Hybrid type-II ARQ/AMS supported by Channel Predictive Scheduling in a Multi-User Scenario

Nilo Casimiro Ericsson, UU Sorour Falahati, CTH
Anders Ahlén, UU and Arne Svensson, CTH.

IEEE Vehicular Technology Conference - VTC'00-Fall
Boston, MA, USA, September 24-28, 2000, pp. 1804-1811. © 2000 IEEE

Extended Abstract:

In future packet based wireless communication systems, the downlink will put a high demand on the rate of data transmission to mobile terminals. An obstacle in this context is the time-variability of the channel. To achieve a high throughput also over fading channels, adaptive methods for adjustment of, for example, the modulation alphabet and the coding complexity, together with time-slot scheduling can be used. In our previous publications we describe two substantially different approaches to packet data transmission over fading channels, using adaptive methods for modulation and coding.

One approach is the Hybrid type-II ARQ scheme combined with an Adaptive Modulation System (AMS). The idea behind the proposed HARQ-II/AMS is to reduce the signalling constellation size and/or the coding rate at each (re)transmission attempt. Therefore, more resistance against channel impairments can be provided in the latter transmission attempts while still having the prospect of a high transmission rate in the first attempt. The drawback of this system is the delay imposed by the method: The adaptation of coding rate/modulation is only controlled by the ACK/NAK feedback signal which increases the system delay in poor channel conditions. The performance of this system was previously studied and evaluated only for the single user case, where no scheduling algorithm was implemented.

The other approach is based on the knowledge and prediction of the channel quality, allowing for scheduling of the transmission for one or multiple users. Predictions of the different users' channel conditions provide a basis for a detailed scheduling of the transmission, combining time-slot allocation and adaptive modulation. This approach can also take into account the desired error-probability and priority of the different users, and, the traffic situation at hand. Moreover, it utilizes the frequency band efficiently, since the different users are allocated time-slots when their transmission conditions are predicted to be favorable, allowing them to use a high modulation level. The resulting constant and low (user-specified) error-rate provides the error correcting codes with manageable data, avoiding bandwidth consuming re-transmissions. The main drawbacks are the computational complexity, and the sensitivity to channel prediction errors.

In this paper, the two methods are combined and the performance of the combined approach is evaluated and compared to the two individual methods. The combined method successfully reduces the drawbacks of the individual methods described above: Knowledge of the channel conditions gives a hint on the initial coding rate/modulation to use, reducing the delay due to numerous NAKs, and, the ARQ-scheme provides robustness against channel prediction errors.

Based on the required bandwidth for each user, and their priorities, the scheduler allocates the time-slots to them, giving priority to re-transmissions due to NAK-signals. In case of a NAK-signal, the coding rate/modulation is chosen to be less than the one used in the previously unsuccessfully transmitted packet, according to the Hybrid type-II ARQ/AMS scheme which is demonstrated in Table 1 where the modulation scheme and coding rates are ordered with respect to the retransmission attempt. This scheme is based on the rate compatible convolutional codes at the parent rate 1/3 with constraint length 7 and puncturing period 2. The codes at higher or lower rates than the parent code rate are obtained by utilizing optimum puncturing or repetition, respectively. The transmitter uses 16-QAM, 8-PSK, QPSK and BPSK where a constant symbol rate and constant average symbol energy are assumed. Generalizing to the N-user system and employing M time slots, for each new packet, the rate and the modulation level are picked from the M-by-N matrix of predicted values for the different users, described above.

This approach results in a reduction of the system delay due to the prior information of the channel which is available at the transmitter. Additionally, applying channel coding improves the error correction capability of the system and provides robustness against the channel prediction errors, especially in situations where there are considerable changes in the channel conditions during the prediction time interval. Moreover, the error correction is done in an adaptive way, suitable for the time varying channel. The full-length paper contains a full description of the proposed system, the simulation results and proper references to the related work.

Related publications:

Ph.D. Thesis by Nilo Casimiro Ericsson, Oct. 2004.

Licenciate Thesis by Nilo C. Ericsson.
IEEE VTC-fall 99 conference paper on scheduling over fading links.
PCC Workshop'98 conference paper
Globecom'99 conference paper

Source:

Postscript, 326K ; Pdf, 107K