Chapter 3 Game Theoretic Spectrum Access with Intra-cell
3.4 Computer Simulations
The simulation environment is shown in Fig. 3.4 where the simulation parameters are listed in Table 3.2. In the environment, there is a femtocell overlaying on a macrocell. In order to verify the proposed method in a densely populated scenario, we adopt the maximum number of supported FUEs, which is mentioned in specifications provided by a femtocell integrated circuits manufacturer in [9]. We increase the number of FUEs scattered in the FAP coverage from four to 32 by four each time. In Fig. 3.5, the utility values per RB for the group spectrum access with the Bayesian game for two and four FUEs per group are shown, and both improved obviously from the original Bayesian game and from the lower bound of no coordination. Besides, the proposed group spectrum access also approaches the performance in both small and large number of FUEs obtained in the centralized scheme where the exhaustive search is adopted.
The results show the ability of providing simultaneous supports of a larger number of FUEs in the LTE environment. Fig. 3.6 shows the utility values per FUE, and we can
Step 1: All FUEs transmit random access preambles to FAP Step 2: FAP feedbacks group index and system information Step 3: FAP transmits access admittances
Step 4: FUEs who have admittances start accessing
Step 5: If any two FUEs access the same RB, the FAP starts the contention resolution: the one FUE with the highest priority in group is picked up to access the RB
Step 6: Break the loop if all RBs are used up
Step 7: Go to Step 3 for different groups until all groups have already performed the access of RBs
see that if the number of FUEs is larger than two, utility values in the Bayesian game are close to zero. In contrast, the proposed method can still provide positive utility values and be close to the values provided by the centralized scheme.
Table 3.2 Simulation parameters
Parameter Value
FUE transmit power 26 dBm
Path loss model [1] 3GPP TR 36.814 v9
Fading channel Rayleigh
Resource block bandwidth 180k Hz
The maximum number of FUE [9] 32
The number of available RBs 4
Thermal noise PSD [10] -174 dBm/Hz
-600 -400 -200 0 200 400 600
Fig. 3.4 Two-tier femtocell network
24
Bayesian game with Group SA: 2 FUEs/group Bayesian game with Group SA: 4 FUEs/group Bayesian game
No coordination
Fig. 3.5 Comparison in utilities per RB between different schemes
1 2 3 4 5 6 7 8
Bayesian game with Group SA: 2 FUEs/group Bayesian game with Group SA: 4 FUEs/group Bayesian game
No coordination
Fig. 3.6 Comparison in utilities per FUE between different schemes
3.5 Summary
In this chapter, we give an introduction of the Bayesian game, and also describe the vital requirement for a distributed mechanism for the spectrum access. To fulfill the requirement, we propose a Bayesian game theoretic spectrum access with intra-cell interference avoidance. The new spectrum access scheme can be adapted to the existing LTE protocol architecture. Compared with the centralized scheme, the proposed scheme improves greatly from the original Bayesian game and exhibits similar utility performances to the centralized scheme.
26
Chapter 4
Multicell Spectrum Access with Inter-cell Interference Avoidance
In this chapter, we incorporate a multicell interference avoidance scheme into the proposed spectrum access in Chapter 3. In the proposed scheme, we exploit some self-organizing techniques. Based on these techniques, the femtocells can cooperate to avoid excessive inter-cell interference in femtocell networks and the system capacity can be improved.
This chapter is organized as follows. In Section 4.1, the motivation of inter-cell interference avoidance is presented. The proposed scheme is described in Section 4.2. In Section 4.3, complexity and overhead analysis is shown. Section 4.4 presents computer simulation results, and Section 4.5 summarizes this chapter.
4.1 Motivation
To improve the spectrum efficiency, LTE-Advanced proposed the deployment of a large scale cost-effective low-power femtocells in local area environments. Local area environments include indoor office scenarios, or outdoor hotspot scenarios with several low-power FAPs. One of the major challenges for the deployment is the inter-cell interference due to the unplanned deploying of femtocells, which can degrade the effectiveness of femtocells. When both tiers share the whole spectrum, indoor and
outdoor user communications are affected by interference from undesignated femtocell and macrocell devices. This problem is more severe when FAPs are randomly deployed by their subscribers. In [20] the authors point out that femto-to-femto interference becomes an important issue for the indoor performance, especially when femtocells are densely deployed. Another challenge for the deployment is, in the femtocell networks, the high speed backhaul cannot be implemented without a direct X2 interface [21].
Therefore, the centralized interference management is not viable due to the heavy information exchange required and intolerable feedback delays.
The optimal configuration will depend on the offered traffic and the location of FUEs, which are likely to be time-variant. Moreover, a subset of the available RBs is enough to guarantee the quality of service (QoS) of a FUE, so a scheme to allocate RBs to serve many FUEs to achieve the optimal spectrum utilization is required. Generally, this spectrum utilization problem in a dynamic multicell multiuser environment is a non-linear, non-convex NP-hard optimization problem [24]. In summary, the centralized spectrum access for the two-tier femtocell networks becomes impractical, and a distributed scheme should be employ for this problem.
In the LTE frequency reuse scenario, co-channel deployment is attractive to operators due to low cost and backward compatibility [21]. In a co-channel deployment, Femtocells are enabled with some cognitive techniques developed to sense their surroundings and change their spectrum access operations to minimize interference to the macrocell. Since femtocells are overlaid within macrocell networks, this cognitive techniques can be exploited to avoid cross-tier interference. After sensing macrocell activities, the utilization of spectral resources is available, and FAPs can exploit unoccupied spectrum. As a result, we only concern the co-tier inter-cell interference issue within femtocell networks.
28
Some techniques for the self-organizing (SON) network proposed by 3GPP provides some potential solutions. SON enables femtocells to integrate themselves into the network of the operator, learn about their environment (neighboring cells, interference) and change their actions accordingly. Generally, the techniques of SON rely on UE measurements in Releases 8 [24], and support some important SON related objectives, include interference control, coverage optimization, and energy saving management.
4.2 Proposed Multicell Spectrum Access
There are several studies focusing SON techniques for spectrum access, and the SON functionalities adopted in this thesis are the interference measurement report by FUEs, the spectrum utilization reports and broadcasting and sniffing functionalities by FAPs [2]. Borrow the idea from [23], we propose a distributed interference measuring technique. The fundamental principle of the technique is the estimation of potential interference based on downlink received reference signal power (RRSP) strength level measurements performed by FUEs. In the proposed technique, FAPs need to transmit reference signals in the coordinated subframes, which occupy a limited portion of the allocation phase mentioned in Chapter 2. FUEs could learn the potential interference to other femtocells by reciprocity of RRSP, and each FAP gathers the interference information from FUEs. This information gathered by FAP is used in the contention resolution process, allowing each FAP to select the most attractive spectrum access in the view of system capacity. In summary, our scheme relies on the existing UE interference measurement reports and information exchange processes, which means the minimal changes to the standard.
After the procedure of obtaining interference measurement reports, we refer to the
dynamic fractional frequency reuse (FFR) [22] and propose an interference avoidance scheme by priorities of femtocell. In LTE-Advanced, OFDMA systems support FFR for interference mitigation and divide frequency and time resources into several resource sets especially in the heterogeneous network with severe interference. The FFR should be dynamic and be adapted to variant traffic based on interference conditions obtained by FAPs. To make the frequency reuse be beneficial, the capacity gain from the increase in SINR must be able to overcome the loss from spectrum. Since the effect of SINR is scaled by a logarithm function, the benefit can only be obtained when the SINR is not too high. A trade-off between bandwidth and SINR is depicted in Fig. 4.1. Here FAP a and FUE a are for femtocell a, and FAP b and FUE b are for femtocell b.
Fig. 4.1 Illustration of two femtocells with asymmetry
30 examine the potential received interference power to neighboring cells with the received signal power as a threshold for interference avoidance.
The deployment of femtocells might be random and the geometric asymmetry of deployment causes some femtocells have more interference to others. This can be known by the interference measurement reports. To minimize the potential interference in the femtocell networks, it is obvious that the femtocell with the lowest interference to others is given with the highest priority. The priority means the right to access the spectrum, and for a FAP, the higher priority, the more RBs can access. As the result, the interference can be reduced, because less potential interference occurs. Incorporate the spectrum utilization reports, interference measurement reports, and femtocell priorities into the scheme we proposed in Chapter 3, we depict a sequence diagram in Fig. 4.2.
The sequence diagram of the proposed procedure is based on the architecture of LTE. We add four phases into the sequence diagram in Fig. 3.1 in Chapter 3 for the interference reports and spectrum usages reports. In the phase one, FUE chooses a random access preamble and transmits it on PRACH. In the next phase, when FAP receives the preamble, FAP gives a response which includes a group index for the FUE.
In the phase three, the FUE measures the received reference signal strength from FAPs and computes the interference reports by reciprocity. In the phase four and five, the FAP broadcasts and sniffs the interference reports from other FAPs. In the next phase, FAP
also sniffs the RB usages from the prior FAPs. Based on the sniffed RB usages and the interference reports, the FAP can ensure that the interference from its serving FUEs to other FAPs is not excessive. If all the prior femtocells perform the spectrum access, the FAP will enter the phase seven, which is transmitting group access admittances to the group from the index one in sequence, and in the phase eight the FUEs which receive the admittances can request a RB from the FAP. In the final phase, an interference-checking function is added to the contention resolution phase in Fig. 4.2, i.e., if there are excessive interferences to other FAPs, the FAP will check and close the transmission.
Fig. 4.2 Sequence diagram of proposed multicell spectrum access
32
Based on the sequence diagram in Fig. 4.2, we depict flowcharts of spectrum access by the FUEs and FAP in Fig. 4.3 and Fig. 4.4, respectively. Different from the flowchart of spectrum access in Chapter 3, we install the step three in Fig. 4.3 for the FUEs. The step three is for the interference reports exchange between the FAP and FUEs.
We also install the step two and the step three to the flowchart in Fig. 4.4. In the step two, FAP is able to collect and broadcast the interference reports from FUEs. In the step three, FAP collects and sorts the interference reports from other FAPs by the sniffer function. The procedure of the proposed spectrum is shown in Table 4.1.
Fig. 4.3 Flowchart of spectrum access by FUEs in femtocell networks
Fig. 4.4 Flowchart of spectrum access by FAP in femtocell networks
34
Table 4.1 Procedure of the proposed multicell spectrum access in femtocell networks
4.3 Complexity and Overhead Gain Analysis
We evaluate the complexity and overhead of the proposed spectrum access presented in Section 4.2. The complexity is measured in terms of the average number of floating point operations to floating points, excluding the complexity and overhead of the operations defined in the original LTE protocol architecture. All real additions, subtraction, multiplications, and comparisons are treated equally, and divisions are weighted by two as shown in Table 4.2.
The complexity of the proposed scheme comprises three parts: the complexities of calculation for cost function, of comparison with threshold, and sorting, and is shown as follows:
1
2
FAP FUE aRB FAP FUE FAP aRB
N N N N N N N (3.1)
Step 1: All FUEs transmit random access requests to FAP Step 2: FAP feedbacks group index
Step 3: FUEs feedback interference reports Step 4: FAP broadcasts interference reports
Step 5: FAP sniffs interference reports from other FAPs Step 6: FAP sniffs RB spectrum usage from prior FAPs
Step 7: If all prior FAPs’ usage is obtained, FAP transmit access admittances and system information
Step 8: FUEs who have admittances start accessing
Step 9: If any two FUEs access the same RB, the FAP starts contention resolution
Step 10: Break the loop if RBs are all used up
Step 11: Go to Step 7 for different groups until all groups have already performed the access of RBs
Table 4.2 Complexity weight of different operations
The transmit overhead is measured as the overhead per second, and we assume the two phases of spectrum access should be renewed in a coherence time, and the coherence time of a object with the speed of 5 km/hour is 4 10 3 second.The major part of transmit overhead of the centralized exhaustive scheme is for the channel state information, which is shown as follows (in real numbers):
3
/ 4 10 250
FAP FUE aRB FAP FUE aRB
N N N N N N (3.2)
The transmit overhead of the proposed scheme comprises three terms: the terms for the group index and access admittance, the interference measurement reports, and the spectrum usage reports, and is shown as follows (in real numbers):
2
250 2 NFAP NFUE NFAP NFUE NFAP NFAP log NaRB (3.3) The major difference of transmit overhead of both schemes is for the centralized scheme, the channel state information is needed on every RBs. In contrast, the number of overheads of the proposed scheme is independent of the number of RBs.
4.4 Computer Simulations
In this section, the simulation results of the proposed scheme are presented. The simulation environment is shown in Fig. 4.5, where the simulation parameters are listed in Table 4.3. In the environment, the FAPs are closely located, and the FUEs are
36
randomly distributed within the femtocell coverage. In order to verify the proposed method in a densely populated scenario, we adopt the maximum number of supported FUEs [9]. We increase the number of FUEs scattered in the FAP coverage from four to 32 by four each time. Fig. 4.6 shows comparison between multicell spectrum access with interference and without interference, and we can see that the inter-cell interference really degrades the capacity greatly, which leave a room for interference avoidance and capacity improvement. In Fig. 4.7, the capacity per channel (bits/sec/hertz) for the proposed spectrum access with inter-cell interference avoidance (IA) is shown, and is improved from the spectrum access without IA, and both of these are away from the lower bound of no coordination. Besides, the proposed spectrum access with IA also approaches the performances in both small and large numbers of FUEs obtained in the centralized scheme where the exhaustive search is adopted. The results show the ability of providing simultaneous supports to a larger number of FUEs in the two-tier femtocell networks. Fig. 4.8 shows the capacity per FUE (bits/sec), and we can see that with the increase in number of FUEs, the capacities provided by the proposed schemes and the centralized scheme decrease, and when the number of FUEs per RB is larger than two, the difference between them is small. However, the capacity of no coordination scheme is close to zero in both small and large numbers of FUEs, which shows the benefits in capacity from coordination in the multi-cell spectrum access.
The comparisons in complexity and overhead are shown in Fig. 4.9 and Fig. 4.10, respectively. Referring to the specification of spectrum release for LTE-Advanced in Taiwan [19], the bandwidth for an operator is up to 45 MHz, and the number of divisions of the total bandwidth by the minimum bandwidth configuration is 37.
Applying this figure into simulations, we can see that in Fig. 4.9 the complexity for the centralized scheme grows exponentially with the number of FUEs, while the
complexity for the proposed scheme grows linearly. In Fig. 4.10, the transmit overhead of the centralized scheme and the proposed scheme both grow linearly with the number of FUEs, and the exponent of the latter is less than the former by one, which means the transmit overhead of the proposed scheme is only tenth of the centralized scheme.
Table 4.3 Simulation parameters
Parameter Value
FUE transmit power 26 dBm
Path loss model [1] 3GPP TR 36.814 v9
Fading channel Rayleigh
Resource block bandwidth 180k Hz
The maximum number of FUE [9] 32
The number of available RBs 4
The number of femtocells 4
Thermal noise PSD [10] -174 dBm/Hz
-600 -400 -200 0 200 400 600
Fig. 4.5 Multiple femtocells in two-tier networks
38
Fig. 4.6 Comparison in capacity per channel
1 2 3 4 5 6 7 8
Fig. 4.7 Comparison in capacity per channel
1 2 3 4 5 6 7 8
Fig. 4.8 Comparison in capacity per FUE
4 8 12 16 20 24 28 32
Fig. 4.9 Comparison in complexity
40
4 8 12 16 20 24 28 32
104 105 106 107
No. of FUEs
Transmit overhead (numbers/sec) Centralized scheme Proposed scheme
Fig. 4.10 Comparison in overhead
4.5 Summary
In this chapter, we first describe the requirement for distributed inter-cell interference coordination. To fulfill the requirement, we propose a multicell spectrum access with inter-cell interference avoidance. The new spectrum access scheme is based on the scheme proposed in Chapter 3 and can be adapted to the existing LTE protocol architecture. Analyses of complexity and overhead are also presented. Compared with the centralized scheme, the proposed scheme improves over the spectrum access without inter-cell interference avoidance and exhibits similar capacity performance to the centralized scheme.
Chapter 5
Conclusions and Future Works
To fulfill the increasing demands on the spectrum resource and to develop an efficient communication system in the next generation wireless communication systems, the femtocell network has been proposed. Femtocells feature no coordination center, low computation ability, and limited backhaul, which make the spectrum access hard to be efficient. Besides, uncoordinated network planning also introduces the interference issue. In this thesis, we propose a novel distributed multicell spectrum access with interference avoidance.
The network architecture and system model for the two-tier femtocell are introduced in Chapter 2, where two SON functionalities built in the system are also mentioned. The channel model of links between FUEs and FAPs is depicted, and the spectrum resource unit suggested by LTE is also adopted. We also briefly describe the mechanism of spectrum access in LTE and differences between femtocell networks and traditional macrocell networks. In the following Chapter 3 and 4, we propose a game theoretic distributed spectrum access in the femtocell networks.
In Chapter 3, we first give an introduction of the game theory, and also describe the vital requirement for a distributed mechanism. To meet the requirement, we propose a Bayesian game theoretic spectrum access with intra-cell interference avoidance. The new spectrum access scheme can be adapted to the existing LTE protocol architecture.
Compared with the centralized scheme, the proposed scheme improves greatly from the
42
original Bayesian game and exhibits similar utility performances to the centralized scheme.
In Chapter 4, we first describe the requirement for inter-cell interference
In Chapter 4, we first describe the requirement for inter-cell interference