Maintaining high Capacity for Centralized DCA with Limited Measurements and Signaling

Transcription

1 Maintaining high Capacity for Centralized DCA with Limited Measurements and Signaling Miguel Berg Radio Communication Systems Lab. Dept. of Signals, Sensors and Systems Royal Institute of Technology SE Stockholm, Sweden Fax: ABSTRACT Centralized dynamic channel allocation strategies can, in theory, achieve very high capacity in terms of achievable load at a certain quality level. Unfortunately, these strategies have often been considered non-feasible in practice due to the massive amount of measurements and signaling typically required. In this paper we evaluate some methods to maintain high performance while drastically reducing the measurement and signaling requirements. Our DCA algorithm is based on measurements of the link gain matrix. We show that for a microcellular Manhattan environment, only the gains from the three strongest beacons have to be measured and signaled by each mobile in order to achieve almost the same capacity as an ideal system where all beacons can be measured. I. INTRODUCTION The high user data rates that become possible in third generation systems will require new and more efficient methods to manage the radio resources. Static assignment techniques like FCA will not work when, for instance, a single user temporarily needs a large part of the total system bandwidth. Centralized Dynamic Channel Allocation (DCA) systems can give a high capacity but they suffer from delay and complexity problems when they become too large. Recently, some research has been performed on locally centralized or bunched systems. Interference between bunches is handled in a decentralized manner, either with interference averaging or some kind of distributed solution. Even though the interbunch interference can be handled in this way, we loose some capacity on the borders compared with a single bunch system. Therefore, we would like for instance a hotspot to be covered by a single bunch, or as few as possible. This could be a problem if the hotspot contains too many BSs; the computational, signaling and measurements needed in such a system could easily become overwhelming. In this paper we present some ideas on how to reduce the measurement and signaling complexity for large bunches and still maintain a high capacity in terms of allowed load at a certain Quality-Of-Service level. Our study uses the bunch concept as presented in [1],[2]. It is based on the concepts discussed in [3],[4],[5] where a limited number of Remote Antenna Units (RAUs) or sub-base stations connected to a Central Unit (CU) form a bunch. Interbunch interference is handled by a modified Autonomous Reuse Partitioning (ARP) scheme in [5] but it is not considered in [3] and [4]. At low loads, the scheme in [5] is a decentralized system since it only uses the central base station in each cell but at higher loads it starts to utilize the sub-base stations which are controlled from the central base. The RAUs in our bunch are connected to the CU with highspeed communication links, which makes communication cheap and enables bunch-wide burst synchronization. The RAUs could in some cases even be fed an RF signal directly through optical fibers [7]. All information like transmitter powers, receiver noise factors, channels used, channel measurements, etc. is known by the CU. A bunch is ideal for hotspots like central parts of a city or a building floor but in some cases we will need several bunches to cover an area effectively. The aim of this study is to show that large bunches can indeed be manageable from the measurement and signaling point of view. II. SYSTEM DESCRIPTION The bunch concept is very flexible and allows a wide variety of Radio Resource Management (RRM) approaches. Our version of the bunch concept uses a SIR-based dynamic channel allocation and power control scheme as described in [1],[2]. Two different multiple access modes have been selected by ETSI for the third generation UMTS system. Both modes were proposed by the ACTS project FRAMES. The first one is a wideband CDMA (W-CDMA) and the other one is a hybrid between TDMA and CDMA (TD/CDMA). In this paper we have only considered the TD/CDMA mode. It is based on FRAMES FMA1S [4] but some parameters have been adapted to be more compatible with the W-CDMA mode. The current proposal has 16 timeslots per 10 ms frame and a 5 MHz wide carrier. Time-Division-Duplexing (TDD) is used which allows a great flexibility in assigning slots to up- and downlink. Joint detection is used to cancel intracell interference. The number of spreading codes that can be used in each slot varies between eight and twelve but in this study we have only used eight codes. The smallest possible resource that can be assigned within a frame is a {code, slot, carrier} triple.

2 Within the bunch, RRM is based on knowledge of all transmitter powers and the link gain matrix, G [8], which is constructed from downlink measurements on the beacon channel broadcasted by all RAUs as shown in Figure 1. The powers and gain values are then used to calculate if admission of a new user is feasible and which power levels to use for all cochannel mobiles. The objective is that all users should achieve a certain SIR target. The actual algorithm is described in more detail in [2]. It is similar to the feasibility equation in [9] with the main difference being that we use an iterative approach to find the new power levels instead of Gaussian elimination. g 31 Cell 3 g 11 g 13 RAU 3 g 12 MS4 MS3 RAU 1 MS1 RAU 2 Cell 1 Cell 2 Figure 1. Gain matrix measurements. g22 MS2 Ideally, we would like each mobile station (MS) to be able to measure the link gain from every RAU in the bunch. In reality, a mobile can not measure on a beacon if the Signal-to-Noise- Ratio (SNR) is too low. This means that RAUs far away can not be identified due to the low signal strength. Since they contribute little to the total interference, this is not a big loss anyway. Another problem is that beacon channels must be reused, sometimes even within a bunch, in order not to waste valuable radio resources. This limits the number of beacons that the MS can hear since the beacons will interfere with each other. Furthermore, in a large bunch it would take to much time and effort to measure on a large number of beacons and transmit the measurement reports on the radio link to the RAU. In the following section we will describe some methods to maintain a high performance in spite of missing gain values, i.e. gaps in the gain matrix. III. METHODS TO FILL IN MISSING LINK GAINS In this study there are two different limits on the number of measurable beacons: the first is due to the required beacon SNR and the second is a variable upper limit to the number of beacons that the mobiles can measure on. This means that even if we allow the mobiles to measure on infinitely many BSs, they can only reach a fraction of the total number. In order to compensate for the missing gain values we will evaluate the following different methods: I. Use a static cell-to-cell value for the beacons that can not be measured. (based on median or some percentile of previously measured gains) II. III. Increase the SIR target when performing the admission decision. Later when the MS is admitted, we reduce the SIR with the power control scheme. Use a channel selection method that is more tolerant against measurement problems. Method I works as follows: Assume that MS2 in Figure 1 can not measure the gain from RAU3 (g 23 ). The CU can then use the median or some other percentile of the previously measured (by other mobiles) gain values from RAU3 to cell 2 instead. The main motivation for this approach is that the path loss curve flattens out at long distances, i.e. the average path loss per meter decreases. Therefore, interference from RAUs far away from our cell varies little over our own cell compared with interference from our neighbors. As long as we can measure the gain from our neighbors on a regular basis, this will work fine. The problem is that we might never be able to measure on anything else than our neighbors, which would mean that we do not have any values to calculate the median from. A more practical way to find the median values is to let the RAUs measure on each other (RAU-to-RAU gain). This would mean that in our example, the CU could use the gain from RAU3 to RAU2 instead of g 23. The main advantage is that these values do not have to be transmitted on the radio link. If we do not consider the shadow fading, the RAU-to- RAU gains will be quite close to the median values. Method II is perhaps the simplest: if some gain values are missing then increase the SIR target in the admission control by some amount that depends on the number of beacons received. One drawback with this method is that it can be difficult to find the optimum amount to add to the SIR target. Method III uses a different approach; instead of compensating for missing values we change the channel search method to a more robust one. If the MSs can measure interference on all timeslots then the CU can use this info to try to allocate the MS on the least interfered channel first. Since the MS will experience less interference, the risk of making an erroneous admission decreases. The drawback is that the MS must be able to measure interference on all slots in a frame but this is already done in current systems, e.g. DECT. IV. PERFORMANCE EVALUATION A. Models and Parameters We have evaluated the performance of our system with snapshot simulations on a Manhattan grid as shown in Figure 2. The simulation area is finite i.e. no wraparound is used. Perfect estimates of SIR and link gains are assumed in the simulations. The block size is 200 by 200 meters and the street width is 30 meters. The models used are similar or equal to the ones recommended by ETSI for UMTS radio transmission technology selection in [10]. All results are for the downlink since we believe that the majority of traffic will be in this direction. Mobiles have a uniform distribution over the streets and are placed in the street center. Mobility is not modeled. We assume that base station antennas are placed below rooftops. The

3 propagation model is a recursive model [11] where the path loss is calculated as a sum of line-of-sight (LOS) and non-los segments. A dual slope behavior is included with a breakpoint at 300m. Furthermore we calculate the path loss over the rooftops according to the COST Walfish-Ikegami model: L = log(d+20) where d is the shortest geographical distance. The path loss we use in the simulations is the minimum between the recursive and the COST model. Log-normal shadow fading with zero mean and 10dB standard deviation is included. North-south distance [m] Manhattan scenario, 12 by 12 blocks and 72 RAUs calculations in order to lower the computational complexity since there are only 15% nonzero elements in the co-timeslot gain matrix (11/72). We found that this could reduce the computational complexity by at least an order of magnitude. First of all, we study the performance when link gains for unreachable beacons are simply set to zero and channels are selected randomly. This is our reference system. The result is shown in Figure 3 where we see that performance degrades rapidly if we only can measure on three or fewer of the strongest beacons. The dashed horizontal line is the 2% assignment failure rate where we measure the capacity. With beacons=x we mean that each MS can measure on a maximum of x beacons. Inf means that there is no restriction for the number of beacons that the MS can measure on apart from that the beacon SNR must exceed 0 db TD/CDMA,Rand BDCA,zero, East-west distance [m] Figure 2. A Manhattan scenario, 12 by 12 blocks with 72 RAUs. The filled circles show the RAU positions. Our performance measure is the user assignment failure rate,, which we define as the fraction of arriving users that either do not get their minimum amount of channels, or the channels they get are unusable due to insufficient quality (SIR). The total number of available channels in our system is C. We define the relative traffic load, ω C, as the fraction of C requested for in each cell (users/cell/channel). Capacity is defined as the load where the assignment failure rate equals two percent. We have used eight time slots with eight codes each for the downlink. The transmitter power range is 0 to 20 dbm. The SIR target for traffic channels is 10 db. Finally, we assume that the RAU beacons have sufficient coding to be received at SNR values down to 0 db. Beacons do not interfere with each other. Figure 3. Performance for maximum 1,2,3,4 and infinitely many reachable beacons when random channel selection is used and unknown link gains are set to zero. Instead of just setting the unknown gains to zero we tried to replace the zero values with the median of the corresponding cell-to-cell values instead. The result of this is shown in Figure 4 where we see that the capacity for three beacons improved from 0.5 to 0.7. The capacity with fewer beacons did not improve as much since they have their optimum percentile higher than 50% (the median). B. Numerical Results We have studied how the maximum number of reachable beacons affects the performance. In any case, the MSs can not measure on a beacon if it has an SNR lower than 0 db. This gives the 10 th and the 90 th percentile of the number of reachable beacons at 8 and 15 respectively. It means that even the performance when the maximum number of beacons is infinite ( ) can vary with different channel selection methods etc since the MSs can only reach about 11 out of 72 beacons on average. This could be exploited in sparse matrix

4 10 0 TD/CDMA,Rand BDCA,median can measure on one or two beacons. Even with only one beacon we can achieve a load of 0.3 which is slightly higher than the best performing FCA scheme (cluster size three) from [2] TD/CDMA,Lea-Tot-Int BDCA,zero, Figure 4. Random channel selection is used and unknown gain values are filled in with the median method. A disadvantage with the median method is that we can no longer use sparse matrix calculations in the RRM algorithms since there are no zeros in the gain matrix. Thus, it will be a tradeoff between capacity and complexity. We studied which SIR margin (in the admission decision) that would give the best performance when combined with our reference system. The result from a simulation with the optimum margins is shown in Figure 5. With two beacons we can now allow twice the load compared with our reference system. For the other cases we get about the same performance as with the median method TD/CDMA,Rand BDCA,zero, =7dB, =4dB, =1dB, =0dB, =0dB, Figure 6. Least-Total-Interfered (as measured by the MS) channel selection is used and unknown link gains are set to zero. Finally, we combined all three methods to see if the gains were additive. The results in Figure 7 shows that this is in fact true; the combination of all methods gives the best performance so far. With only three beacons we can allow a load of 0.75 which is about 90% of the ideal system s capacity TD/CDMA,Lea-Tot-Int BDCA,median, =7dB, =4dB, =1dB, =0dB, =0dB, Figure 5. The optimum SIR margin (rounded to the nearest db) is used in the admission control () and unknown link gains are set to zero. Our third method is to use another channel selection method than random. Figure 6 shows that the performance is much better than for the previous methods, especially when we only Figure 7. Performance with all three methods combined: Least-Total- Interfered channel selection, median method to fill in missing gains and raised SIR target in the admission control. V. CONCLUSIONS Our ideal system can measure on all available beacons as long as the beacon SNR is above 0 db. The results presented show that we can achieve 90% of the ideal system s capacity even if we only can measure on the three strongest beacons (including

5 the serving cell) and use the methods discussed (median values, increased SIR-margins, and improved channel selection). With only two beacons we get about 60% of full capacity. If we set the unknown gains to zero it is possible to trade measurement and signaling against computational complexity since the gain matrix will be very sparse and thus allow more efficient calculations. The results show that large centralized or bunched systems can indeed be manageable from the measurement and signaling complexity point of view. It is important to remember that the results are not necessarily applicable in other environments since the Manhattan environment is more 1-dimensional than 2-dimensional (due to the effective screening by the buildings). This means for instance that more beacons might be needed in an indoor system. In the future we will study how measurement errors affect the performance. [9] N. Bambos, G.J. Pottie, On the Maximum Capacity of Power-Controlled Cellular Networks, Wireless Communications. Future Directions. Kluwer Academic Publishers, Dordrecht, Netherlands, 1993, pp [10] Universal Mobile Telecommunications System (UMTS); Selection procedures for the choice of radio transmission technologies of the UMTS, UMTS Technical Report 30.03, [11] J.-E. Berg, "A Recursive Method for Street Microcell Path Loss Calculations", Proc. PIMRC 95, Vol 1, pp VI. ACKNOWLEDGMENT This work has been performed in the framework of the ACTS project FRAMES, partly funded by the European Commission. The author would like to acknowledge the contributions of his colleagues from Siemens AG, Roke Manor Research Limited, Ericsson Radio Systems AB, Nokia Corporation, Technical University of Delft, University of Oulu, France Telecom CNET, Centre Suisse d Electronique et de Microtechnique SA, ETHZ, University of Kaiserslautern, Chalmers University of Technology, The Royal Institute of Technology, Instituto Superior Técnico and Integracion y Sistema. REFERENCES [1] M. Berg, S. Pettersson, J. Zander, A Radio Resource Management Concept for Bunched Personal Communication Systems, Proc. Multiaccess, Mobility and Teletraffic for Personal Communications Workshop, MMT 97, Melbourne, Dec [2] Miguel Berg, "A Concept for Hybrid Random/Dynamic Radio Resource Management", Proc. PIMRC'98, Boston, USA, September 8-11, [3] Broddner, Lilliestråhle, Wallstedt, "Evolution of Cellular Technology for Indoor Coverage", Proc. ISSLS 96, Melbourne, Feb [4] U. Dropmann, X. Lagrange, P. Godlewski, "Architecture of a Multi-Cell Centralized Packet Access System", Proc. PIMRC'95, Toronto, Sept 1995, pp [5] K. Madani, H.A. Aghvami, Performance of Distributed Control Channel Allocation (DCCA) Under Non- Uniform Traffic Condition in Microcellular Radio Communications, Proc. ICC 94, pp , [6] A. Klein, R. Pirhonen, J. Sköld, and R. Suoranta, FRAMES multiple access Mode 1 - wideband TDMA with and without spreading, Proc. PIMRC 97, Helsinki, Sept. 1997, pp [7] H. Ichikawa, M. Ogasawara, T. Yoshida, A Microcell Radio System with a Dynamic Channel Control Method, Proc. VTC 93, pp [8] J. Zander, "Radio Resource Management - an overview", Proc. VTC 96, Atlanta, GA, May 1996.