Hybrid ARQ method for packet data transmission

ES 2 289 205 T3

DESCRIPTION

Hybrid ARQ method for packet data transmission with one control channel and one data channel.

The present invention relates to retransmission techniques in mobile communication systems, in particular CDMA systems, and more specifically to a hybrid ARQ (automatic retransmission request) method for packet data transmission combining previously transmitted packets with retransmitted packets. Redundancy increases with each merge operation, and the packet is more likely to be received correctly even in harsh communication environments.

In more detail, the present invention relates to a hybrid ARQ method according to the preamble part of claim 1. This method is commonly referred to in the art as hybrid ARQ type II or III or incremental redundancy.

A common technique for detecting non-real-time service errors is based on Automatic Repetition Request (ARQ) schemes that are combined with No Return Channel (FEC) Error Correction, called hybrid ARQ. If a Cyclic Redundancy Check (CRC) error is detected, the receiver requests the transmitter to send additional bits of data.

Of the different existing schemes, the continuous ARQ of selective repetition is the one most frequently used in mobile communications. This scheme in connection with FEC will be used for next generation mobile communication systems, such as UMTS. A relay unit of the RLC (Radio Link Control) layer is called a PDU (Protocol Data Unit).

Three different types of ARQs are commonly defined in the art as specified below. Examples of corresponding prior art documents are:

Performance ofpunctured channel codes with ARQ for multimedia transmission in Rayleighfading channels; Lou, H. and Cheung, AS; 46th. IEEE Vehicle Technology Conference, 1996;

Analysis of a type II hybrid ARQ scheme with code combing, S. Kallel, IEEE Transactions on Communications, Vol. 38 # 8, August 1990; Y

Throughput performance of Memory ARQ schemes, S. Kallel, R. Link, S. Bakhtiyari, IEEE Transactions on Vehicular Technology, Vol. 48 # 3, May 1999.

* Type I: Bad PDUs are discarded, and a new copy of that PDU is retransmitted and decoded separately. There is no mix of earlier and later versions of that PDU.

* Type II: The wrong PDU to be retransmitted is not discarded, but is combined with some incremental redundancy bits provided by the transmitter for subsequent decoding. The retransmitted PDUs sometimes have higher coding rates and are combined at the receiver with the stored values. That means only little redundancy is added in each retransmission.

* Type III: This is the same as Type II with the only difference that each retransmitted PDU is now self-encoding. This implies that the PDU is decodable without the need to form a combination with previous PDUs. This is useful if some PDUs are so damaged that there is almost no reusable information.

Type II and III schemes are obviously more intelligent and show some performance gain, because they have the ability to adjust the encoding rate to changing radio environments and to reuse the redundancy of previously transmitted PDUs.

To support incremental redundancy, the SN sequence number of the transmission unit has to be encoded separately. The data stored with the known SN can then be combined with subsequent retransmissions.

In prior art the SN is encoded in the PDU header or in the time slot header (eg EP-A-0938207) and is transmitted together with the PDU. If the PDU is corrupted, the header is likely to be destroyed as well. Therefore, the encoding has to be done with a lower encoding rate to be able to read the SN even when the data is wrong. That means there will be a large coding load to ensure the reliable transmission of the sequence number. Therefore, the encoding for the SN has to be different from that used for the PDUs, leading to more complexity. To ensure that the Sn is correct, a CRC parity check could be applied, but a reliable CRC in a few bits is not very efficient.

In addition to the signaling header that is introduced with prior art methods, the complexity of implementation is what has prevented the use of this technique. A large amount of memory is required in the receiver to store the erroneous packets in order to combine them with retransmissions. Since the SNs are not known

ES 2 289 205 T3 before receiving the retransmission, it is not possible to start the combining process before the SNs have been decoded.

The object underlying the present invention is to provide a hybrid ARQ method with less signaling overhead and low implementation complexity.

This object is achieved by a hybrid ARQ transmission method as set forth in claim 1, a reception method according to claim 6, a transmitting apparatus according to claim 9, a system according to claim 12 and a receiving apparatus according to claim 13.

The present invention overcomes the problems of the prior art since the sequence number is transmitted on a separate control channel. This allows to reduce the complexity of the receiver since the sequence number can be transmitted previously, which allows a more efficient decoding and combination of the PDUs that can follow at a later time. Instead of storing the entire frame, decoding the SNs, combining the stored packets with now identified retransmitted packets, and finally decoding the packets, only combining and decoding has to be performed. Furthermore, the distribution of the SNs in a separate channel facilitates the introduction of this method in existing systems, since the PDU format and the full application function in the medium MAC access control layer can be left unchanged in comparison. with a retransmission scheme that does not use a type II / III combination.

According to preferred embodiments, different channelization codes, different time slots and different frequencies are used for the control channel to transmit the sequence numbers and the data channel to transmit the PDUs. This provides an additional performance gain due to the time and frequency diversity and separate physical channels of the PDU and the SN.

Preferably, the data channel for transmitting the PDUs is a channel shared by several users, which allows a more efficient use of the channel resources.

According to a preferred embodiment, the control channel for transmitting the SNs is a dedicated low rate channel or control channel shared by several users to save channel resources.

According to another advantageous embodiment, the QoS service quality of the control channel is independent of the QoS of the data channel for transmitting the PDUs by suitably controlling at least one of the parameters of transmission power, coding rate and spreading factor. Consequently, the transmission efficiency as well as the reliable transmission of the number of sequences is achieved by controlling the QoS of Sn and PDU separately.

For higher data rates, it is advantageous to combine multiple sequence numbers in a sequence number data unit SNDU to compress signaling and increase CRC efficiency. Preferably, the SNDU is multiplexed with other signaling data or user data to save channel resources. According to another preferred embodiment, the SNDU is sent together with an assignment message on the control channel for a shared uplink or downlink channel transmitting at a high data rate.

Depending on the physical channel used and the access technology, the reception of SNs and PDUs are not or are less correlated with respect to time. Although it is advantageous for the SNs of the SNDU to arrive in the order of the received PDUs, the high rate packet transmissions are less time bound and allow a time offset between the SN and the corresponding PDU.

According to another preferred embodiment, the SNDU is applied to more than one frame of the control channel that allows interleaving.

Furthermore, it is preferred that the correct reception of an SNDU is indicated from the mobile station to the base station or vice versa as part of a transmission protocol.

If the sequence number is additionally included in the header of each PDU, ARQ of type

III.

According to another advantageous embodiment of the invention, the method includes that a network control unit transmits a signal whether or not the hybrid ARQ method is used. Alternatively, the signal can be transmitted from the mobile or the base station. As a variant, the base station and / or the mobile station can recognize, from the existence of an SNDU, whether or not the hybrid ARQ method is to be used.

The present invention will now be described in more detail with reference to the accompanying figures, in which:

Fig. 1 shows a frame and interval structure of a DCH frame to which the present invention can be applied.

Figure 2 shows a frame and slot structure of a DSCH frame to which the present invention can be applied.

ES 2 289 205 T3

Figure 3 shows the temporal relationship between the DCH frame and the associated DSCH frame.

Figure 4 shows the DCH frame data structure to be multiplexed into a 10 ms frame.

Figure 5 shows a flow chart explaining the principles of the present invention.

Next generation mobile communication systems, such as UMTS, will provide the capabilities to transmit packets at a variable bit rate. Traffic characteristics can be very rough and require a fast channel allocation strategy. An example for a fast allocation scheme is the use of a shared channel, where only one high rate packet channel is allocated to users who actually have data to transmit. This minimizes downtime for dedicated high-rate channels. An example of a shared channel concept is disclosed in WO-A-00/02326. The invention can be advantageously used with a high rate shared channel.

Creating a dedicated DCH channel as a permanent resource is not very efficient to support packet traffic, since establishing a DCH will take considerable time. For CDMA communication systems using orthogonal codes also the available code resource is limited. Using a DSCH downlink shared channel with fast resource allocation is considered important because, for packet data, the data flow could have high maximum rates, but low duty cycles.

The invention will now be described by way of example only in connection with a downlink shared channel called DSCH. When using a shared channel, the broadcast codes for high rate code users are assigned frame by frame. There will be a signaling channel for assignment messages in parallel to the DSCH. This could be a shared control channel or an associated low rate channel. In the described example a dedicated low rate DCH channel is assigned to each user to maintain CDMA power control and to inform the mobile station when there is data on the shared channel to be decoded. The DCH will be assigned a high spread factor code (for example SF = 256), but will still represent a considerably large header.

Figure 1 shows the frame and slot structure of the low rate DCH containing pilot bits for coherent detection, TPC (Transmit Power Control) bit for power control, TFCI (Transport Format Control Indicator) to indicate the transport format and a data field.

As indicated in the figure, a time slot contains 2,560 slots and 15 slots # 0 to # 14 form a complete frame that has a duration of 10 ms.

Figure 2 shows the frame and slot structure of the DSCH containing only data. The DSCH can transmit variable data rates while different diffusion factors (SF) are applied (k = 0 .. 6 refers to SF = 256 .. 4). The TFCI information in the DSCH includes information about the broadcast factor, the data rate, and the channelization code of the DSCH.

Figure 3 shows the timing relationship of the DSCH with a mobile station (with a low rate DCH) that could get data on the DSCH when there is data to transmit to that user. The timing of the DSCH is known since it is synchronous with other common channels. The high rate channel (DSCH) will only be assigned on demand and will be shared by multiple users. Thus, the data on the DSCH only has to be decoded if there is data indicated by TFCI. At the same time, the continuous DCH can be used to carry other data (eg circuit switched data or other delay limitation) or signaling data. DSCH and DCH operate asynchronously since the different DCHs have different timing from each other, but the relative timing is known to the mobile station and the data can be decoded correctly.

According to one aspect of the invention, the PDU sequence numbers will be sent on a separate physical channel. In the preferred embodiment the SNs are sent along with the allocation message to minimize the signaling header required for packet transmission and the incremental redundancy scheme.

For CMDA communication systems this implies that the channel where the signaling data is applied is broadcast with different channelization codes before modulating the signal. This allows the QoS on this channel to be controlled separately from the channel where the PDUs are sent. For example, the DCH power level can be increased to improve reception of SNs. In future mobile communication systems, such as UMTS, it is also possible to transmit some fields with different power. For example, the power of the DCH data field may be different from the TFCI, TPC, or pilot power. The separation of control and user data provides additional flexibility. Therefore, some systems are also using separate protocol stacks for the user and control plane from the ISO (Organization for International Standardization) OSI (Open Systems Interconnection) protocol stack. One benefit of separating control information from data is that signaling can be combined with other signaling thus providing more efficient transmission. Sending the SNs on a different physical channel can also mean sending them on a different interval (eg TDMA) or at a different frequency (eg FDMA, OFDM).

In prior art systems the sequence numbers are sent along with the PDU for unambiguous allocation and minimal delay. A strong block code is typically used to encode sequence numbers

ES 2 289 205 T3 unique since only one pair of bits has to be encoded. New packet data applications allow some delay that was not acceptable for traditional circuit switched applications (eg voice). In the preferred embodiment the DCH frame containing the assignment message (TFCI) for the shared channel also supplies the SNs for the PDUs to be transmitted in the corresponding DSCH frame. Combining these two methods minimizes the signaling overhead of the shared channel concept and incremental redundancy by using the channels together. By this combination also the delay just introduced is kept to a minimum, because the allocation message is needed in any case if a high rate channel is shared by multiple users. Simulations carried out have shown that the delay for packet data can be reduced even compared to a circuit switched connection since the “large pipe” shared by multiple users is a more appropriate transmission scheme for applications where data does not arrive continuously. . The time difference between the allocation message and the data packets has to be very small, since in a mobile communication environment conditions can change quite frequently.

The sequence numbers will be supplied as an upper layer signaling message in the DCH data field. Since shared channels are only used for higher data rates, it is possible to combine them for reliable coding and use more suitable codes, such as convolutional or turbo codes. In the following, the packet with the SNs will be called the sequence number data unit - SNDU. Figure 4 shows the simplest arrangement of SNs. The sequence numbers for all packets in the next DSCH frame are arranged in order and encoded by a rate 1/3 convolutional encoder. Before encoding, 8 bits are attached to the SNs for code completion as a queue. Other encoding methods could also be used, such as Turbo or BCH encoding. To ensure reliable reception, the data field is protected by a CRC code that can be variable in size of 8,12,16 or 24 bits. The number of PDUs in the DSCH frame, and consequently the number of SNs in the DCH frame that is transmitted, can vary from 1 to more than 100, depending on the PDU size and the chosen data rate of the DSCh. After encoding, punching or repeating is applied to correlate the data on the physical channel. Before interval segmentation, the data is interleaved in one frame (10 ms). Naturally, it should be understood that this coding and multiplexing processing is given only as a simplified example of embodiment of the invention.

It is also possible for the SNDU to be multiplexed together with other signaling data with user data on the DCH. A main advantage of the proposed scheme is that it is possible to group multiple SNs. ARQ protocols generally use a sliding window technique. That means that, except for retransmissions, which are often sent with higher priority, all packets are sent in order. Different arrangements of the SNs can be used to compress the actual information in the SNDU that is sent over the air interface. For example, they do not have to be sent as a list, each SN having around 6 to 12 bits. Instead, they could be sent serially, for example 1-4 or 1 + 3, 7-12 or 7 + 5 instead of 1, 2, 3, 4, 7, 8, 9, 10, 11, 12.

For a high rate shared channel transmitting several PDUs per frame it will be difficult to put the SNDU in a single frame while maintaining the high spread factor (eg 256, 512). A decrease in the diffusion factor should be avoided to minimize the resources allocated in idle times. Therefore, it will be possible to apply the SNDU on more than one frame. The time offset between the DCH and the DSCH should take into account the maximum number of frames per SNDU. The interleaving size can also be increased to multiple frames or remain on a frame basis to make the SNs available as soon as possible. SNs could also be sent in multiple SNDUs to avoid large packet losses if one SNDU is corrupted.

An example will be given below. The SNDU applies to two frames, while the interleaving is only performed in 10 ms. The DCH / DSCH offset is defined to a minimum of one frame. That means that the first SNDU frame is received before the corresponding DSCH frame, while the second is received simultaneously.

The retransmission window size and consequently the number of bits required for the sequence number should also be kept as small as possible to reduce the PDU signaling header. A small window size requires that the round trip delay be as small as possible to speed up retransmissions and the recognition process.

SNs in the DCH data field easily identify whether or not incremental redundancy is used before PDUs are received. This once again reduces the complexity of the receiver, since reconfiguration of the receiver can be done prior to receipt of the PDUs. Incremental redundancy can be easily activated / deactivated with the proposed method, for example when the receiver memory is exhausted.

The sequence numbers identify which PDUs will be combined with each other. Therefore, for correct operation it is essential that the sequence numbers are correct. The CRC will provide an effective means to ensure that the SNDU is received correctly. However, means must be provided in the protocol to resolve sequence number errors that are not detected. A high FEC encoding will ensure that the SNDU is received correctly even when some or all of the PDUs are in error. There is a compromise between reliability and encoding header. It might be more efficient to account for regular failures rather than encode overly reliable data. A recognized problem is that, if the SNDU is lost, all the PDUs of the corresponding frame are sent on the DSCH even though they cannot be identified.

ES 2 289 205 T3

A variant of the invention is that the mobile station will send an indicator on the DCH uplink to the base station after the correct reception of an SNDU. Only when this indicator is received by the base station, the PDUs are sent on the DSCH. If the indicator is not received, PDUs will not be sent and interference will be minimized.

For Type III hybrid ARQ, each PDU is self-encoding, which means that it can theoretically be decoded without any combination with previous PDUs. Enough information is obtained in each PDU to decode it without combining. A different beneficial approach has been found for such schemes. The SNDU is also sent on a separate channel, but it is not encoded very strongly. At the same time, the sequence number is further transmitted as part of the header in the PDU, as in the usual operation. The header is included in the RLC layer. If the SNDU is received correctly, the reception can be improved by combining PDUs. If the SNDU is lost, the PDUs can still be decoded without combination (if reception quality allows), because the sequence number in the PDU header identifies the PDU for the RLC layer. Therefore, the encoding header of the SNDUs is decreased and the protocol can still operate efficiently if the SNDU is lost. This approach has other advantages since it is possible to completely separate the RLC relay protocol from the recombinant process at the physical layer. If the intention is not to use SNDU transmission, the RLC layer protocol is exactly the same as without Type III hybrid ARQ. This allows you to disable the merge operation without any impact on the RLC protocol, PDU structure, or DSCH transmission in general. The drawback is that there is redundant information in the header of the sent PDU in cases where the SNDU is received correctly.

A preferred embodiment of the method of the invention is explained below with reference to Figure 5.

When a mobile station establishes a packet data session in step 100 (eg, Internet access), the base station may decide, depending on the application, to use the DSCH for that user. A dedicated uplink and downlink channel is established. A transport format control indicator TFCI defining the possible data rates in the dSch is assigned by the base station and signaled to the mobile station.

If packets arrive at the base station, the data will be segmented in step 200 into PDUs. Now, the SNs are assigned to the PDUs (step 210), before being stored according to step 220 for possible retransmission. Once enough PDUs have accumulated to be sent on the DSCH, the base station will schedule a frame on the DSCH for this user (step 230). The sequence numbers will be multiplexed, encoded according to Figure 4 and applied on the control channel as shown in the figure at step 240. The base station then transmits the control channel including the TFCI on the DCH to the mobile station. In step 250 the PDUs are multiplexed, encoded and applied on the data channel that is sent on the DSCH. With the specified timing (see figure 3), the mobile station receives the DCH and therefore will be informed by TFCI (step 230) on the DCH (the signal is broadcast with the broadcast code x) about the data to be decoded on the DSCH (the signal is broadcast with the broadcast code y) and its transport format. In the same DCH frame (or subsequent frames, if applied to multiple frames) the sequence numbers will be signaled and decoded by the mobile station (step 260). Therefore, the mobile station knows exactly the start of the DSCH frame and will receive and decode the PDUs on the DSCH (step 270) sent in step 250.

The storing of erroneous PDUs (step 280) and combining with retransmissions (step 270) will take place according to an implemented algorithm that is outside the scope of this description. All correctly decoded packets are transmitted to the higher layers. The unsuccessfully decoded packet will be stored for recombination with retransmissions. Acknowledgment (ACK) and Non-Acknowledgment (NACK) messages (step 290) will be sent to the transmitter according to the implemented RLC protocol.

The mobile station will wait for new packets to transmit while the session continues (return to step 220) and the user is likely to use the DSCH.

In future systems it will be common for there to be multiple logical channels applied on the physical channel. A logical channel could consist of control data or user data and can belong to different applications or protocol entities. Transport channel multiplexing does not necessarily take place at the physical layer, but is likely to be done by the Medium Access Control (MAC) Layer. For incremental redundancy, this higher layer multiplexing is problematic, because a transport block that is passed to the physical layer for transmission can consist of data from different logical channels. After decoding, one of the blocks could be received correctly while the other is wrong. The retransmission has to be done based on the data originally sent. The exact data block including the correctly received data part would have to be retransmitted to make the recombinant process work. Some logical channels might not even use ARQ if they have a low QoS requirement.

Another feature of the present invention is to disable MAC multiplexing to make incremental redundancy more efficient. This can be done in connection with the decision to use incremental redundancy or not. This will ensure that if incremental redundancy is used, different logical channels are passed to the physical layer as separate transport channels. In addition to the transport blocks for each transport channel, additional information is given to the physical layer on whether or not incremental redundancy will be used. Incremental redundancy is only possible for logical channels that apply ARQ (that are in recognized mode).

ES 2 289 205 T3

Which transport channel will use incremental redundancy will also depend, on the downlink, on the capabilities of the mobile terminal. The main limitation of the terminal will be the lack of memory to store the soft decision values. If the mobile terminal cannot support incremental redundancy for all transport channels, incremental redundancy can be disabled for some transport channels.