GB2348583A - Automatic re-transmission of data - Google Patents

Abstract

The integrated channel access mediation and automatic repeat query (ARQ) mechanism is provided wherein an acknowledgement array is generated at a receive terminal corresponding to the successful receipt of data packets from a sender terminal. The position of the flags in the array correspond precisely to the order in which the data packets were sent, and hence the position of each flag conveys packet identifier information, thus, reducing network overheads in sending identifier information explicitly. This has the primary advantage that spectral efficiency is increased.

Description

2348583 A Method and System fQr Autmatic Re-Transmission of Data The

present invention relates to the automatic re-transmission of data which has been received incorrectly at a receiver terminal.

More particularly, the present invention introduces a method and system for the automatic re-transmission of data packets which have been transmitted onto a point-to-multipoint network without the use of Forward ErTor Correction (FEC) and m a bandwidth efficient manner, and wherein the nework supports a wide variety of traffic types.

Automatic Repeat Query (ARQ) is well known m the field of networked data communications, providing for the automatic re- transTrUssion of data packets which have been transmitted to an intended receiver terminal, but somehow corrupted or otherwise damaged during transmission and receipt so as to have been received in error. Various error-checking codes which can be added to data packets and which can identify the erroneous receipt of a particular data packet are well known in the art.

The network channels over which data may be sent can by nature be unreliable, resulting in bit errors occurring within data packets. Bit errors may be detected and corrected by the use of forward error correction, but in some applications this is not appropriate,_ for example because the majority of terminals m a network operate on low bit error rate channels making the extra FEC overhead unnecessary. Selective repeat automatic repeat query is a more efficient manner of correcting the types of bit errors that may occur. Selective Repeat ARQ is a form of ARQ wherein only those packets received in error are re- transmitted.

Typical prior art implementations of ARQ locate it outside the mechanism that grants channel bandwidth reservations. Typically, networks are designed in a layered manner, with automatic repeat query implemented in a Data link controller that is separate from the media access controller. A system designed in this manner could potentially cause bandwidth reservation requests to be 2 transmitted unnecessarily when a re-transmission is required, lengthening data latency and wasting precious on-air bandwidth. In a data network supporting a wide variety of traffic types ranging from available bit rate (ABR), with no restrictions on data latency, to continuous bit rate (CBR), which requires a fixed maximum latency of the data through the network, such reservation request delays for re-transmission are unacceptable with respect to complying with quality-of service (QOS) requirements.

Moreover, typical prior art implementations of selective repeat,_ARQ generally require the transmission of a packet identifier from the receiver terminal to the sender terminal in order to identify which of the packets were received in error, and hence require re-transmission. This transmission represents a network overhead and hence reduces overall network efficiency at transporting actual payload data traffic.

The present invention described herein improves upon the pirior art arrangements by integrating the channel access mediation, and automatic repeat query (ARQ) all into a single network layer, providing for fast selective repeat ARQ of data traffic while still supporting traffic latency requirements.

According to a first aspect of the present invention, there is provided a method of performing integrated data traffic scheduling and automatic retransmission of erroneous data received at a receiver terminal, comprising the steps of - a) scheduling one or more data portions waiting to be transmitted from a transmitter terminal to the receiver terminal by queuing said data portions in order in a traffic 25 queue, each of said data portions including an error-check code; b) transmitting one or more of said data portions to said receiver terminal in the same order in which said data portions were queued in said traffic queue; 1 3 C) detecting the error-check code of each transmitted data portion at the receiver terminal to determine whether each transmitted data portion has been received either correctly or erroneously; d) generating an acknowledgement array containing one or more acknowledgement flags, each acknowledgement flag indicating either the correct or the erroneous receipt of a particular corresponding transmitted data portion; e) transmitting said acknowledgement array from the receiver terminal to the transmitter terminal; and f) removing from the traffic queue those data portions indicated in the array as having been cor-rectly received by the receiver terminal; wherein steps a) to f) are continuously repeated in turn such that retransmission of "those data portions that have been received erroneously occurs automatically due to those data portions remaining in the traffic queue after step f).

The acknowledgement flags may be arranged in the array III the same predetermined order in which the respective corresponding data portions were transnUtted at step b), whereby the transmitter terminal may determine from the position of the acknowledgement flags in the acknowledgement array which of the data portions may be-r-emoved from the traffic queue at step f).

Furthermore, a particular data portion remains in said traffic queue and is repeatedly retransmitted -until a positive acknowledgement indicating correct receipt of said particular data portion is received from said receiver terminal.

The method of the present invention may be used in a point-to multipoint network compns-m'-g a central control node and one or more remote subscriber nodes, wherein the central control node is always either one or other of the receiver terminal or the transmitter terminal.

4 The central control node can control all access of each of the subscriber nodes to the network channel.

The present invention also provides a system for perfon-ning integrated data traffic scheduling and automatic retransmission of erroneous data received at a receiver terminal, comprising:- a) scheduler means for scheduling one or more data portions waiting to be transmitted from a transmitter terminal to the receiver terminal by queuing said data portions in ordr in a traffic queue, each of said data portions including an error check code; b) means for transmitting one or more of said data portions to said receiver terminal in the same order in which said data portions were queued in said traffic queue; C) effor-detection means for detecting the error-check code of each transmitted data portion at the receiver terminal to determine whether each transmitted data portion has been received either correctly or erroneously; d) acknowledgement generation means for generating an acknowledgement array containing one or more acknowledgement flags, each acknowledgement flag indicating either the correct or the erroneous receipt of a particular corresponding transmitted data portion; e) means for transmitting said acknowledgement array from the receiver terminal to the transmitter terminal; and f) removal means for removing from the traffic queue those data portions indicated in the array as having been correctly received by the receiver terminal; wherein said above-mentioned means a) to f) each repeat their operations in turn whereby retransmission of those data portions that have been 1 received erroneously occurs automatically due to those data portions remaining in the traffic queue after the operation of the removal means.

It is an advantage of the present invention that as the ARQ mechanism is integrated with the channel access mechanism, there need be no further contention for bandwidth for sending an acknowledgement or a repeat request, thereby improving overall bandwidth efficiency. Furthermore, as a result of this integration, the delay for receiving feedback from the far end of the link is always deterministic.

The present invention has another advantage in that it allows the access network to provide error correction without unnecessarily sending forward error correction overhead to each subscriber terminal.

There is a flirther advantage in that it allows a fast turnaround ARQ mechanism for support of fixed bit rate, low latency traffic, by integrating the ARQ mechanism with the channel access mechanism.

It is a feature of the present invention that position of data within a portion conveys information. More specifically, acknowledgements of the receipt of data portions are ordered the same as the portions themselves were transmitted; therefore the acknowledgement does not require an additional identifier to be transmitted over the air. This further increases bandwidth efficiency.

Further features and advantages of the present invention will become apparent from the following description of a particularly preferred embodiment thereof, and in particular by reference to the accompanying drawings, in which:

Figure 1 shows- the overall frame structure of the data signal used in the present invention; Figure 2 shows the frame structure of the control data portion of the data signal used in the present invention; Figure 3 shows the frame structure of a single reservation request acknowledgement cell; - - Figure 4 shows the frame structure of a single downstream 6 acknowledgement cell; Figure 5 illustrates the structure of a single downstream payload cell; Figure 6 shows the structure of a single subscriber reservation request cell of the data signal; Figure 7 shows the frame structure of a single upstream acknowledgement cell; Figure 8 illustrates the frame structure of an upstream payload cell with reservation request; Figure 9 illustrates the frame structure of a single upstream payload cell without reservation request; Figure 10 illustrates the data structures used in the present invention; and Figure 11 shows a demonstrafion of the method of the present invention when transmitting multiple payload cells upstream to the central access is point from a remote subscriber terminal.

The method and system of the present invention are chiefly although not exclusively for use within a wireless access network, and the invention may also find application within wired networks. In the present embodiment to be described herein, however, the invention is employed in a wireless network deployed m a cellular configuration. Each cell of the network consists of a central access point and multiple subscriber units. Subscriber units communicate to the network only through the access points, making each cell of the network a point multi point architecture. The access point is the centre of all wireless network communication for the particular cell, and thus is the locus of control of access to the wireless medium for the cell.

Communication can occur on the wireless medium in both directions, and hence a means of duplexing the wireless medium is required. Two common methods are frequency division duplexing (FDD) and time division duplexing (TDD). In the case of FDD, the medium is broken into a downstream 1 7 (data originating from the access point) frequency band and an upstream frequency band (data originating at the subscriber unit). TDD breaks a single frequency band into downstream time slots and upstream time slots. The present invention uses TDD.

For the purpose of understanding the present invention, it is useful to View the network of the present embodiment as consisting of multiple switches.

On one end, corresponding to the access point, there is a switch with a single physical wired data port, and multiple wireless data ports. Disbursed throughout the cell are two port switches, each located at a subscriber terminal. Each subscriber terminal has a single wireless port and a single physical wired port During the subscriber terminal initial registration process, a subscriber terminal negotiates With an access point to be assigned a temporary port identifier, referred to as a subscriber unit access identification (SU- AID). Once a subscriber terminal has been granted an SU-AID, it is capable of proceeding with higher layer signalling to gain access to the network.

Access of the subscriber terminals to the wireless medium is controlled through central control of the subscriber terminals by the access point.

In order to achieve this the access point is provided with a medium access controller Q which administers the medium control. Similarly, each remote subscriber terminal is also provided with a compatible medium access controller for responding to the central MAC in a master-slave manner: the subscriber terminals request access to the medium and the access point has the ability to grant access or fail to grant access based on the current level of network utilisation.

Access to the network is granted m the form of time slots - when a subscriber terminal is granted the ability to access the wireless network medium, permission is granted in the form of one or-m- ore time slots in which the subscriber terminal can transmit. Within the granted time slot the entire medium capacity is available to the subscriber terminal to transmit its payload data. By referring to a medium access controller, it is to be understood that either a hardware or software based 8 control means is envisaged and that reference to a controller as such implicitly includes reference to those control means required at both the central access point and at the subscriber terminals. In this respect, the medium access controller (MAC) therefore corresponds to those network means, whether hardware or software based, that would approximate to the Network-level and the Data- level of the ISO Open Systems Interconnection 7-layer Reference Model. As an example, a suitable hardware implementation can be achieved using a Field

Programmable Gate Array (FPGA). j The MAC operates by controlling transmissions on the medium by the definition of a MAC frame, being the framework in which data transmissions take place. In order to fully understand the various features and advantages of the present invention, it is necessary to first describe the constituent parts of a MAC frame, followed by a description of the various data structures used in the MAC.

This will be performed by reference to Figures 1 through 9.

Figure 1 shows the overall structure of a single MAC frame. The MAC frame consists of a downstream portion, generated by the access point and broadcast to all subscriber terminals, and an upstream portion, which consists of a contention interval and all data bursts being sent from subscriber terminals back to the access point.

The downstream portion consists first of a downstream preamble (2).

The downstream preamble is a Physical layer synchronization sequence of fixed length, used for frame acquisition and channel estimation. Only one downstream preamble may occur per MAC frame. Immediately following the preamble is the frame descriptor header (FDfIDR) (4). The FDHDR describes the complete contents of the remainder of the MAC frame. The size of the FDHDR may vary.

The I'DHDR contains a map of all traffic, upstream and downstream, to occur within the MAC frame. After achieving bit synchronisation on the MAC frame via the preamble, subscriber terminals demodulate the FDHDR and from that gain complete knowledge of the traffic that will occur within the remainder of the 1 9 frame. Only one FDHDR may occur per MAC frame. The precise contents of the FDHDR are shown in Figure 2 and described in detail in Table I below.

Field Tag Description

SYNC Short 4 symbol sync burst.

BD_cnt Bursts Downstream Count. Number of subscriber units having payload data sent to them in this MAC firame BU-cnt Bursts Upstream Count. Number of subscriber units that will be sending payload data in this MAC frame.

AP-ID Access Pomt ID. Identifies the access point that originated the frame descriptor header.

RRA_cnt Reservation Request Acknowledgment Count. Number of acknowledgments being sent in response previous requests.

DA_cnt Downstream Acknowledgment Count Number of upstream cell acknowledgements being sent downstream in this MAC frame.

Downstream Identifies the subscriber unit being sent cells, the number of cells to Map be sent, and the traffic type being sent.

RR_cnt Reservation Request count. Total number of reservation request slots that will be made available in this MAC frame.

UA_cnt Upstream Acknowledgment count. Total number of downstream cell acknowledgments being sent upstream in this MAC frame.

Upstream Identifies the -subscriber units that have been granted reservations, Map the number of cells to be sent by each, and the traffic type allowed.

CRC Cyclic Redundancy Check. Allows each subscriber terminal to verify corTect receipt of the frame descriptor.

SU-ID, Subscriber Unit ID. Identifies the subscriber unit acting as the data source or sink in the burst.

Cell-Cnt Cell Count. Total number of ATM cells to be sent in this particular burst.

_Field Tag Description

Tr_type Traffic Type. Defines the type of traffic that the subscriber unit is allowed to send or will be receiving during the current frame.

Table 1: Frame Descriptor Header FDHDRILS-tructur Following the FDHDR is the reservation request acknowledgement (RRA) portion 6. The RRA contains acknowledgements of requests by subscriber terminals for upstream time slots and can also communicate signal propagation delay. There is a single RRA for each reservation request that was made during the contention interval from the previous MAC frame, although in the case where no reservation requests were made in the previous MAC frame, then no acknowledgements will be sent. The precise contents of the RRA are shown in Figure 3 and described in detail M Table 2 below.

_Field Tag Description

Sync 8 bit framing synchronization sequence SU-ID ID of the subscriber unit that originated the reservation request, and to which the reservation request acknowledgment is directed.

is RTRN Return Code. Communicates reservation status to SUs and SU_AID status to SUs Performing registration.

DELAY Delay compensation bits. These bits are assigned during subscriber unit registration and cause a shift in subscriber unit timing.

CRC Cyclic Redundancy Code. Used by the subscriber unit to verify that the frame has been received error free.

Table 2: Reservation Request Acknowl ement (RRA) Structure -Ce M Following the RRA comes the Downstream Acknowledgement (DACK) portion 8 containing DACK cells. Each DACK cell contains a downstream ack or nack of a single upstream burst from a previous MAC frame.

There is a single DACK cell for each upstream burst from the previous MAC frame, although in the event that there were no previous upstream bursts then no DACKs will be sent. As the DACKs are always located m the same place in the MAC frame, there need be no separate identifier symbols. There also need not be any packet sequence numbers accompanying the acks/nacks, thus improving overall bandwidth efficiency. The precise contents of a DACK cell are shown in Figure 4 and described m detail in Table 3 below.

Field Tag Description

Sync 4 symbol synchronization burst SU-1D ID of the subscriber unit that originated the cells being acknowledged.

UU Unused AckINack One acknowledgment bit per cell. 1 = successful cell receipt.

map CRC Cyclic Redundancy Code. Used to verify that the downstream acknowledgment has been received error free.

Table 3: Downstream Acknowledgement (DACK) cell structure.

Following the DACK portion comes the Downstream Burst (9). The MAC operates on a principle of cell bursts for commUrUcating payload data between the access point and the subscriber terminals by allowing multiple cells of data to be sent to or from a particular subscriber unit at a time. A burst must always consist of at least one cell. In upstream bursts, this single cell must be an upstream cell with reservation request (UCELLR) (18). Additional cells M the upstream burst are in the-fo-rma of a UCELL - an upstream cell without reservation request (20). Upstream cells are discussed m more detail later.

Downstrearn bursts can also- consist of multiple cells, but there is only one type - the downstream cell (WELL) 10. There can be many DCELLs either several directed to a single subscriber te rminal, or several directed to several subscriber terminals. Each DCELL contains one ATM-compatible cell of payload data.

12 Currently the MAC allows bursts to have a maximum size of six cells, although more or less cells may be designated per burst if required in a future implementation without departing from the scope of the present invention. The structure of a DCELL is shown in more detail in Figure 5, and described in Table 4 below.

Field Tag Description

Sync 4 symbol synchronization burst SU-1D ID of the subscriber unit to which the payload data is difected.

SEQ Sequence number. Used by the MAC to re-sequence cells that get out of sequence due to cell loss and subsequent cell repetition.

Condensed Includes Virtual Path Identifier (VPI), Virtual Channel Identifier ATM Header (VCI), Traffic Type, Cell Loss Priority.

Payload Payload data.

CRC Cyclic redundancy code. Used to verify correct receipt of the downstream cell.

Table 4: Downstream Cell (I)CELL) Structure The downstream burst concludes the downstream portion transmitted by the access point and received at all subscriber terminals. There then follows a slight delay due to subscriber turnaround time (STT) 12. The STT varies with distance to the farthest subscriber unit. A typical maximum distance to a subscriber unit could be, for example, 51an, although this obviously depends on the network configuration and the size of each network cell.

Following the STT comes the Upstream Portion of the MAC frame, being data transmitted from the subscriber units to the access point. The entire expected structure of the upstream portion has already been communicated to each and every subscriber terminal in the FI?HDR transmitted in the downstream portion. Therefore, each subscriber terminal knows whether or not it is permitted to transmit in the upstream portion, what data it is to transmit, and when it is to 13 transmit this data. In this way absolute control of the contents of the upstream portion can be controlled by the access point. With such a mechanism, however, it becomes necessary to define a period in which subscriber terminals can first communicate a request for transmission permission to the access point, without which no subsequent penrussion would ever be granted. This period forrns the first part of the upstream portion, being the subscriber reservation request (SRR) portion 14.

The SRR is a contention based reservation request interval.. If a subscriber terminal has been sitting idle with empty data queues, the arrival of a burst of data on its physical port will force it to request a time slot reservation from the access point. Because the subscriber terminal has no active reservations, and because it is believed that at any given time the number of terminals making initial bandwidth requests Will be small, it is reasonable to force the subscriber terminals to contend for reservations. This contention window is kept as small as possible while still allowing reasonable success probability by employing a novel implementation of aloha contention control schemes. Once the subscriber terminal's reservation request has been acknowledged by the access point, the subscriber terminal ceases requesting bandwidth in the contention slots, allowing other terminals access to the contention mterval. The number of SRR's that may occur in one MAC frame is communicated to the subscriber terminals in the FDHDR. Multiple slots can be made available during times of heavy request traffic. Furthermore, the start of the contention interval can be calculated by the subscriber terminals by virtue of the FDHDR indicating to each terminal the number of RRAs, DACKs and the structure of the downstream burst M the subsequent downstream portion of the MAC frame. The contention interval then begins immediately after the end of the downstream burst, allowing for the STT.

The structure of a single SRR to be transmitted during the contention interval is shown in Figure 6, and de-sc-nbed-in detail in Table 5 below.

14 Field Tags Description

Preamble Physical layer synchronization sequence.

Sync 8 bit MAC framing synchronization burst SU-1D ID of the subscriber unit requesting a reservation.

Cells Number of cell time slots being requested by the subscriber unit.

Tr_Type Traffic type of the data for which time slots are being requested.

CRC Cyclic redundancy code. Used to verify correct receipt of this upstream burst.

Table 5: Subscriber Reservation Reauest (SM structure Following the contention interval comes the upstream acknowledgement portion 16, containing upstream acknowledgement (UACK) cells of each downstream burst received during the downstream po. rtion. Each UACK indicates upstream ack or nack of a single downstream burst-from a previous MAC frame. As many UACKs may be transmitted in each upstream acknowledgement portion as there were downstream bursts in the downstream portion. As with DACKs, as the UACKs are always located in the same place in the MAC frame, there need be no separate identifier symbols. There also need not be any packet sequence numbers accompanying the acks/nacks, thus improving overall bandwidth efficiency The structure of each UACK cell is shown in Figure 7, and described in detail below in Table 6.

Field Tag Description

Sync 4 symbol synchronization burst SU-1D ID of the subscriber unit acknowledging receipt of downstream cells.

UU Unused Ack/Nack One acknowledgment bit per cell. 1 = successful cell receipt.

Map 1 Field Tag Description CRC Cyclic Redundancy Code. Used to verify that the downstream acknowledgment has been received error free.

Table 6: Ilpstream Acknowledgement (UACK) structure Following the upstream acknowledgement portion comes the upstream burst portion 22, containing cell bursts from subscriber units which were granted permission in the FDHDR to transmit payload data to the access point.

The FDHDR from the downstream portion contains the instructions to the subscriber terminals on when to transmit a burst in the upstream burst portion, and what the burst is expected to contain. Each upstream burst contains one or more data cells with the same traffic type being sent from a particular subscriber terminal. Each upstream burst made in the upstream burst portion may be from a different subscriber unit, or alternatively may be from the same subscriber unit, depending upon the channel allocations granted to the subscriber units. In this way channel allocations can be dynamically arranged between the subscriber terrminals from MAC frame to MAC frame, depending on the network traffic loading and the traffic priority. As mentioned -earlier, each upstream burst must contain a sing e upstream cell with reservation request (UCELLR) 18, and zero or more upstream cells without reservation request (UCELL) 20. The condition that a burst must contain a UCELLR allows a subscriber terminal to maintain its channel reservation until all of its payload data has been sent, thus"meaning that the subscriber terminal need not transmit again during the contention interval to request channel allocation to transmit the remainder of its data. This combination of the reservation maintenance request and the upstream cell into one message allows a single downstream acknowledgement to serve as both reser-vation request acknowledgement and payload cell acknowledgement, thus improving bandwidth efficiency.

The structure of a UCELLR is shown in Figure 8, the contents of 16 which are described below in Table 7.

Field Tag Description

Preamble Physical layer synchronization sequence Sync 8 bit MAC framing synchromzation sequence SU-1D ID of the subscriber unit from which the payload data is originated.

RSV-MAINT Reservation maintenance. Used by the subscriber ternunal to continue requesting time slot reservations without contending for them. Cells Number of time slots being requested by the subscriber unit for

future MAC frames Tr-Type Traffic Type of the data to be sent by the subscriber unit in future MAC frames.

SEQ Sequence number. Used by the MAC to resequence cells that get out of order due to cell loss and re transmission Condensed Includes VPI, VCL Traffic Type, Cell Loss Priority.

ATM Header Payload Payload data CRC Cyclic redundancy code. Used to verify correct receipt of the downstream cell.

Table 7: Upstream Cell with Reservation Request (UCELLR) structure The structure of a UCELL is shown in Figure 9, the contents of which are described below in Table 8.

Field Tag Description

Preamble Physical layer synchronization sequence Sync 8 bit MAC framing synchromzation sequence 1 _Field Tag Description

SU-1D ID of the subscriber unit from which the payload data is originated.

SEQ Sequence number. Used by the MAC to resequence cells that get out of order due to cell loss and retransmission.

Condensed Includes VPI, VCI, Traffic Type, Cell Loss Priority.

ATM Header Payload Payload data CRC Cyclic redundancy code. Used to verify correct receipt of the downstream cell.

Table 8: Upstream Cell with no Reservation Request (UCELL) The various data structures used within the MAC which are pertinent to the present invention will now be described with reference to Figure 10.

The first relevant data structure is the subscriber unit list 102. The subscriber unit list contains a list Of actively registered subscriber unit access IDs, with a pointer (104) from e-ac ntry to a set o tra ic queues.

The next structures are the subscriber unit traffic queues 105, 106 and 107. For each subscriber unit that is actively registered with an access point, there is a set of traffic queues, each set containing one queue per traffic type. For each queue there is a Cell acknowledgement map 109. Each time a cell is sent from the access point to a subscriber terminal, the access point expects to receive a positive (ack) or negafive--ac ow e gement nac) o e ce 's arriv e acknowledgement bit map -must -contain as many bits as the maximum number of cells that can be sent in a burst. In the present embodiment, as a non- limiting example, no more than six cells of a given traffic type for a particular subscriber unit can be sent within a single MAC frame, and hence it is not necessary to retain the acknowledgement status for more than six cells of a given traffic queue.

Therefore, each queue has a six bit acknowledgement map corTesponding to the 18 acknowledgement state of the most recently sent six cells. It is to be understood that more or less cells may be sent in a burst without departing from the scope of the present invention, in which case the acknowledgement bit map must contain the same number of bits The next relevant data structures are the pending reservation queues 110, 111, 112, 113. For each traffic type, there is a circular buffer (the pending reservation queue) containing a single entry for each subscriber unit with traffic awaiting a reservation.

The final data structure is the downstream frame reservation schedule 116. Contained within this structure is a map of all bursts to be sent during the subsequent downstream portion of the MAC frame. Entries for each burst include the subscriber unit access identifier (SU-AID), the traffic type of the burst, and the number of cells contained within the burst. The downstream frame reservation schedule may be in exactly the same format as the burst Map contained within the FWDR. By making the format identical, hardware implementability can be improved, and in particular in a hardware implementation using a FPGA, as mentioned earlier. In this case, the number of gates required in the array is reduced, as the same data structure can be re-used for both the burst map in the I'DHDR, and the frame reservation schedule. This re-use allows the MAC to be implemented within present technological constraints, and in particular with regards to the number of gates in presently available FPGAs.

Having described the various signal and data structures used within the present particularly preferred embodiment, the mechanism by which the novel method of ARQ of the present invention is implemented will now be described M detail. For ease and brevity of description, the following description will concentrate on cell transmission and ARQ in the downstream direction only. It will be readily apparent to those skilled in the art how the following is also applied to ARQ in the upstream direction, but an example is also given later of ARQ in the upstream direction.

19 With reference to Figure 10, when a cell enters the MAC, it is de multiplexed into the traffic queue corresponding to its subscriber unit and traffic type. If the queue had been empty upon cell amval, the subscriber unit's SU_AfD is entered into the pending reservation buffer (110, 111, 112, 113) for the corresponding traffic type. With traffic in the subscriber unit queues a cell scheduling algorithm and means (not shown) is then activated.

MAC frame burst maps (schedules) are generated less than one frame in advance of the time that they are to be sent over the air. Th cell scheduler first checks the pending reservation request buffer for the highest priority traffic type to see if any cells of that type are awaiting transmission. If there is an entry in the buffer (indicating that there may be traffic of that type queued up), the scheduler reads the SU-AID from the buffer and uses it as an index into the subscriber unit list 102 to reference the subscriber unit's traffic queue for that traffic type.

Because of the implementation of the cell scheduling algorithm, it is possible for entries to remain in the pending reservation queue even though no more traffic is pending. In that case, the scheduling controller removes the entry from the pending reservation buffer and makes no corresponding entry in the downstream schedule / burst map.

If there is additional traffic in the queue at the time the pending reservation is processed, the scheduling controller places the subscriber uni"t access identifier into the tail end of the buffer, in order to keep the reservation open for potential cell retransmissions.

When the cell scheduler does determine that an entry is to be made in the downstream schedule, it does so in the manner of figure 10. That is, an entry is made including the particular subscriber terminal's SU-AID, the traffic type and to be transmitted, and the number of cells to be transmitted. As mentioned earlier, the frame reservation schedule is identical to the burst map transmitted in the FDHDR.

Cell Acknowledgement is achieved as follows. The bits of the acknowledgement map for a traffic queue correspond to the acknowledgement state of the most recently sent cells. The data are sent to the subscriber terminal during the downstream portion of the MAC frame, and are expected to be acknowledged during the upstream portion. The acknowledgement sent by a subscriber terminal contains a bit map containing the same number of bits as the number of transmitted cells, with a I bit indicating positive acknowledgement and a 0 bit indicating negative acknowledgement. This bit map is transferred into the acknowledgement map for the corresponding SU-AlD and traffic type. The bit map is incorporated directly into the traffic queue data structure itself. Once this transfer has taken place, those cells that have been acknowledged as having been successfully received are then removed from the traffic queue by referring to the acknowledgement bit map. The count of pending cells is the queue is decremented accordingly. By storing the acknowledgement bitmap directly in the traffic queue, cells t hat have been successfully transmitted may be removed at some other time during the MAC frame.

Automatic retransmission of cells that were negatively acknowledged (or alternatively not positively acknowledged) happens automatically as a result of the sequence of operations described above. Two aspects of the MAC design force this to be true. First, only positive cell acknowledgment causes a cell to be removed from a traffic queue, therefore leaving a cell in the traffic queue until it is confirmed received by the destination.

Second, because the cell scheduler continues to add the subscriber terminal access identifier to the pending reservation buffer, the reservafion for that subscriber terminal and traffic type is kept open until all cells have been positively acknowledged.

At the cell destination, cells are buffered and resequenced by a processor that accepts all data from the MAC. Because each cell of data contains a sequence number in its header, resequencing can be performed outside the 21 MAC/ARQ mechanism without any additional complexity.

In order that the present invention be fully understood, an example scenario where multiple cells are transmitted m the upstream direction and ARQ occurs will be described with reference to Figure 11.

This scenario demonstrates the MAC's combination of payload data acknowledgement and reservation maintenance request acknowledgement into a single message, thus improving overall bandwidth efficiency. Furthermore, by combining the reservation maintenance request with a cell of payload data in the form of a UCELLP, the MAC is able to ensure that the subscriber terminal's reservation is kept open until all payload cells have been sent, without the need for further reservation requests from the SU. Moreover, because cell acknowledgement and retry is handled at the MAC layer, it is possible for the wireless access network to maintain short cell latencies even during retries.

When the payload data (80) arrives at the SUs MAC layer, it is put into the appropriate traffic queue for transmission. The arrival of the data causes the SU to initiate the reservation request process. As described before, the SU waits for the next downstream burst (82), then demodulates the FWDR to determine the location of the contention interval within the current MAC frame.

Currently the SU can request a maximum of six slots. The SU generates an SRR (84) and transrrUts it during the contention interval. Since the number of cells to be transmitted in this example is greater than the maximum number that can be requested, the SU requests six slots. The SU generates the request and transmits it during the contention window. When the AP receives the SRR, it acknowledges the request by placing an RRA (86) M the downstream portion of the next MAC frame.

Some time later-, e AP grants time slots to the SU (88). The grant is communicated via the F1)HDR, which includes burst maps for each burst to be sent in the remainder of the MAC frame. The SU detects its SU-AID in one of the burst maps of the FWDR, and takes that as an indication that it is to send some 22 of its traffic upstream. The number of cells granted is also part of the burst map; the access point could grant anywhere from one to six cells. In this example six cells are granted to the subscriber terminal.

The SU generates an upstream cell (UCELLR) which, M addition to the payload data, includes a reservation maintenance request that indicates to the AP that it is requesting time slots for four additional cells. The SU also generates cells to fill the remaining five slots that it has been granted for the current burst. These remaining cells need not contain reservation requests -, they follow the format of the UCELL (see figure 9). The burst of six cells is then sent upstream (90) during the time slots allocated to the SU in the upstream portion of the MAC frame.

Since the AP MAC granted reservations to the SU in the downstream portion of the current frame, it is expecting to receive the same number of cells in the upstream portion of the same frame, and is expecting to send a smgle DACK for the entire upstream burst in the subsequent MAC frame. By the time the upstream cell arrives, the access point MAC has assembled the DACK, minus the bit map containing the acknowledgements of the individual cells. When the AP MAC verifies that the upstream cells have been received correctly, it sets the corresponding bits in the DACK.

The AP MAC expects the first cell of the burst to contain a reservation request. In this case, it finds that the SU has requested four additional time slots. As mentioned previously, the DACK also acts as an acknowledgement of the reservation maintenance request. The AP therefore grants four time slots M the upstream burst portion of the present MAC frame (92). The SU generates its UCELLR plus three UCELLs, and places a reservation maintenance request of 0 slots in UCELLR. The upstream burst is sent as before (94).

The AP has now updated the number of reservations being maintained for the SU. However, if any of the four cells that the AP expected is received in error, the SU will need to retransmit it. It is not practical to require the 1 23 SU to request a reservation and await the reservation grant M order to re- send the single cell that was in error. Rather, any time the AP receives a cell in error, it increments the number of slots reserved for the particular SU. In this example, the AP had received a UCELLR with a reservation maintenance request of 0 cells (94). However, it also received one cell in error, so it increments the number of cells reserved for the SU by one. When the SU receives the FWDR and the DACK of the next MAC frame, it Will know that it has one slot reserved for it, and it wil.1 know which cell to re-transmit from the position of the nack in the DACK map If the AP venfies correct reception of the four cells, it sets the corresponding bits in the DACK. Since the UCELLR contained a reservation maintenance request for 0 slots, no fin-ther reservations are needed and the transfer of cells from SU to AP is complete upon downstream receipt of the DACK (96) by the SU.

As demonstrated by the above example of the method and system of the present invention in operation, the present invention allows both selective repeat ARQ to be fully integrated with the mechanisms for traffic scheduling and medium access control, thus providing for fast selective repeat ARQ of data traffic, and in particular ATM-compatible traffic, while still supporting traffic latency requirements.

24

Claims (12)

CLAIMS:

1. A method of performing integrated data traffic scheduling and automatic retransmission of erroneous data received at a receiver terminal, comprising the steps of - a) scheduling one or more data portions waiting to be transmitted Erom a transmitter terminal to the receiver terminal by queuing said data portions in order in a traffic queue, each of said data portions including an error-check 10 code, b) transmitting one or more of said data portions to said receiver terminal in the same order in which said data portions were queued in said traffic queue; C) detecting the error- check code of each transmitted data 15 portion at the receiver terminal to deterrnine whether each transmitted data portion has been received either correctly or erroneously; d) generating an acknowledgement array containing one or more acknowledgement flags, each acknowledgement flag 20 indicating either the correct or the erroneous receipt of a particular corresponding transmitted data portion; e) transmitting said acknowledgement array from the receiver terminal to the transmitter terminal; and f) removing from the traffic queue those data portions indicated 25 in the array as having been correctly received by the receiver terrainal.; wherein steps a) to f) are continuously repeated in turn such that retransmission of those data portions that have been received erroneously occurs automatically due to those data portions remaining in the traffic queue after step 1 0

2. A method according to claim 1, wherein the acknowledgement flags are arranged in the array in the same predeten-nined order in which the respective corresponding data portions were transmitted at step b), whereby the transmitter terminal may determine from the position of the acknowledgement flags in the acknowledgement array which of the data portions may be removed from the traffic queue at step f). z

3. A method according to claim I or 2, wherein a particular data portion remains in said traffic queue and is repeatedly retransmitted until a positive acknowledgement indicating correct receipt of said particular data portion is received from said receiver terminal.

4. A method according to any of the preceding claims whereinfin-ther data portions may be scheduled for transmission together with said one or more data portions by queuing said further data portions in said traffic queue.

5. A method according to any of the preceding claims wherein each of said data portions further includes a sequence 'index, wheremi said receiver terminal may re-sequence--g-aid data portions into said original predetermined order.

6. A method according to any of the preceding claims for use in a poM"t-to-multipoint network comprising a central control node and one or more remote subscriber nodes.- wh-ereiri the central control node is always either one of the receiver terminal or the transmitter terminal.

7. A method according to claim 6, wherein the central control node 26 indicates to each of the subscriber nodes when each particular node is to be either of the said receiver terminal or the said transmitter terminal.

8. A method according to claim 7 wherein the central control node controls access of each subscriber node to the network.

9. A method according to any of the preceding claims, wherein each portion of data traffic is assigned a priority relative to every other portion of data traffic, and wherein a separate traffic queue is maintained for each priority type, wherein the scheduling step a) further comprises the step of queueing each portion of data traffic in the respective traffic queue corresponding to the respective data portion's priority, and wherein the transmitting step b) further comprises transmitting those data portions queued in a higher priority traffic queue before those data portions queued in a lower priority traffic queue.

10. A method according to any of the preceding claims wherein the transmission of said data portions and said acknowledgement arTay takes place on a wireless channel.

11. A method according to any of the preceding claims wherein said data portions contain ATM compatible cells.

12. A system for performing integrated data traffic scheduling and automatic retransmission of erroneous data received at a receiver terminal, comprising:- a) scheduler means for scheduling one or more data portions waiting to be transmitted from a transmitter terminal to the receiver terminal by queuing said data portions in order in a traffic queue, each of said data portions including an error- 1 27 check code; b) means for transmitting one or more of said data portions to said receiver terminal in the same order in which said data portions were queued in said traffic queue; C) error-detection means for detecting the error-check code of each transmitted data portion at the receiver terminal to determine whether each transmitted data portion has been received either correctly or erroneously; d) acknowledgement generation means for generating an acknowledgement array contairung one or more acknowledgement flags, each acknowledgement flag indicating either the correct or the erroneous receipt of a particular corresponding transmitted data portion; e) means for transmitting said acknowledgement array from the receiver terminal to the transmitter terminal; and f) removal means for removing from the traffic queue those data portions indicated in the array as having been correctly received by the receiver terminal; wherein said- above-mentioned means a) to f) each repeat their operations in turn whereby retransmission of those data portions that have been received erroneously occurs automatically due to those data portions remaining in the traffic queue after the operation of the removal means.