Interference Management for 4G Cellular Standards

WIMAX/LTE UPDATE Interference Management for 4G Cellular Standards Nageen Himayat and Shilpa Talwar, Intel Corporation Anil Rao and Robert Soni, Alcatel-Lucent ABSTRACT 4G cellular standards are targeting aggressive spectrum reuse (frequency reuse 1) to achieve high system capacity and simplify radio network planning. The increase in system capacity comes at the expense of SINR degradation due to increased intercell interference, which severely impacts cell-edge user capacity and overall system throughput. Advanced interference management schemes are critical for achieving the required cell edge spectral efficiency targets and to provide ubiquity of user experience throughout the network. In this article we compare interference management solutions across the two main 4G standards: IEEE 802.16m (WiMAX) and 3GPP-LTE. Specifically, we address radio resource management schemes for interference mitigation, which include power control and adaptive fractional frequency reuse. Additional topics, such as interference management for multitier cellular deployments, heterogeneous architectures, and smart antenna schemes will be addressed in follow-up papers. INTRODUCTION Most fourth-generation (4G) systems, including WiMAX 802.16m [1 3] and Third Generation Partnership Program Long Term Evolution (3GPP-LTE) [4], are targeting single-frequency deployments. Although aggressive frequency reuse results in a significant increase in system capacity, it also severely degrades the performance experienced by cell edge users due to the increased interference caused by out-of-cell transmissions. Figure 1 illustrates the degradation in signal-to-interference-plus-noise ratio (SINR) for reuse 1 relative to reuse 3, which is approximately 10 db. While the increase in capacity due to the availability of increased bandwidth can typically offset the capacity loss due to SINR degradation, the capacity of users with very weak SINR (cell edge users) still degrades. Hence, interference management schemes are critical to improve the performance of cell edge users. Both 802.16m and 3GPP-LTE, therefore, have focused on several interference management schemes for improving system performance. These techniques include semi-static radio resource management (RRM) through adaptive fractional frequency reuse (FFR) mechanisms, power control, and smart antennas techniques to null interference from other cells. Together, these techniques aim to address the aggressive requirements of > 2 improvements in cell edge user throughput and absolute spectral efficiency over prior releases [1, 4]. This article describes and evaluates the performance of key interference management techniques across the 802.16m and 3GPP-LTE standards. In particular, we focus on RRM schemes, which include FFR and power control. Smart antenna schemes, although extremely important for interference management, will be discussed elsewhere due to limited space. We focus on the standard cellular network deployments, and interference management for multitier/heterogeneous network deployments in which low-power nodes are placed throughout a macro cellular network [5] is deferred to follow-on papers. The organization of the article is as follows. The next section focuses on the downlink (DL) interference management scheme covering adaptive FFR techniques. We then cover uplink (UL) techniques focusing on power control and UL FFR algorithms. Final conclusions are presented in the final section. DL RRM Multicellular RRM efficiently partitions resources across cells in order to manage per resource interference experienced in each cell. Both 802.16m and 3GPP-LTE have focused on semi-static RRM techniques, which adapt frequency reuse across cells based on user distribution and traffic load. In particular, a mix of high and low reuse frequency resources (e.g., reuse 1 and 3, respectively) are allowed in each cell. Resources governed by reuse 1 can be assigned to users that are closer to the center of the cell and hence experience less interference from other cells, while the lower reuse resources are assigned to interference-limited users at the cell edge. Allowing a combination of frequency reuse patterns overcomes the capacity limitation inherent with lower frequency reuse, while also retain- 86 0163-6804/10/$25.00 2010 IEEE

ing a low interference environment to retain throughput and coverage for cell edge users. Also note that the definition of what constitutes cell center vs. cell edge users is an important part of FFR design and is typically based on SINR metrics rather than actual user location within the cell. In the following section we discuss the use of FFR schemes for interference management in the DL. UL-FFR is closely tied to power control mechanisms for interference management and hence is discussed together with UL power control techniques. DL ADAPTIVE FREQUENCY REUSE IN 802.16M Soft FFR DL FFR in 802.16m combines reuse 1 resource with either reuse 3 or reuse 2 resources. Both soft and hard schemes can be supported. Hard reuse refers to the case where a higher reuse factor (e.g., reuse 2, 3) is achieved by shutting off the interfering base station (BS) on certain resources. In contrast, soft reuse refers to the case where higher reuse factors are supported by restricting the interfering BS DL transmit power on certain resources, rather than turning them off. For all reuse schemes, the total DL transmission power is kept constant and below the maximum allowed value. Soft FFR is beneficial because these lower-power resources can still be used in the cell to service additional cell center users with good link conditions, without causing much interference to cell edge users in other cells. Figure 2 illustrates the soft FFR scheme used in 802.16m, where logical OFDMA resources are divided into four frequency partitions comprising reuse 1 and soft reuse 3 resources. In the figure all cells (sectors) transmit on the reuse 1 partition with equal power, while the transmit power on the remaining reuse 3 partitions is based on the primary partition assigned to the cell for transmission. The actual power allocation across the frequency partitions is a function of user distribution across the cell and is optimized cooperatively among cells based on user feedback. Normalized Spectral Efficiency Resource (power) allocation across the sectors results in an associated cost for each partition, which captures the spectral efficiency (SE) penalty implied by lower reuse. For example, a resource belonging to a hard reuse 3 partition will use three times the cell bandwidth when compared to one in a reuse 1 partition, hence will incur three times the cost in terms of lowered spectral efficiency. This cost-weighted spectral efficiency associated with a resource is referred to as the normalized spectral efficiency and is computed as Normalized SE (resource) = Expected SE (resource)/resource Metric (partition). The resource metric indicates the cost or spectral efficiency penalty associated with the soft reuse factor of the partition. The normalized spectral efficiency indicates the true spectral efficiency of a given partition, and is used to F(x) 1 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0-10 0 Empirical CDF 10 20 30 40 50 60 Geometry SINR (db) Figure 1. Geometric SINR distribution for a network with multiple frequency reuse factors (500 m cell). Sector 1 Sector 2 Sector 3 Tx power Tx power Tx power Partition (0) Partition (1) Figure 2. Soft FFR reuse partitioning in IEEE 802.16m. determine the preferred FFR partition (PFP), corresponding to the maximum average normalized SE, for each user. Dimensioning of the FFR partitions and the associated resource metrics is based on cooperative sharing of the PFP by all users in the system. Various optimization schemes may be used to derive the optimal FFR parameters from these reports. The exact algorithm is implementation-dependent. 16m FFR Protocol The initial FFR partitions and the corresponding Resource Metrics are available to users as broadcast information. Upon network entry, the user measures the average SINR on each frequency partition and computes an average normalized SE. It then computes the maximum normalized SE across all FFR partitions and reports the corresponding partition as its PFP. These PFP reports are aggregated across base stations in the system to update the FFR configuration, including partitions size and power level. A user can periodically update its PFP with changing SINR conditions across the partition. The user will also report channel quality indicator (CQI) metrics on the best M resources in Partition (2) Reuse 1 partition Reuse 3 partitions Logic PRU index Reuse 1 Reuser 3/2 Reuse 3 Partition (3) 87

Scheme Channel model Average SE (b/s/hz/cell 2 2) Gain in SE Cell-edge SE (b/s/hz/cell) Gain in cell-edge throughput Localized baseline ITU PedB- 5.84 0.067 Localized AFR-S 3 km/h 5.93 1.6% 0.068 3.3% Distributed baseline ITU PedB- 3.68 0.032 Distributed AFR-S 3 km/h 4.52 22.8% 0.060 88.9% Table 1. Performance results for IEEE 802.16m adaptive soft FFR. Cell border throughput (kb/s) 700 600 500 400 300 200 100 0 0.8 1.0 Reuse 7/6, frequency selective CQI Reuse 1, frequency selective CQI Reuse 7/6, wideband CQI Reuse 1, wideband CQI 1.2 1.4 1.6 1.8 2.0 2.2 2.4 2.6 Sector SE (b/s/hz)] Figure 3. Performance of the static DL FFR scheme for frequency selective and wideband CQI feedback. the preferred FFR partition. These metrics are used by the BS for dynamic resource allocation among users in a cell. The base stations can adjust the resource metrics periodically to ensure adequate use of resources across partitions. Such updates may happen locally at the BS level unless a consistent trend in use of a particular partition is observed. The BS may report this trend to the central RRM function for it to make the necessary changes in the FFR partition configuration. Thus, the FFR configuration may be managed through faster but localized updates of the resource metrics coupled with a slower but more system-wide change to the FFR partitions and configuration. Performance Results Table 1 illustrates the gains with adaptive FFR for 802.16m systems. The results are based on simulation methodology compliant with IEEE 802.16 evaluation methodology [2]. Further details are provided in [6]. The FFR partitions and cost update mechanisms are based on the market price iteration algorithm in [6]. For OFDMA subchannelization schemes with localized permutations, limited improvement is observed with FFR as substantial gains are already captured with frequency selective scheduling. However, adaptive FFR yields significant cell edge throughput (89 percent) benefit for the case of distributed sub-carrier permutations when compared to a baseline 16e system. The average cell spectral efficiency is also improved by 23 percent. The cell edge performance may be further improved by trading off cell edge gains with gains in average spectral efficiency, by appropriately trading off fairness of the proportional fair (PF) scheduler through a weighted PF-metric. Note that additional techniques such as multi-user multiple-input multiple-output (MIMO), and beamforming with nulling may also be used to achieve the 2 gains in DL performance targets, particularly for the localized permutation. DL FFR IN 3GPP LTE The 3GPP LTE standard allows for very generic FFR schemes to be implemented in the DL, depending on the distribution of mobiles or traffic load. The basic mechanism is the use of a relative narrowband transmit power (RNTP) indicator, which is exchanged between BSs on the X2 interface [7]. The RNTP is a per physical resource block (PRB) indicator which conveys a transmit power spectral density mask that will be used by each cell. This feature results in arbitrary soft reuse patterns being created across the system. For instance, the soft FFR pattern shown in Fig. 2 can easily be created. The idea would be that each cell would have a specific subband for which it will generate low interference with its reduced transmit spectral density. The DL scheduler can exploit this induced frequency selective interference in one of two ways. First, if frequency selective subband CQI reporting is used, these CQI reports will inform the scheduler that there is a particular subband which has low interference and hence improved CQI. Second, if wideband CQI reporting is used to reduce the uplink overhead, the scheduler can be made aware of the identity of the strongest interfering cell for a particular mobile it is serving. This is done through the Event A3 reporting mechanism [8]. Based on knowledge of which cell is causing the dominant interference in the DL, the scheduler can consult the RNTP report from this cell to see which subband is being transmitted at reduced power and hence generating less interference, and can choose to schedule mobile in that subband so that it experiences higher SINR. Performance Results We use a simple static reuse scheme [9] to illustrate FFR performance in Fig. 3. Here seven subbands are utilized in each cell, of which six are transmitted at the normal power level, but the seventh is transmitted with 10 db lower power. We refer to this scheme 88

gamma Sector SE (b/s/hz) Cell edge SE (b/s/hz) 0.2 0.9074 0.0596 0.4 1.0404 0.0606 0.6 1.1471 0.0584 0.8 1.2098 0.0528 1.0 1.2752 0.0471 1.2 1.3286 0.0417 1.4 1.3403 0.0375 Table 2. Sector spectral efficiency and cell edge performance trade-off by adapting IoT levels with gamma. as reuse 7/6. We have used the simulation assumptions for case 1 described in [4] in 10 MHz bandwidth with 10 mobiles/sector, and studied the performance of the case of using both frequency selective subband CQI feedback as well as wideband CQI feedback. The mobile speed is 3 km/hr, for which the subband CQI feedback can effectively exploit frequency selective scheduling gains. Different points on the curve are generated using different degrees of fairness in the DL scheduler. As in the case of 16m, we see that the gain from using the static FFR scheme is relatively low when frequency selective CQI feedback is being used. In the case of utilizing wideband CQI feedback, which would be done to reduce the uplink overhead, or with high mobile speed, we see the gain of the FFR scheme is clearer, but only when the scheduler is tuned to be fair. For example, if we consider the case of high fairness in reuse 1, which obtains a best cell edge rate of 425 kb/s, the FFR scheme can maintain this edge rate while improving the sector spectral efficiency by 25 percent. Alternatively, the FFR scheme can be used to improve the cell edge rate for a given sector spectral efficiency; for the case of proportional fairness we see the cell edge throughput is improved by 30 percent while maintaining the same sector spectral efficiency. UL POWER CONTROL AND FFR Orthogonal frequency-divison multiple access (OFDMA) systems operate with tight synchronization across cells, and the main source of interference is intercell interference. Power control is not typically used in the DL in order to avoid dynamic fluctuation in signal power across resources. However, uplink power control is critical for managing intercell interference. UL POWER CONTROL FOR IEEE 802.16M Open loop power control is used for data transmission in 16m. In this case power levels are adjusted to track long-term fading while fast fading variations are tracked through adjustments in adaptive modulation and coding. Closed loop power control is enabled for control channels. Open loop uplink power control in 802.16m is designed to manage the average interference in the system to some desirable interference over thermal (IoT) level. Specifically, the power update algorithm is derived based on the concept of maximum sector throughput, in which power is increased for a user (mobile station, MS) if the gain in spectral efficiency is greater than the net spectral efficiency loss in other cells due to the increased interference. Several simplifying assumptions are used to derive a power update mechanism, that effectively imposes a target SINR for each user based on its location within the cell [10]. Here users closer to the BS are allowed to maintain a higher SINR target as their transmissions are less likely to interfere with those of the neighboring cells. Key in determining the target SINR is the reciprocity assumption applicable to time-division duplex (TDD) systems, which implies that the uplink interference caused to other BSs by an MS is proportional to the DL interference experienced by the user. Specifically, the target SINR (in db) per user is based on the following equation: SINR 10 ^ min, SINRTarget = 10 10log10 max. 1 γ SIR DL N r Note that the SIR DL serves as an estimate for the uplink SINR and is a ratio of DL signal power vs. interference power measured at one MS receive antenna. The parameter γ controls the level of interference seen by other cells and is set by the BS and broadcast periodically. N r is the number of receive antennas at the BS. SINR min is the minimum SINR in db corresponding to the reliable reception of the lowest modulation coding scheme. The resulting power update equation per subcarrier is P(dBm) = L + SINR Target + NI + Offset. Here L is the path loss between the MS and the BS. NI is the average noise plus interference measured at the BS and broadcast periodically. The offset is the additional MS-specific power correction used by the BS to make additional MS specific adjustments. Performance Results The performance of uplink power control is shown in Table 2. The parameter γ can be used to trade off average spectral efficiency vs. cell edge performance. The IoT level in the system is a function of γ. Further details are given in [10]. We note that with appropriate choice of γ, uplink power control alone can effectively meet the uplink sector and cell edge SE target requirements specified in [1] (sector SE = 1.3 b/s/hz, cell edge SE = 0.5 b/s/hz). UL FFR IN 802.16M The 802.16m standard supports uplink FFR operation [3] by allowing for multiple reuse par- OFDMA systems operate with tight synchronization across cells, and the main source of interference is intercell interference. Power control is not typically used in the DL in order to avoid dynamic fluctuation in signal power across resources. However, uplink power control is critical for managing intercell interference. 89

In our evaluation further application of UL-FFR in addition to uplink power control provides limited additional gain (less than 10 percent). This is because the 802.16m power control algorithm is quite powerful in managing intercell interference and captures most of the performance improvement. titions to be configured within a cell. When UL FFR is used, the above power control algorithm is generalized to allow power adjustment per partition. The power adjustment is determined based on the IoT level to be allowed per partition, which is controlled by introducing a perpartition IoT parameter γ, broadcast by the BS. In our evaluation further application of UL-FFR in addition to uplink power control provides limited additional gain (less than 10 percent). This is because the 802.16m power control algorithm is quite powerful in managing intercell interference and captures most of the performance improvement. UL POWER CONTROL FOR 3GPP-LTE The uplink power control specification in 3GPP LTE allows for a wide variety of power control modes to be utilized. In fact, we show that one of the modes that is possible is quite similar to the power control method described for 802.16m. The baseline uplink power control method for data transmissions on the physical uplink shared channel (PUSCH) in 3GPP LTE is slow, open loop power control. The BS broadcasts a parameter called P 0_PUSCH, which is expressed in dbm and can be set as P 0_PUSCH = SINR Target,Nominal + I 0, where I 0 is the total measured uplink interference (thermal noise plus interference from other cells) level in dbm. The mobile sets its total transmit power (in dbm) as [11] P = 10log10(M) + P 0,PUSCH + αpl, where M is the number of scheduled physical resource blocks (PRBs), which are 180 khz wide in 3GPP LTE, PL is a long-term path loss measurement by the mobile in the DL, and 0 α 1 is a fractional path loss compensation factor broadcast by the BS. Using the expression for P 0,PUSCH, the target SINR achieved by the mobile will be SINR Target = SINR Target,Nominal (1 α)pl. Since α 1, the target SINR always decreases with increasing path loss. The parameter α allows a flexible trade-off between sector throughput (i.e., overall spectral efficiency) and cell edge bit rate; the smaller the value of α, the higher the sector throughput and the smaller the cell edge bit rate, for a fixed IoT level [12]. It should be noted that for transmission of control information on the physical uplink control channel (PUCCH), which carries acknowledgment (ACK)/negative ACK (NACK) and CQI to support the DL, α is always set to 1 as per the 3GPP standard, and a separate value called P 0_PUCCH is broadcast which the mobile will use instead of P 0_PUSCH. In this way a separate open loop target SINR can be maintain for control channel transmissions, which is the same for all mobiles regardless of their path loss; this is desired due to the fact that the QoS is the same for all of the fixed bit rate control channel transmissions. While the baseline uplink power control method is open loop, aperiodic closed loop power control corrections can be sent by the BS in the uplink scheduling grant which the mobile will apply on top of the open loop power control set-point for transmission on the PUSCH. Separate closed loop power control commands for controlling the PUCCH power are sent in the DL scheduling grants. The closed loop power control rate is typically chosen to be much faster for the PUCCH due to the tight QoS constraints and lack of HARQ. The power control mode can be set to accumulate commands received over multiple subframes [11]. For example, the uplink scheduler can maintain an internal uplink target SINR for mobile, and based on the SINR measured on the uplink (on either the PUSCH/PUCCH or the periodic sounding reference signals), the uplink scheduler can send closed loop power control corrections in the accumulated mode to adjust the mobile s transmit power to achieve the desired target SINR. As an example of target SINR settings on the PUSCH, the desired uplink target SINR for a particular mobile can be based on the long-term DL SINR of that mobile similar to the formulation in 802.16m. However, unlike 802.16m, the 3GPP-LTE specification does not allow the DL SINR to be used as part of the open loop power control set-point. Instead, a measure of the longterm DL SINR can be inferred from long-term averages of the CQI, which is fed back in the uplink, and this would be used to set the uplink target SINR internally in the uplink scheduler. One advantage of the DL-SINR-based method over the fractional power control (FPC) method is that it allows differentiation of mobiles that may have low path loss but generate a high amount of interference, such as mobiles located to close the BS but near the sector boundary. One disadvantage of the DL-SINR-based method in the 3GPP LTE context is that it depends on the CQI feedback from the mobile, and there may be notable variability in the CQI measurement reports for the same radio frequency (RF) condition from different mobile manufacturers. One formula for uplink target SINR we have simulated using the DL SINR method takes a form similar to that of the FPC rule, but uses DL SINR instead of path loss as the metric to differentiate the uplink target SINR between users: SINR Target = SINR Target,Nominal (1 β)sinr DL where SINR DL is the DL SINR experienced by the mobile (in db), 0 β 1 is a factor that allows a trade-off between sector throughput and cell edge bit rate, and SINR Target,Nominal is adjusted in order to achieve the desired IoT operating point. When the formula is viewed in linear scale, it is clear that it is similar to the method used in 802.16m, although it can only be achieved through closed loop power control in 3GPP LTE. Performance Results In Fig. 4 we illustrate the performance of both the FPC and DL-SINRbased methods in terms of edge of cell user spectral efficiency vs. sector spectral efficiency. The simulation was performed for a 10 MHz LTE carrier at a 700 MHz carrier frequency using an outdoor Hata suburban path loss model and a 2 km intersite distance with an extended typical 90

urban channel model at 3 km/hr mobile speed. We provide results for two different IoT operating points: 6 db and 10 db. Because this is an interference limited deployment, we see both the sector throughput and cell edge bit rate increase with increasing IoT. The different points on each curve have been generated by selecting different α and β factors for the FPC and DL-SINR-based methods, respectively. We used α = {1, 0.8, 0.7, 0.6} and β = {1, 0.8, 0.7, 0.6, 0.5, 0.3, 0.1}. For each point, the SINR Target,Nominal value was chosen to achieve the specified IoT target value. We see that the DL SINR based method for setting the uplink target SINR does similarly to the FPC method when using high values of α and β in order to get high cell edge bit rates at the expense of sector throughput, but the DL-SINR-based method does better for lower values of α and β when we desire to trade cell edge bit rate for increased sector throughput. UL FFR IN 3GPP-LTE In 3GPP LTE a high interference indicator (HII), which is defined per PRB, can be exchanged between cells via the X2 interface [9] to implement uplink FFR. When the HII bit is set to 1 for a particular PRB, it signifies that this PRB has high sensitivity to uplink interference for this cell; when the HII bit is set to 0 for a particular PRB, it signifies that this PRB has low sensitivity to uplink interference. The exchange of HII reports between cells allows the creation of fractional reuse patterns through uplink scheduling and power control. Upon receiving a HII report from a particular neighbor cell, the uplink scheduler in a given cell can intelligently choose to schedule mobiles that generate significant interference to this neighbor cell only in those PRBs for which the HII report signifies low sensitivity to uplink interference in that neighbor cell. Additionally, the uplink scheduler can reduce the power level of mobiles that need to transmit in the PRBs with high sensitivity to interference for the neighbor cell. The scheduler in a given cell is aware of its mobiles generating significant interference towards a particular neighbor cell because of the Event A3 reporting mechanism, in which a mobile reports to its serving cell the identity of its strongest neighbor cell when the DL signal strength of the neighbor cell comes within a certain range of the DL signal strength of the serving cell [8]. We evaluate a static uplink FFR algorithm based on the inverted-reuse pattern described in [9]. In this scheme we designate either one third or one ninth of the total PRBs in each sector of a three-sector system as an interference-bearing zone. The HII bit is set to 0 for these PRBs to inform neighboring cells that interference will be concentrated in this zone (these PRBs have high sensitivity to uplink interference, while the remaining PRBs have low sensitivity to interference). The way interference from other cells is concentrated in this interference-bearing zone is by configuring a frequency-dependent power restriction via the uplink scheduler only for mobiles located near the cell edge. Mobiles located toward the interior of the cell do not have their transmit power level altered from the normal power control rule. Cell edge SE (b/s/hz) 0.055 0.05 0.045 0.04 0.035 0.03 0.025 0.6 β = 1 α = 1 FPC (lot = 6 db) DL SINR based (lot = 6 db) FPC (lot = 10 db) DL-SINR-based (lot = 10 db) 0.65 α = 0.6 Figure 4. Performance as a function of FPC α and DL-SINR-based β factors, for IoT levels of 6 db and 10 db. Cell edge mobiles can only transmit at their current power level as configured by uplink power control if the mobile can be scheduled within the PRBs of the interference-bearing zone of its strongest neighbor cell. If there are no resources available in this particular zone and this cell edge mobile must be scheduled outside the interference-bearing zone of its strongest neighbor, the scheduler instructs the mobile to transmit with reduced power by issuing an absolute power control command in the corresponding scheduling grant; a value of 4 db is allowed by the 3GPP specifications [11]. The basic proportional fair scheduling algorithm is not altered; rather, the priority metrics for PRBs located outside the interference-bearing zone of a particular mobile s strongest neighbor cell will automatically be lower due to the transmit power reduction. Hence, the scheduler will prefer to schedule cell edge mobiles in the interference-bearing zone of their strongest neighbor. Performance Results In Fig. 5 we illustrate the performance of the UL FFR scheme using both 1/3 and 1/9 inverted-reuse schemes. For this simulation we have used the simulation case 1 assumptions described in [4], which is an interference limited deployment of a 10 MHz carrier at a 2 GHz carrier frequency. The FPC method of power control is used, as it is compatible with issuing absolute power control commands needed for the UL FFR algorithm. An Event A3 threshold of 6 db is used to classify users as cell edge; this classifies approximately 50 percent of the mobiles as cell edge in this deployment. Results are provided for mobile speeds of 3 and 120 km/h. At low speeds, frequency selective scheduling works well as an inherent form of fast interference coordination between cells, because the scheduler obtains information on short-term other-cell interference variations through the channel sounding provided by its own mobiles. Hence trying to impose additional restrictions through the UL FFR mechanism does not provide any additional gain, and in fact can hurt performance as the natural frequency selective β = 0.1 0.7 0.75 0.8 0.85 0.9 0.95 Sector SE (b/s/hz) 91

Cell edge SE (bps/hz) 0.045 0.04 0.035 0.03 0.025 0.02 0.015 0.4 no FFR (3 km/hr) FFR inverted reuse-3 (3 km/hr) FFR inverted reuse-9 (3 km/hr) no FFR (120 km/hr) FFR inverted reuse-3 (120 km/hr) FFR inverted reuse-9 (120 km/hr) α=1 0.45 α=0.8 α=0.7 0.5 0.55 0.6 0.65 0.7 0.75 0.8 Sector SE (bps/hz) Figure 5. Performance of UL FFR for 3km/hr and 120 km/hr using FPC α = 1.0, 0.8, and 0.7. IoT in non-interference-bearing zone is 6 db for all points. scheduler operation is disrupted by power restrictions for particular mobiles on particular PRBs. This behavior is similar to that observed for 802.16m UL FFR. At high speeds the frequency selective scheduling is not effective, and the UL FFR algorithm can provide some improvement. There is approximately a 10 15 percent improvement in sector throughput for a given cell edge bit rate when considering FPC α = 0.8 or 0.7. Notable gains in cell edge rate are only seen if we operate with FPC α = 0.7 with reuse-1 (the low cell edge rate region), in which case the UL FFR scheme improves the cell edge bit rate by 35 percent while maintaining the sector throughput. SUMMARY AND CONCLUSIONS This article has described advanced interference management schemes across IEEE 802.16m and 3GPP-LTE standards for enabling universal frequency reuse. Special focus on advanced RRM schemes, including DL/UL FFR and UL power control techniques, has been provided. It is observed that both IEEE 802.16m and 3GPP-LTE standards utilize similar interference management schemes, and that there are common elements in how each scheme is used. However, the exact details and the relative emphasis of each technique differ across the two standards. For example, 3GPP-LTE uplink power control may be configured to give a similar uplink SINR distribution as 802.16m; however, it requires usage of the closed loop power update mechanism instead of the pure open-loop approach taken by 802.16m. Despite differences in certain details, the schemes in both standards are effective in managing interference and substantially improving the cell edge user performance to meet 2 improvement targets set forth by both standards. ACKNOWLEDGMENTS Several colleagues have developed the ideas and results presented in this article. We especially acknowledge Clark Chen, Hongmei Sun, Hua Yang, Vladimir Kravtsov, Yuval Lomnitz, Ali Koc, Ronghzhen Yang, Tolis Papathanasiou, Wendy Wong, Hujun Yin, Christian Gerlach, Andreas Weber, and Micheal Wilhelm. REFERENCES [1] IEEE 802.16m, System Requirements Document (SDD), IEEE 802.16m-09/0002r10, Jan. 2010. [2] IEEE 802.16m, Evaluation Methodology (EMD), IEEE 802.16m-09-0004r5, 2009. [3] IEEE P802.16m/D4, Advanced Air Interface, Feb. 2010. [4] 3GPP TR 25.814, Physical Layer Aspects for Evolved UTRA. [5] 3GPP TR 36.814, Further Advancements for E-UTRA; Physical Layer Aspects. [6] C. Chen et al., Proposed Text for Interference Mitigation in 802.16m AWD, IEEE C802.16m-09_1022r2, 2009. [7] 3GPP TS 36.423, X2 Protocol Specification. [8] 3GPP TS 36.331, Radio Resource Control (RRC) Protocol Specification. [9] C. Gerlach et al., ICIC in DL and UL with Network Distributed And Self Organized Resource Assignment Algorithms in LTE, Bell Labs Tech. J., vol. 15, no. 3, Fall 2010. [10] R. Zhang et al., Supporting Material for UL OLPC Proposal, IEEE C80216m-09_0845, 2009. [11] 3GPP TS 36.213, Physical Layer Procedures. [12] A. M. Rao, Reverse Link Power Control for Managing Intercell Interference in Orthogonal Multiple Access Systems, IEEE VTC-Fall, 2007. BIOGRAPHIES NAGEEN HIMAYAT (nageen.himayat@intel.com) is a senior research scientist with Intel Labs, where she works on several aspects of cellular system design, covering multitier heterogeneous networks, cross-layer radio resource management, and MIMO-OFDM techniques. Prior to Intel, she was with Lucent Technologies and General Instrument Corp, where she developed standards and systems for broadband access networks. She obtained her B.S.E.E degree from Rice University and her Ph.D. degree from the University of Pennsylvania in 1989 and 1994, respectively. SHILPA TALWAR (shilpa.talwar@intel.com) is a principal engineer in the Communications Technology Laboratory at Intel, where she is conducting research on mobile broadband technologies for increasing cellular capacity and coverage. Specifically, she is researching techniques for advanced interference mitigation, MIMO, and novel cellular topologies. Prior to Intel, she held several senior technical positions in wireless companies over the past 10 years. She graduated from Stanford University in 1996 with a Ph.D. in applied mathematics and an M.S. in electrical engineering. She is the author of numerous technical publications and patents. ANIL RAO (anil.rao@alcatel-lucent.com) is a Distinguished Member of Technical Staff in Alcatel-Lucent s wireless R&D organization in Naperville, Illinois. He received his M.S. and Ph.D. degrees in electrical engineering from the University of Illinois at Urbana Champaign where he held a National Science Foundation graduate research fellowship. His work at Alcatel-Lucent has involved various aspects of system design, performance analysis, and algorithm development for UMTS, HSPA/HSPA +, and LTE. He has actively contributed to both the standardization and product realization of these technologies. ROBERT SONI (robert.soni@alcatel-lucent.com) is a system architecture manager with Alcatel-Lucent, where he leads a team developing next-generation cellular technologies and standards covering physical and MAC layer related techniques for MIMO OFDMA and CDMA systems. He received his Ph.D. degree in electrical engineering from the University of Illinois at Urbana Champaign in 1998. He has also served as an adjunct professor at Columbia University and New Jersey Institute of Technology. 92