Common Feedback Channel for Multicast and Broadcast Services

Transcription

1 Common Feedback Channel for Multicast and Broadcast Services Ray-Guang Cheng, Senior Member, IEEE, Yao-Yuan Liu, Wen-Yen Cheng, and Da-Rui Liu Department of Electronic Engineering National Taiwan University of Science and Technology Taipei, Taiwan, R.O.C. Abstract Multicast and broadcast service (MBS) is one of the important services for next generation wireless systems. Normally, the base station (BS) may perform advance radio resource management functionalities based on the channel quality indicator (CQI) reports sent by MBS subscribers. However, the CQI reports lead to high signaling overhead at the uplink. This paper proposed a common feedback channel to reduce the uplink signaling overhead. A simple dynamic modulation and coding scheme (MCS) is then presented to improve the spectral efficiency based on the limited information carried by the common feedback channel. Simulation results showed that the proposed method could enhance the spectral efficiency with reduced signaling overhead. The results also demonstrated the robustness of the common feedback channel in the presence of feedback errors. Index Terms anonymous feedback; common feedback channel; dynamic modulation and coding scheme (MCS); multicast and broadcast services (MBS); NACK-based feedback M I. INTRODUCTION ULTICAST and Broadcast Services (MBS) (which is also known as multimedia broadcast/multicast services, MBMS in 3GPP Long Term Evolution) is one of the important services to be supported by the next generation wireless systems. MBS is a point-to-multipoint service where data packets are transmitted simultaneously from a single source to multiple destinations [1]. In MBS, the same data is transmitted via a common broadcast or multicast channel to multiple MBS subscribers (MSs). Normally, the network may use the most robust modulation and coding scheme (MCS) to transmit the MBS in order to guarantee the quality of service (QoS) of the MS. Basically, the base station (BS) can use either single-bs access scheme or multi-bs access scheme to transmit MBS. In single-bs access, each BS transmits data with its own MCS, while multi-bs access is implemented by multiple BSs with the same MCS level over the same frequency channel at the same time. Moreover, MSs should be able to receive MBS in both connected state (consists of active mode and sleep mode) and idle state. Potential MBS applications include streaming services, file download services, and carousel services (combines aspects of both the streaming and file download services with repetition and update to reflect changing circumstances) [2][3]. Several techniques such as dynamic MCS and hybrid automatic repeat request (HARQ) can be used to enhance the spectral efficiency of MBS. However, these techniques rely on MSs feedback information to make proper decision. Properly chosen feedback information and feedback channel may reduce the signaling overhead. The selection of the feedback channel and the feedback information depends on the state of the MS and the required signaling overhead. Potential uplink (UL) channels that can be used to carry the feedback information include dedicated control channel (DCCH), shared channel (SCH), and random access channel (RACH). Only connected state MSs may transmit their feedback information through DCCH or SCH. In contrast, both connected and idle state MSs may transmit their feedback information over RACH [4]. Possible feedback information indicating the satisfaction of MS s QoS requirement includes acknowledgment (ACK) or negative acknowledgment (NACK), channel quality indicator (CQI), signal-to-noise ratio (SNR) and received signal strength indicator (RSSI). Several MCS selection methods have been proposed to enhance the spectral efficiency of MBS. In [5], Motorola proposed a method to determine an MCS level before starting MBS transmission. The BS indicates a potential MCS level and MSs transmit NACKs if their radio conditions can t support that MCS level. However, this method relies on user-specific feedback and thus, may result in high signaling overhead in the UL direction. Moreover, the method can only be used for initial MCS selection since the selected MCS level may be useless if MSs change their locations or dynamically join/leave the network. TD-Tech [6] proposed a method for the BS to determine the MCS level via the worst CQI sending from the unsatisfactory MSs. However, the CQI reporting may result in huge signaling overhead in the UL. In order to optimize the system throughput, Lu et al. proposed a method to improve channel condition of the weakest terminal by exploiting channel diversity of frequency-selective attenuation [7]. Each MS should feedback CQI on each sub-channel to determine the sub-channel with better channel condition. With better sub-channel condition could optimize system throughput performance while guaranteeing system coverage. However, the CQI reporting on each sub-channel for each MS may result in high signaling overhead. For MBS, there is a tradeoff between transmitted bit rates (or spectral efficiency) and intended coverage areas [8]. The spectral efficiency may be maximized by adopting radio resource management techniques subject to the given coverage constraint. Existing radio resource management of MBS [5-7] tried to improve the radio spectral efficiency based on accurate feedback information provided by each MS. It results in high signaling overhead in the uplink. Cai et al /11/$ IEEE 1958

2 [8] proposed three schemes to reduce the uplink feedback. In the proposed schemes, MSs are sorted by their path-loss, G-factor (the value of SINR excluding fast fading), or short-term BLER. The BS selects N MSs with the largest path-loss, lowest G-factor, and highest BLER to form a feedback set. Only MSs belonging to the feedback set are eligible to report their CQI to the BS. MCS is then selected based on the MS with the poorest CQI in the feedback set [8]. The feedback set reduces the uplink signaling overhead by minimizing the number of MSs reporting CQI. However, the BS requires complete knowledge of the MSs to maintain a list of sorted MSs. For example, the BS may rely on the status reports sent from all MSs to maintain an up-to-date feedback set. Hence, the signaling overhead of the status report may not be ignored if the status of MSs are frequently changed. This paper presents a dynamic MCS based on a proposed anonymous common feedback channel. Different from [8], which requires N (i.e., the size of the feedback set) dedicated feedback channels to obtain CQI information from portion of the MSs, this paper uses a common feedback channel to obtain ACK/NACK feedback information from all MSs. The proposed method can be used to enhance the spectral efficiency of MBS applications with uplink minimum signaling overhead. The rest of the paper is organized as follows. Section II defines the system model and the objectives of this paper. The proposed common feedback channel is also elaborated. Section III presents the design concepts of a dynamic MCS selection algorithm based on the proposed anonymous common feedback channel. The simulation results are shown in section IV. Finally, conclusions are drawn in section V. II. SYSTEM MODEL This paper considers a network adopts a single-bs access scheme to offer MBS. For simplicity, a single MBS service is considered. Figure 1 shows a timing diagram illustrating the MBS transmission. The time axis is divided into fixed interval of radio frames. Each radio frame consists of a downlink (DL) portion and an uplink (UL) portion. Each MS shall receive the MBS media access protocol (MAP) [9] for the radio resource allocated for MBS data burst and uplink feedback channel (if available) is this radio frame. MBS data burst is sent by the BS every MBS scheduling interval (MSI). Each MSI is equal to m radio frames. Assume that the BS will set a feedback condition every query interval, T query, which is equal to n MSIs. Hence, the BS has to reserve an UL channel for MBS feedback every n MSIs. In this paper, a NACK-based anonymous common feedback mechanism [10] is proposed to provide information from all of the MSs subscribing the MBS. NACK-based means that only a signal is transmitted indicating the non-satisfactory to a specific feedback condition. Anonymous means that all MSs send identical information without carrying users identifications [11]. Common means that both idle and connected states MSs shall transmit their feedback information through the same feedback channel. In the implementation, a BS may reserve a radio resource unit as an anonymous common feedback channel. The BS may have to assign a service-specific code division multiple access (CDMA) code for each MBS. The MS shall transmit the NACK with CDMA code through the anonymous common feedback channel if the feedback condition is not met. The anonymous common feedback channel behaves as an RACH used in the conventional CDMA system. Hence, MSs in either connected or idle state can use the single anonymous common feedback channel to indicate their NACK. Moreover, multiple feedbacks from different MSs will be combined at BS as multipath since identical CDMA code is sent at the pre-defined radio resource unit reserved for the anonymous common feedback channel. As shown in Fig. 1, BS may specify a feedback condition in its downlink (DL) multicast control channel (MCCH) to all MSs within its coverage in an event-triggered or periodical way to perform radio resource management functions such as MCS selection or hybrid ARQ (HARQ) [9]. Upon receiving the feedback condition set by the BS, the MS shall send a NACK if the feedback condition is not satisfied. In this paper, signal-to-interference plus noise ratio (SINR) is used as the feedback condition. Other parameters can also be used herein. The BS may also set a reporting probability p (0 < p 1) to limit the number of simultaneous feedbacks for the purpose of, for example, user counting [11]. That is, unsatisfied MSs may send a NACK only if a probability test is passed. Note that the NACK may not be successfully decoded due to multiple access interference (MAI). In this case, the unknown signal can still be interpreted as a NACK if the receiving power of the anonymous common feedback channel exceeds a threshold [10]. The effectiveness of the dynamic MCS selection scheme can be evaluated in terms of the spectral efficiency. The spectral efficiency is defined as [11] R spectral efficiency, (1) BW * TR where R is the aggregate cell/sector throughput, BW is the used channel bandwidth, and TR is the time ratio of the link reserved for MBS. Note that R depends on the intended coverage of the BS and thus, it may affect the percentage of the satisfactory MSs receiving the MBS. For example, the spectral efficiency measured over 95% of the coverage area with a target packet error rate of 1% is chosen as a performance metric in [12]. III. DYNAMIC MCS SELECTION SCHEME In this section, a dynamic MCS selection scheme using the proposed NACK-based anonymous common feedback channel is described. The proposed scheme selects the MCS level by setting SINR threshold, T SINR, as feedback condition. The MSs shall indicate a NACK through anonymous common feedback channel if its SINR is less than T SINR. In contrast, MSs shall keep silence if its SINR is greater or equal to T SINR. Note that the BS can only be notified when any NACK is transmitted from anonymous common feedback channel. The proposed dynamic MCS selection scheme may work with or without power control. The inter-cell interference can be reduced by adopting power control but at 1959

3 the risk of reduced coverage. In the implementation, the BS always transmit MBS data using the maximum transmit power if the power control is not enabled. The proposed MCS selection scheme is developed based on two basic design concepts. The first concept is to adopt a query-before-action approach to prevent from the ping-pong effect resulted from blind decision. The query-before-action approach may also reduce the risk of a wrong MCS selection due to the lost of NACK. Generally, a BS may select wrong MCS selection (i.e., a MCS level which is higher than the one that all MSs can supported) only if at least two successive NACKs are lost. The second concept is that the BS may either select a lower transmission power to reduce the inter-cell interference, or choose a higher (or less robust) MCS level to enhance the spectral efficiency if all MSs are all satisfied with the modification. In the implementation, the BS shall set T SINR according to a target action (i.e., transmit power control or MCS adjustment) and query for the satisfactory of MSs before taking the action. To increase the operational flexibility, the decision may be made based on the result obtained from a single query or multiple queries. Let N u, and N d be positive integers denoting the thresholds required to upgrade and downgrade the MCS level, respectively. A lower N d (or a higher N u ) is used if we aim to ensure a strict QoS. In contrast, a higher N d (or a lower N u ) is adopted if we want to enhance the spectral efficiency at the cost of degraded QoS. Note that a higher N u may help to combat with the NACK transmission due to the bad wireless channel. The details of the proposed scheme are described below. Let P current be the current BS transmission power; P new be new transmit power that may be used in the upcoming MSI; T SINR (MCS i ) be the minimum SINR requirement of MCS level i. For simplicity, upgrade is used to indicate the change of MCS level from MCS i to MCS i+1 or increase the transmit power; downgrade is used to indicate the change of MCS level from MCS i+1 to MCS i or decrease the transmit power. The proposed dynamic MCS selection algorithm is summarized below. Initially, the network assigns a CDMA code to MBS subscribers during service creation [9]. The BS will use the most robust MCS level and the maximum transmit power to offer the best transmission quality. The initial parameters are set as MCS i = MCS 0, P current = P new = P max, and T SINR = T SINR (MCS 0 ). The BS periodically announces the feedback condition T SINR in the MCCH and adjusts its MCS (and transmit power) according to the result received from the common feedback channel. In the procedure, the case of N u = N d = 1 was illustrated. The general case was illustrated in Figs. 2 and 3. Dynamic MCS Selection Scheme: Step 1: Send a message with feedback condition T SINR = T SINR (MCS i ) + (P max - P new ). - If any NACK is received and P current = P new, go to Step 2 (i.e., increase the transmission power or decrease the MCS level). - If any NACK is received and P current P new, go to Step 3 (do not adjust the transmission power). - If no NACK is received, go to Step 4 (i.e., decrease the transmission power or increase the MCS level). Step 2: - If P current < P max, set P current = P new = P max (increase the transmission power). Return to Step 1. - If P current = P max and MCS i > MCS 0 (i.e. the minimum MCS level), set MCS i = MCS i-1 (decrease the MCS level) and P current = P new = P max. Return to Step 1. Step 3: Set P new = P current and keep the current transmission power. Return to Step 1. Step 4: - If P current = P new (new transmission power is not adjusted). Set P new = P current Δ (i.e. adjust the new transmission power). Return to Step 1. - If P current P new (new transmission power is negatively adjusted but no NACK is received) and (P max - P new ) >= T SINR (MCS i+1 ) T SINR (MCS i ) (i.e., power margin is good enough to accommodate an MCS adjustment), set MCS i = MCS i+1 and P current = P new = P max (increase the MCS level and restore the transmission power). Return to Step 1. - If P current P new and (P max - P new ) < T SINR (MCS i+1 ) T SINR (MCS i ) (i.e., power margin is not good enough to accommodate an MCS adjustment), set P current = P new (decrease the transmission power). Return to Step 1. Figures 2 and 3 shows the general procedure used to upgrade and downgrade the MCS level, respectively. It was assumed that the BS used MCS level i (MCS i, 0 i i ) at the beginning of the observation interval. Figure 2 shows the procedure used to upgrade the MCS level. The BS must ensure that all MSs are satisfied with the existing setting (i.e., no NACK is received in the past query) before upgrading the MCS level. The BS then sets an aggressive condition (i.e. higher T SINR ) according to the requirement of a target action (i.e., decreases the transmission power or increase the MCS level). The action is performed only if the BS ensures that MSs are all satisfied with the modification (i.e., no NACK is received in successive N u queries). Figure 3 shows the procedure used to downgrade the MCS level. This procedure is triggered if a NACK is received during periodic query. The BS shall increase the BS transmission power or select a lower MCS level if N d successive NACKs are received. IV. SIMULATION RESULTS Simulations were conducted on top of a C-based platform to verify the effectiveness of the proposed dynamic MCS selection scheme. In the simulation, a single BS with cell radius of 750 meters and three sectors per cell was considered. All MSs were randomly distributed in the cell. The effect of path loss was considered in the simulations. The modified COST 231 used in [12] is chosen to model path loss effect for the carrier frequency of 2.5 GHz. The antenna height for BS and MS are set to be 32m and 1.5m, respectively. The path loss is given by PL[ db] log10 ( R) (2) where R is the distance from the transmitter to the receiver in kilometer. Outer-ring interference and 10% channel error probability were considered. The effects of multipath fading and intra-cell interference were neglected. A random way max 1960

4 point mobility model was adopted here. The MS may change it speed and moving direction if the MS moves exceeds an uncorrelated distance of 100 meters. In the simulation, 20 samples were observed. Each sample was obtained by averaging 5x10 5 outcomes. Each outcome was collected within 0.05 seconds (i.e.10 frames). The possible MCS levels and the corresponding SINR requirement were obtained from [13]. In the simulations, 5 ms frame length was used; MSI = 10 frames; TR = 5/8, T query = 100 frames, Δ = 1.5 db, N d = N u = 1 and the reporting probability p = 1. For simplicity, it was assumed that the BS uses the 100% of the DL resources for transmitting MBS. The two main performance indices for MBS of spectral efficiency and the coverage were observed. As suggested in [14], the coverage was obtained as the MBS subscribers who correctly received the MBS content divided by the total number of MBS subscribers averaged during the whole simulation interval. The performance of the proposed common-feedback-based dynamic MCS and the reduced-feedback-based dynamic MCS presented in [8] was first investigated. In the simulation, HARQ is not enabled. Similar to [8], the two MCSs selected the MCS based on the MS with the poorest quality (i.e., 100% coverage). It was assumed that one radio resource unit is used to carry the CQI from an MS in the reduced feedback scheme [8] and the signaling overhead resulted from the status report for updating the feedback set was ignored. As in [8], 10 MSs per sector were investigated. In the reduced feedback scheme, 4 MSs was selected in the feedback set as suggested by [8]. In the simulations, it was found that MCS algorithms developed based on the two feedback mechanisms achieved a similar performance. However, the reduced feedback scheme consumed four times of the radio resource units than that used by the anonymous-common-feedback scheme. Due to the space limitation, the results of reduced feedback scheme were not explicitly shown since its performance is similar to that of the proposed anonymous common feedback scheme. Figure 4 shows the performance of the proposed dynamic MCS scheme for different number of MSs. In the simulation, the MCS level was dynamically selected from MCS 0 to a maximum MCS level based on the information provided by the common feedback channel. Each point in the figure was obtained by setting the maximum MCS level to be MCS 0, MCS 3, MCS 6, MCS 9, MCS 12, and MCS 15, respectively. The maximum MCS level of MCS 15 was located at the leftmost endpoint of the curve, which achieved the highest spectral efficiency but had the lowest percentage of satisfactory MSs. In contrast, the MCS 0 was located at the rightmost endpoint of the line, has the lowest spectral efficiency but achieves the highest percentage of satisfactory MSs. The observations were listed as below. First of all, the dynamic MCS scheme can fulfill the coverage requirement of MBS accommodating different number of MSs. The coverage was greater than 97% in all of the scenarios. Second, the trade-off between spectral efficiency and coverage can be controlled by limiting the maximum MCS level. The dynamic MCS scheme may achieve a larger dynamic range of the spectral efficiency (or, a higher spectral efficiency) if the BS serves less MSs. For example, the maximum spectral efficiency of 1.8, 1.3, and 0.76 bps/hz are achieved if the BS served 2, 5, and 40 MSs, respectively. It was because that the proposed dynamic MCS scheme selected the MCS based on the MS with the poorest quality. The BS may have a higher chance to receive a NACK and thus, select a robust MCS if it serves more MSs. Third, enabling power control may decrease the inter-cell interference but at the cost of slightly degradation of the coverage and the spectral efficiency in the serving cell. The proposed NACK-feedback scheme may fail if NACKs are lost in transmission. Hence, a worst case scenario was studied to investigate the robustness of the NACK-feedback scheme. The worst case scenario was demonstrated for a BS accommodating few MSs in a bad wireless channel. The few MSs result in less NACKs to be sent by the MSs and thus, the lost of a NACK may result in erroneous decision. A bad wireless channel may lead to a high NACK error probability. In the simulation, a BS accommodating only 5 MSs with a NACK error probability of 0.1 was investigated and the results were shown in Fig. 5. Note that the packet error of a single MS may result in 20% (i.e., 1/5) coverage lost if the BS serves 5 MSs. Hence, the NACK error probability of 0.1 may only constitute a coverage lost of 2%. In the simulation, it was found that the NACK error probability of 0.1 only resulted in a maximum coverage lost of 1.5%. The proposed query-before-action approach helps to smooth out the effect due to the lost of NACKs. In the query-before-action approach, the dynamic MCS scheme will select a higher MCS level if no NACK is received in successive N u queries. Hence, a wrong MCS will be chosen (i.e., select a higher MCS level than that can be supported by all MSs) only if N u successive NACKs are all lost. Thanks to the time diversity, for a NACK error probability of p, the probability of the lost of N u successive NACKs was greatly reduced to p Nu. Hence, the lost of NACKs can be controlled by increasing N u but at the cost of slightly reduce the spectral efficiency. In the simulation, it was found that the coverage was still higher than the 95% target in such a worst case environment. The impact of the NACK error was insignificant when lower maximum MCS levels were used. In these cases, the BS was prohibited to use higher MCS levels and thus, the performance lost due to NACK error can be reduced. The proposed common-feedback-based dynamic MCS scheme is expected to work well for a cell accommodating more MSs. In these cases, more MSs are able to indicate their NACKs and thus, it reduces the probability that BS receives no NACK due to NACK error. V. CONCLUSION This paper presents an anonymous common feedback channel to reduce the uplink signaling overhead of MBS feedback. A dynamic MCS scheme is then proposed to enhance the spectral efficiency of MBS based on the limited information obtained from a common feedback channel. Instead of using the exact signal quality of MSs, BS adjusts the MCS based on MSs simple NACK indication sent over the common feedback channel. In the simulation, it was found that the performance of the dynamic MCS scheme based on the information provided by the common feedback channel and the reduced-feedback-scheme proposed in [8] 1961

5 was similar. However, the proposed anonymous common feedback channel greatly reduced the uplink signaling overhead more than the reduced-feedback-scheme did. The performance of the dynamic MCS scheme for a BS accommodating different number MSs in a wireless channel with high NACK error rate were also investigated. The results showed that the proposed dynamic MCS scheme still worked well for a BS accommodating more than 5 MSs in such a worst-case feedback situation. ACKNOWLEDGEMENT The authors would like to thank anonymous reviewers for their valuable comments, which help to improve the quality of the presentation. This work was supported in part by the National Science Council (NSC), Taiwan, under Contract NSC E REFERENCES [1] T. Jiang, W. Xiang, H. H. Chen, and Q. Ni, Multicast broadcast services support in OFDMA-based WiMAX systems, IEEE Communications Magazine, vol.45., pp.78-86, Aug [2] M. Knappmeyer and R. Toenjes, Adaptive data scheduling for mobile broadcast carousel services, IEEE Vehicular Technology Conference Spring, pp , April [3] 3GPP TS Multimedia Broadcast/Multicast Service (MBMS) user services; Stage 1 (Release 6) [4] M. Malkowski, Spectrum efficient multicast and asymmetric services in UMTS including performance simulation results, IST / OverDRiVE/ WP1/D15, [5] Mortorola, MBMS modulation and coding state selection, 3GPP R , June [6] TD-Tech, MBMS rate control and feedback suppression, 3GPP R , Feb [7] S. Lu, Y. Cai, L. Zhang, J. Li, P. Skov, C. Wang, and Z. He, Channel-aware frequency domain packet scheduling for MBMS in LTE, IEEE Vehicular Technology Conference Spring, [8] Y. Cai, S. Lu, L. Zhang, C. Wang, P. Skov, Z. He, and K. Niu, Reduced feedback schemes for LTE MBMS, IEEE Vehicular Technology Conference Spring, [9] S. Hamiti, IEEE m system description document, IEEE m-08/003r8, April [10] R. G. Cheng and F. M. Yang, Proposed text for indication of multiple NACKs in a single PRU (E-MBS), IEEE C80216m-09/1958, Aug [11] A. Reznik, and E. Zeira, Contentious feedback in Cellular Systems, IEEE Wireless Communications and Networking Conference (WCNC), pp. 1-5, [12] R. Srinivasab, IEEE m evaluation methodology document (EMD), IEEE m-08/004r5, January [13] Fujitsu R&D centre, Modulation and coding set design for IEEE m system, IEEE C802.16m-09/0216, Jan [14] L. Zhang, Z. He, K. Niu, B. Zhang, and P. Skov, Optimization of coverage and throughput in single-cell embms, IEEE Vehicular Technology Conference Spring Fall, Figure 1. MBS transmission timing diagram. Figure 2. The procedure used to upgrade the MCS level. 1962

6 Figure 3. The procedure used to downgrade the MCS level. Figure 4. The coverage and spectral efficiency of the proposed common-feedback-based dynamic MCS scheme. Figure 5. The coverage and spectral efficiency of the proposed common-feedback-based dynamic MCS scheme. 1963