst errors, as the information it provides indicates the sequence number of an initial PU that was not correctly received. Further to this it indicates the number of consecutive PDUs following this initial one that have also been received in error.

RLIST With the RLIST SUFI an initial sequence number is specified, this value is then proceeded by a length value. The initial sequence number indicates a PDU in error and the length indicates the distance between this initial PDU that is in error and the next PDU that is in error. RLIST can also operate in a similar fashion to the LIST SUFI with an initial sequence number and the number of consecutive PDUs in error.

BITMAP This SUFI indicates the status of PDUs by means of a bit fields where each bit, depending on its value, indicates data correctly or incorrectly received. The BITMAP

reports a sequence number, which the first bit in the BITMAP refers to i. e. the starting point. This SUFI differs from the other two in that it indicates data that has been received correctly also. This is due to the fact that information is conveyed by bit fields which are multiples of 8 bits. So unless there are eight consecutive errors

some of the bits will indicate that the PDUs they correspond to were correctly f received.

The standard does not specify when to use which of these SUFis. If one is used for all circumstances, this would simplify the implementation. It is an object of the invention to provide improved error signalling.

Summary of the invention According to a first aspect of the invention, there is provided a mobile terminal for use with a base station or RNC of a telecommunications system, the terminal having: a protocol for sending an acknowledgement to the base station or RNC for data received from the base station or RNC, the protocol having a number of different methods of signalling information about errors in the data, for sending to the base station or RNC to enable the base station or RNC to retransmit some of the data, the terminal being arranged to analyse the information to determine efficiencies of the signalling methods for that information, and select according to their efficiencies, one or more of the signalling methods to send the information to the base station or RNC.

An advantage of this invention is that it can significantly reduce traffic from the mobile terminal during a session where a guaranteed delivery is employed. Also, this will enable greater throughput of data to the terminal. This is because data sent or received by layer 2 is buffered and the data cannot be removed from the buffers until it is acknowledged. As the buffers start to fill up flow control will restrict the transmission window and stop the transmission of data. Throughput of the system will be increased, as more data can be acknowledged simultaneously and thereby the transmitter window will not be restrained so much.

Preferably, the efficiency includes the amount of bits to be sent, and/or the amount of processing. Preferably the information includes the size of the window to be

acknowledged, and/or the number of errors, and/or a pattern of the errors, such as how bursty it is. Preferably, the different signalling methods are concatenated using a linear programming technique. Alternatively, the concatenation is carried out using a search for byte aligned errors. The selection can be based on the instantaneous size of the error pattern to be encoded. Preferably the protocol is a UMTS L2 radio link controller.

According to other aspects, there is provided a corresponding base station or RNC, and corresponding methods and software.

According to another aspect of the invention, there is provided a method of determining which of a number of signalling methods to use to minimise an amount of the signalling, by using a linear programming technique, with zero and non zero metric cost function, to compute the minimal data size of the signalling message, based on the size of the overhead of each of the signalling methods. Other advantages will be apparent to those skilled in the art. Preferred features may be combined with other aspects of the invention.

I Embodiments will now be described by way of example with reference to the drawings in which :- Figure 1 shows a flow chart of an embodiment of the invention including an overall algorithm and a greedy algorithm.

Detailed Description The embodiments described show ARQ selection to adapt the selective repeat signalling strategy to the error pattern and retransmission window size. To signal the errors, there exist three possible methods (as described above) at any point in time, of which one option is always optimal. The selection can be continuously adapted, to ensure the optimal selection can be made. The signalling methods differ only in how they signal the information, so they allow certain types of errors to be signaled with the minimum amount of overhead.

The embodiments show how the error information being signalled is analysed. The characteristics of the data are determined and used to apply the optimum method of

signalling the errors so as to reduce the amount of control information transmitted and hence overhead required. The three methods above are not used exclusive of each other but rather are used together as the errors to be indicated can usually be classified according to more than one property. With the correct use of the SUFis

based on the pattern of the errors that have occurred the amount of data transmitted , to convey these errors can be greatly reduced.

The SUFis contain an inherent compression of the data needed, where for example the LIST SUFI can transmit information about 50 consecutive PDUs in 16 bits of data the BITMAP SUFI would need 68 bits to represent the same information. With the likelihood of having to report information on up to 4096 PDUs (in case of high data rates) it is obvious that considerable bandwidth can be saved by classifying the errors and applying the SUFis that produce the most economical result.

Below are the patterns to be searched for and the SUFis that best suit the needs of each pattern.

Burst errors are best handled by a List SUFI. If, though, an RUST is currently being created then the burst can more economically be accommodated into the RUST as the creation of a SUFI involves a data overhead in terms of simply creating the SUFI type.

Isolated errors spaced far apart will always be best handled by an RUST.

Errors that do not fall into the above two classifications, where there are errors in non consecutive PUs that cannot be cannot be classified as isolated from one another, are best handled by the BITMAP SUFI.

The different performance, in terms of transmitted data, between an optimal and suboptimal solution depends principally on two factors: * the size of the window to be acknowledged * the number of errors

In order not to waste MtPs (Millions of Instructions per second) (i. e. processing cycles, and therefore terminal battery power supply) the status report generated will use a single SUFI type for small packets (Bitmap Encoded) and for"almost

error free packets", List Encoding (RList if the Packets are small).

I In all the other cases the encoding can be either done with an optimal or greedy SUFI concatenation technique, each of which are described in more detail below.

Summary of overall algorithm 1. Is the data set free from errors 1.1. Yes. Therefore one super field can be used to signal information about the data set and no more unnecessary processing need be done.

1.2. No. Now check to see if the data set size if small 1.2. 1. Yes. Encode the data set with a bitmap, as any processing is unnecessary.

1.2. 2. No. Is It the error distribution bursty? 1.2. 2.1. Yes. Use Greedy algorithm described below 1.2. 2.2. No. Use Optimal algorithm described below

Optimal solution The problem can be stated as an optimization one. An aim is to use the combination of SUFis that describe the error pattern that uses the overall least number of bits.

The error pattern can be thought of as a binary sequence, where 1 s represent errors. Each bit will have to be encoded using one of the three SUFI types. The number of nibbles needed to encode one bit with one of the three SUFI types depends on the SUFI type (if any) used to encode the previous bit.

Let N be the total number of nibbles needed to encode the sequence and n, (a, la, 1, ai-2,... a1) the number of nibbles needed to encode the i-th bit with a SUFI of type al given the SUFis used to encode the previous bits, (al-1, ai-2,..., a1)' It can Npdus be seen that N = #ni(ai#Ai-1,ai-2,...,a1). Fortunately it is not necessary to

1=l consider all previous bits, but it is preferable to keep track of a "state" SI-1 for the bit before the one it is desired to encode. The equation can be rewritten as

Npdus N = yn, (a, I ,-))- The optimal solution is given by the set having the lowest ; =i value of N.

Note that the technique can be reused if different SUFis are introduced and/or the number of bits needed to encode each SUFI is modified. In that case it is necessary to modify the state transition table, either introducing new states and/or modifying the states'transitions costs.

This is a linear programming problem, whose solution can be found using the , Viterbi algorithm. In order to apply the Viterbi algorithm it is necessary to specify the number of states, the state transition table, the cost associated with each transition, and initial and final states.

States 'The initial and final state is the"IDLE"state.

'The"bitmap"SUFI is split into 8 states. The reason is that each field in a"bitmap"SUFI conveys information about 8 consecutive PDUs. They will be denoted as"BMO","BMI", ...,"BM7".

* The "1ist" SUFI is split into 16 states. Each"list"SUFI field allows to specify a maximum of 16 consecutive PDU errors, so 16 states are needed. They will be denoted as"LO", "L1",...."L15".

* The"relative list"SUFI is split into 8 states, because of the need to increase the number of bits to represent the relative distance between PDU errors each time a multiple of 8 is exceeded. The states will be denoted as"RLO","RL1",...,"RL8".

State transition diagram and associated costs The following table shows the transitions between one state and the other and the associated cost. The value"INF"denotes forbidden or redundant transitions.

The SUFI used to encode bit i will have the same name (without the numerical suffix) as the state reached.

RLC metrics update 8M = Bit Map L= List RL= st Metrics cost update for an error (1) In nibbles

None BMO BM1... BM7 LO LI... US RLO RL1... RL7 Previous state s (i-1) : Idle INF INF INF INF INF INF INF INF INF INF INF INF INF BM 0 7 INF INF fNF 2 INF INF INF 7fNF INF INF 7 BM 1 NF 0 INF INF INF INF INF INF INF INF INF INF INF ..., INF INF 0 INF INF INF INF INF INF INF INF INF INF BM 7 INF INF INF 0 INF INF INF INF INF INF INF INF INF LO 6 INF INF INF 6 INF INF INF 2 INF INF INF 6 L1 INF INF INF INF INF 0 INF INF INF INF INF INF INF ... INF INF INF INF INF INF 0 INF INF INF INF INF INF L15 INF INF INF INF INF INF INF 0 INF INF INF INF INF RLO 6 INF INF INF 6 INF INF INF 6 1 1 1 1 RL1 INF INF INF INF INF INF NF INF INF INF INF INF INF ... INF INF INF INF INF INF INF INF INF INF INF INF INF RL7 INF INF INF INF INF INF INF INF INF INF INF INF INF

Next state s (i) Metrics cost update for a non error (0) in nibbles

None BMO BM1... BM7 LO LI... L15 RLO RLI... RL7 Previous state 5 (1-1)' Idle 0 INF INF INF 0 0 0 0 0 INF INF INF 0 BMO INF INF INF INF INF INF INF INF INF INF INF INF INF BMi NF OJNF INF INF INF INF INF INF INF INF INF INF ... INF INF 0 INF INF INF INF INF INF INF INF INF INF BM7 INF INF INF 0 INF INF INF INF INF INF INF INF INF LO INF INF INF INF INF INF INF INF INF INF INF INF INF Li INF INF INF INF INF INF INF INF INF INF INF INF INF INF INF INF INF INF INF INF INF INF INF INF INF INF L15 INF INF INF INF INF INF INF INF INF INF INF INF INF RLO INF INF INF INF INF INF INF INF INF INF INF INF 1 RL1 INF INF INF INF INF INF INF INF INF 0 INF INF INF ... INF INF INF INF INF INF INF INF INF INF 0 INF INF RL7 INF INF INF NF INF INF INF INF INF INF INF 0 INF

Next state s (i) Note that redundant transitions have been assigned a metric of inifinite and represent non allowed transitions.

Table 1

Once the states'sequence with the least number of nibbles N used is found with the Viterbi algorithm, the error sequence can be encoded.

Greedy solution The idea is that the encoding of the list of errors is searched only for"byte aligned solutions"where the boundary of the byte is the first error in the error subsequences.

The control flow is shown in the picture below.

A list based method is used when the distance between consecutive errors is larger than a predefined number of bit octets : Ni, (e. g. N1= 4 x (8 bits).

A bitmap encoding is used for error sub-sequences where the distance in bit octets between consecutive errors is less or equal than N2 (e. g. N2=4 (x8) bits).

Greedy solution"pseudo code" 1. Firstly use the ACK super field to acknowledge all the data in the data set. This will acknowledge positively any data that isn't indicated in error by some other super field. In this way only indicating errors is considered.

2. Find and record the positions of the errors with the minimum and maximum sequence numbers.

3. Reset, to zero, the value of a counter variable,"Bytefree".

4. Set variab1e "currentydu" to the lowest erroneous sequence number.

5. Clear the value, set to zero, of the variable"pducounter".

6. Has the last error in the window been reached? 6.1. No.

Increment the value of currentpdu. As this will be used to sequentially step 6.1. 1. through all the relevant PDUs.

6.1. 2. Increment the value of pducountcr. This counts the number of PDUs encountered since the last error.

6.1. 3. Have eight PDUs without error been counted? 6.1. 3.1. Yes.

6.1. 3.1. 1. Increment byte free counter. This variable tracks the amount of distance in bytes between consecutive errors. It will be used to make a decision on how to encode the errors.

6.1. 4. Is currentpdu in error? 6.1. 4.1. Yes. Now a decision will be made on how to encode this error.

6.1. 4.1. 1. Is the byte free greater than the decision value N.

6.1. 4.1. 1.1. Yes.

6.1. 4.1. 1.1. 1. Encode the error with a List. This is because there is a large distance between the errors.

6.1. 4.1. 1.1. 2. Reset byte free, as this has to count the distance between different errors.

6.1. 4.1. 1.1. 3. N will changed to a new value that reflects the overhead involved in now signalling errors with a method other than list.

6.1. 4.1. 1. 1.4. Go to section 6 6.1. 4.1. 1.2. No.

6.1. 4.1. 1.2. 1. Encode the error with a bitmap. This is because the distance between the last two errors has been small.

6.1. 4.1. 1.2. 2. Reset bytefree, as this has to count the distance between different errors.

6.1. 4.1. 1.2. 3. N will changed to a new value that reflects the overhead involved in now signalling errors with a method other than bitmap.

6.1. 4.1. 1.2. 4. Go to section 6 6.1. 4.2. No. Go to section 6.

6.2. Yes. Merge 6.2. 1. Merge the decisions made on the super fields depending on the differential data size.

6.2. 2. Post process the decisions and encode to a super field.

Claims

Claims 1.'A mobile terminal for use with a base station or RNC of a telecommunications system, the terminal having : I a protocol for sending an acknowledgement to the base station or RNC for data received from the base station or RNC, the protocol having a number of different methods of signalling information about errors in the data, for sending to the base station or RNC to enable the base station or RNC to retransmit some of the data, the terminal being arranged to analyse the information to determine an efficiency of the signalling methods for that information, and select according to their efficiencies, one or more of the signalling methods to send the information to the base station or RNC.
2. The mobile terminal of claim 1, the efficiency being in terms of the amount of bits to be sent.
3. The mobile terminal of claim 1 or claim 2, the efficiency being in terms of the amount of processing.
4. The mobile terminal of any preceding claim, the information including the size of the window to be acknowledged.
5. The mobile terminal of any preceding claim, the information including the number of errors.
6. The mobile terminal of any preceding claim, the information including a pattern of the errors.
7. The mobile terminal of any preceding claim, and arranged to select a combination of the different signalling methods using a linear programming technique.
8. The mobile terminal of any preceding claim, and arranged to select the different signalling methods using a search for byte aligned errors.

<Desc/Clms Page number 11>
9. The mobile terminal of any preceding claim, the terminal having a UMTS L2 radio link controller.
10. A base station or RNC for use with mobile terminals of a telecommunications system, the base station or RNC having: a protocol for sending an acknowledgement to the mobile terminal for data received from the mobile terminal, the protocol having a number of different methods of signalling information about errors in the data, for sending to the mobile terminal to enable the mobile terminal to retransmit some of the data, the base station or RNC being arranged to analyse the information to determine an efficiency of the signalling methods for that information, and select according to their efficiencies, one or more of the signalling methods to send the information to the mobile terminal.
11. The base station or RNC of claim 10, the information including the size of the window to be acknowledged.
12. The base station or RNC of claim 10 or claim 11, the information including the number of errors.
13. The base station or RNC of claim 10 or any claim when dependent on claim 10, the information including a pattern of the errors.
14. The base station or RNC of claim 10 or any claim when dependent on claim 10, the terminal having a UMTS L2 radio link controller.
15. A method of offering a telecommunications service to a subscriber over a telephone network having the base station or RNC of claim 10 or any claim when dependent on claim 10, and the mobile terminal of claim 1, or any claim when dependent on claim 1, the method having the step of making the base station or RNC available to pass subscriber data between the base station or RNC and the mobile terminal.

<Desc/Clms Page number 12>
16. A method of using the mobile terminal of claim 1, or any claim when dependent on claim 1, to pass subscriber data between the mobile terminal and a base station or RNC.
17. A method of communicating between a mobile terminal and a base station or I RNC of a telecommunications system, the method having the steps of: using a protocol in the mobile terminal for sending an acknowledgement to the base station or RNC for data received from the base station or RNC, the protocol having a number of different methods of signalling information about errors in the data, for sending to the base station or RNC to enable the base station or RNC to retransmit some of the data, analysing in the terminal, the information to determine an efficiency of the signalling methods for that information, selecting according to their efficiencies, one or more of the signalling methods, and sending the information to the base station or RNC using the selected method or methods.
18. A method of determining which of a number of signalling methods to use to request retransmission of data received with a pattern of errors, the methods having different overheads, and different ways of coding the errors, the method having the step of: using a linear programming technique, with zero and non zero metric cost functions, to choose methods to suit different parts of the pattern, to minimise an overall amount of the signalling, based on the different overheads and coding method of each of the signalling methods, and different arrangements of errors in the pattern.
19. The method of claim 18 having the preliminary step of specifying a number of states, a state transition table, a cost associated with each transition, and initial and final states.