Interference Analysis and Mitigation

Because of the random distributions of femto BS network and different operation modes, femto BSs will produce strong interference if the joint interference is not solved appropriately. In [15], six interference scenarios of femto BS networks are plotted:

1) Femto user -> macro BS 2) Femto BS -> macro user 3) Macro user -> femto BS 4) Macro BS -> femto user 5) Femto user -> femto BS 6) Femto BS -> femto user

Fig. 2-3 Interference scenarios in the femto BS networks [15]

Macro user is the user who is served by macro BS and femto user is the user who is served by femto BS. In Fig. 2-3, the black solid lines represent data connection and the read dash lines represent interference.

In femto BS interference mitigation algorithm, most of the analyses and proposals are based on two popular radio access technologies: i) CDMA (Code Division Multiple Access) system, and ii) OFDMA system. Here, we will also introduce the preliminary studies based on their access technologies.

2.4.2.1 CDMA System

To analyze the interference that femto BS creates to the macro/femto BS overlapping network, Chandrasekhar and Andrews provided new mathematical models and analysis for the uplink interference problem in the two-tier CDMA-based CSG femto BS networks [22]. In [22], sectoring receiver antennas and time hopping-CDMA mechanism were proposed to be embedded in the femto BS to avoid mutual interference between the macro BS and femto BSs. Das and Ramaswamy investigated the reverse link capacity of femto cells by modeling inter-cell interference as a Gaussian random variable [23]. For CDMA femto cells, power control or saving a “macro BS only” spectrum was proposed by many studies [24]-[26]. In [27], Arulselvan et al. proposed a “geo-static scheme”, which to enable the femto BS to adjust power level in radio frequency based on its physical distance to the macro BS. Based on this adaptive power control scheme, the femto BS network locally achieved a target data rate that is centrally computed by the network.

2.4.2.2 OFDMA System

Chu et al. proposed a decentralized resource allocation scheme for the OFDMA downlink of a shared spectrum hybrid macro/femto network [28]. In [28], each femtocell randomly selects a subset of OFDMA resources for transmission. The proposed approach in [28] is simple in implementation. However, the random selection approach cannot provide the optimal system performance and QoS guarantee to UEs. To eliminate mutual interference, many studies proposed advanced algorithms baed on different directions, which include: a) Frequency planning, b) Power control, c) SON, d) Optimization problem, and e) Game Theorem.

16

a) Frequency plan

In [29]-[31], the authors discussed how to assign frequency carriers to femto BSs through the given frequency plan in macro BS networks. In these studies, fractional Frequency Reuse (FFR) was applied in the macro BS network. Then, femto BSs were proposed to reuse the macro BS spectrum to improve the spectrum efficiency. In [32], Ghosh et al. analyzed the improvement of FFR in the heterogeneous network. Their results can also be applied to the FFR in the macro/femto BS overlapping network.

However, there is one implementation problem in this approach: How to decide if the femto BS could reuse the macro BS spectrum? To solve this problem, G¨uvenc et al.

proposed a “interference-limited coverage area” (ILCA), which is an area within a contour where the received power levels from the macro BS and femto BS are the same [33] . The ILCA will be compared with a threshold (e.g., the area of a user’s premises); if it is larger than the threshold, the femto BS is allowed for co-channel operation (i.e., it is in outer region). Otherwise, femto BS is in the inner region and it cannot reuse the macro BS spectrum. In [34], based on the objective to increase the system spectral efficiency, Bai et al. discussed the tendency of macro BS/femto BS to reuse or to partition the spectrum when the serving user is in different locations of the macro BS coverage. Then, a hybrid frequency allocation algorithm was proposed to improve the spectrum efficiency in the macro/femto BS overlapping network.

b) Power control

In [35], Li et al. formulated the downlink power control problem for femto BSs that operate in the same frequency carrier with macro BSs. Both centralized and distributed solutions were given jointly with a dynamic channel re-allocation procedure to assure the QoS of users. In [36], a distributed utility-based SINR adaptation was proposed for femto BS networks to alleviate cross-tier interference at the macro BS, which was interfered by overlaid femto BSs.

To briefly summarize, the interference elimination approaches above proposed to eliminate the interference between the macro/femto overlapping network by the popular techniques in the cellular netowrks, such as directional antennas, power control, and frequency participation. Next, we will introduce the studies about how to decrease the mutual interference through SON algorithm.

C) SON algorithm

The approaches of SON algorithms can be divided into three cooperation levels, which are shown in Fig. 2-4. Note that the three levels can co-exist in one control algorithm.

i) Self-Measurement

In self-Measurement, femto BS adjusts its operation parameters by measuring the environment itself. In [37], several heuristic frequency assignment schemes were proposed and compared. Based on their simulations, the LIP (Least Interference Power) scheme, which new femto BS selects the frequency band that has the minimum the received total interference power at the receiver side of itself, is the best practical scheme.

However, the performance of LIP scheme is sensitive to the order of femto BSs turning on its radiation power. Furthermore, each femto BS can only access one frequency carrier, which limits the system capacity. To solve this problem, Garcia et al. proposed an

“autonomous component carrier selection” approach [38]. First, each femto BS selects one least interfered primary carrier from a set of carriers based on its measurement. Then, allocation of additional secondary component carriers is possible if and only if the performance impact on neighboring cells is estimated to be acceptable.

In [39], Sundaresan and Rangarajan proposed a distributed random access scheme (DRA). Accoring to DRA, the femto BS decides which resource blocks it occupies based on a hash table. The hash table is generated individually by each femto BS and the size of hash table is decided by the interfering degree, which is also measured by the femto BS itself. Femto BSs will rehash the hash table in the collided resource blocks. The details of resource blocks of OFDM system will be explained in Chap. 5.

Fig. 2-4 Different cooperation levels of SON algorithms

18

Many studies also applied cognitive radio technique to realize the SON algorithm [40]

-[43]. In [40], Lien et al. proposed a cognitive radio resource management (CRRM) scheme for femto BS networks. In CRRM, femto BSs periodically sense the channel to identify which resource block is occupied by the macro BS network. In subsequent data frames, femto BSs only allocate “non-occupied” resource blocks sensed in the sensing phase. To achieve optimal spectrum utility, femto BSs are required to record the following parameters: (i) the traffic loading of the macro BS network, (ii) radio resource allocation correlation probability of the macro BS network, and (iii) percentage of correlated radio resource allocation of the macro BS network. In [41], a localized dynamic spectrum access approach was proposed in a macro BS/femto BS overlapping network. In [41], femto BSs reuse the spectrum for macro-PU/femto-PU, which are primary users served by macro BS and femto BS respectively, by sensing the idle spectrum. Simulation results showed that throughput improves if spectrum sensing is achieved by femto BSs. It is because femto BSs are usually with better sensing capability.

In [42], Jin et al. proposed to combine cognitive radio and multi-hop cooperative communication in the macro/femto BS overlapping network. By requiring every wireless device to be equipped with frequency-agile spectrum sensing units, Jin et al. developed an optimization framework for location-aware cooperative resource management, with jointly employing power control, multi-hop cooperative communication and flow management techniques. Based on stochastic geometry and homogeneous Poisson point process (HPPP), Cheng et al. proposed several corresponding downlink spectrum sharing schemes between femto BSs and macro BSs as well as among femto-BSs [43]. Moreover, by requiring femto BS to measure location information and avoidance region, the proposed Distance Sense Multiple Access (DSMA) and controlled-underlay schemes in [43] provided much more throughput than that in traditional interweave and slotted Aloha schemes.

To summatize, the Self-Measurement approach is easy to be implemented. Without information exchange between UEs and other BSs, the femto BS does not produce much control overhead to the backhaul network. However, the drawback of Self-Measurement approach is: the measurement from the femto BSs does not represent the measuremet of users. To fulfill users’ QoS requirements, femto BSs need the measurement reports from users.

ii) Users’ Reports

In cellular network, users’ reports are already applied in many control algorithms, such as handover process or frequency carrier selection [15]. In the femto BS networks, users’

report is extended to modify the control parameters of femto BS network. In [44], Claussen et al. proposed a novel mobility-event based self-optimization approach to adjust the femto BS radiation power. In this approach, they tried to minimize the increase of unnecessary mobility events, such as passing and handover events. It was shown that mobility event based self-optimization of coverage can both significantly reduce the total number of mobility events caused by femto cell deployments and improve the indoor coverage. In [45], L´opez-P´erez et al. proposed two SON approaches for femto BS in the downlink direction. One is the femto BS selects the least interfered frequency carrier it detects from neighbor BSs. Another one is femto BS selects the least interfered frequency carrier that users measure. Simulations result showed the user based approach gets better performance. Although the femto BS networks can get better performance from the users’ report, the number of calculations in each femto BS would increase exponentially with the number of users and the number of carriers. Furthermore, requiring UEs to perform measurements may enhance the power consumptions of UEs, which are typically power limited.

iii) Inter-BS Cooperation

To optimize the bandwidth efficiency and resource allocation, many researches proposed the capability that BSs exchanging information with neighbor cells through air links or backhaul connections [46]-[49]. In [46], Amirijoo et al. proposed that network detects an outage area autonomously based on measurements, which from both UEs and neighbor BSs. Then, the network alters the configuration of surrounding BSs to compensate the outage-induced coverage. In [47], Li et al. enhanced the joint cooperation of femto BSs by also requiring the information exchange between BSs.

However, even the system performance can be improved by the inter-BS information exchange, the propagation delay of backhaul connection is too long to allow dynamic cooperation between femto BSs and macro BSs [48]. To facilitate the information exchange, information exchange through air links was proposed in many studies. In [49], Adhikary et al. proposed a novel approach for the femto BSs to reuse the macro BS spectrum by listening to the resource allocation map, which is broadcasted by macro BS over the time slots. Furthermore, the femto BS also gets the locations information of UEs

20

through Global Positioning System. Then, femto BS will reuse the macro BS spectrum by limiting joint interference.

To briefly summarize, the inter-BS cooperation improves the system performance futher because of more information gathered during the process. However, it may also increase the signaling overhad and the loading of compulations to the cellular netowrks.

d) Optimization problem

Interference problem can also be solved by using the tools of optimization problem. In [50], L´opez-P´erez et al. proposed “dynamic frequency planing” by modeling the frequency allocation problem as a mixed integer programming. In the backhaul network, a centralized controller is responsible to gather all the measurements from users and BSs.

Greedy algorithms were used in the simulations and the result showed the macro BS femto BS joint cooperation would conduct the best performance during the simulation. In [51], femto BSs were grouped based on the mutual interference information. Then, a central controller determined the minimum number of orthogonal sub-channels for each group to provide target performance. The transmission power of each femto BS was adjusted based on the received signal strength indication (RSSI) in a distributed manner.

However, the above optimization problems require high complexity and centralized computations, which increase the difficulty to be implemented in the femto BS network.

e) Game theorem

Game theorem is also applied by many researchers to analyze spectrum allocation problem in the macro/femto BS overlapping network. In [52], Chen et al. proved the existence of the unique optimal solution of the channel allocation problem. Furthermore, they also proposed A DANCE mechanism for a general femtocell channel allocation problem. In [53], Lien et al. proposed the cognitive radio resource management scheme for femtocells to mitigate cross-tier interference. Under such cognitive framework, a strategic game was further developed for the intra-tier interference mitigation.