Spatial Reuse Algorithm Using Interference Graph in Millimeter Wave

Spatial Reuse Algorithm Using Interference Graph in Millimeter Wave Beamforming Systems Ohyun Jo and Jungmin Yoon

This paper proposes a graph-theatrical approach to optimize spatial reuse by adopting a technique that quantizes the channel information into single bit submessages. First, we introduce an interference graph to model the network topology. Based on the interference graph, the computational requirements of the algorithm that computes the optimal spatial reuse factor of each user are reduced to quasilinear time complexity, ideal for practical implementation. We perform a resource allocation procedure that can maximize the efficiency of spatial reuse. The proposed spatial reuse scheme provides advantages in beamforming systems, where in the interference with neighbor nodes can be mitigated by using directional beams. Based on results of system level measurements performed to illustrate the physical interference from practical millimeter wave wireless links, we conclude that the potential of the proposed algorithm is both feasible and promising. Keywords: mmWave, Spatial reuse, Graph coloring, Scheduling, Beamforming, Modem implantation.

Manuscript received Apr. 15, 2016; revised Dec. 22, 2016; accepted Jan. 4, 2017. Ohyun Jo (corresponding author, [email protected]) is with the 5G Giga Communication Research Laboratory, ETRI, Daejeon, Rep. of Korea. Jungmin Yoon ([email protected]) is with the Department of Electrical Engineering and INMC, Seoul National University, Rep. of Korea. This is an Open Access article distributed under the term of Korea Open Government License (KOGL) Type 4: Source Indication + Commercial Use Prohibition + Change Prohibition (http://www.kogl.or.kr/news/dataView.do?dataIdx=97).

ETRI Journal, Volume 39, Number 2, April 2017 https://doi.org/10.4218/etrij.17.0116.0035

© 2017

I. Introduction Link scheduling algorithms have been widely studied to enhance the performance of wireless communication systems [1]–[4]. Scheduling techniques based on fully informed channel quality information are considered as conventional and straightforward solutions that can guarantee optimal scheduling results. However, the computational complexity and resource overhead of these algorithms are problematic because the best solutions require large amounts of channel information to be sent over the entire network, and the solutions must contend for a single central node to accommodate the requirements for an optimal spatial reuse scheduling procedure. Scheduling algorithms that model the network as vertices of a graph offer advantages relative to interference-based scheduling for several reasons: the implementation is relatively easy, and the computational complexity is, at worst, linear, and the resources required and performance measurement overhead can be minimized [1], [3], [5]. In addition, the interference edge of a graph model can present one-bit quantization information about the interference level. Thus, graph-based algorithms are more feasible to implement. By using a proper threshold value for generating the interference edge, we can avoid performance degradation while obtaining significant gain from spatial reuse. On the other hand, beamforming technology is regarded as one of the key features to enhance system performance by using intelligent scheduling algorithms [6]–[8]. Beamforming technology in millimeter-wave bands is also considered as a key enabler for 5th generation communications systems [9]– [13]. It has been adopted for application to wireless communication systems in order to achieve a maximal high throughput. According to the well-known Friis law [14], millimeter wave signals incur much larger signal attenuation during propagation. As one of the most effective solutions to Ohyun Jo and Jungmin Yoon

255

the large propagation loss, beamforming plays a key role by preventing the transmission of energy from spreading arbitrarily in all directions. Using an array antenna, electrical energy can be concentrated toward the intended destination point such that the electrical signals do not reach unintended areas. In addition, the beamforming feature ensures robust multi-gigabit communication over millimeter wave bands by using a wide bandwidth. The majority of future wireless communication systems including 5th generation systems and the next generation of Wi-Fi systems, such as IEEE 802.11 ad/ay, have already adopted the millimeter wave beamforming technology to exploit the advantages of the beamforming feature. Within the literature, recent mmWave beamforming technologies and the beam training techniques for application to practical wireless communication systems, including 5th generation communications systems, are described [9], [10], [13]; however, this research does not include an efficient spatial reuse scheme. The usage model can be limited because the practical mmWave radio channel for wireless communications is not considered at the system level. While there may exist research in the field of efficient resource management and cooperation with millimeter wave communications [15]–[17], our study considers a practical system model and mmWave radio channels that build upon the feasibility and complexity limitations of this previous research in the application to practical wireless communication systems. To reduce the large amount of feedback information from link scheduling that includes millimeter wave beamforming techniques, a graph model can be constructed for the proposed scheduling algorithm [18]–[24]. Building on this research, we propose a graph-based scheduling algorithm that uses a one-bit quantized indicator of the interference level to reduce the signaling overhead. Relative to this graph model, a vertex corresponds to a wireless communication link. Vertices are connected by an edge when considerable interference is detected between the corresponding wireless links. Thus, we obtain an interference graph that reflects the real wireless communication environments. Then, resource allocation can be efficiently performed to improve spatial reuse efficiency based on the interference graph. With the application of the theoretical analysis to the practical wireless communication systems [25]–[32], we devise, propose, and analyze a graphbased algorithm. Then, we show that the proposed scheduling procedure can avoid link corruption while obtaining significant gain from spatial reuse.

II. System Model In this section, we describe the system model that represents 256

Ohyun Jo and Jungmin Yoon

v7

7 5

e8

6

1 v2

e3

2

v1

3

e6

e1

4

e7

e9

v6

e4

e2

v5

e10

v4 e5

v3

Fig. 1. Colored interference graph obtained from a sample network.

a graph-based scheduling algorithm. The proposed scheduling algorithm is designed from an interference graph with a network topology as shown in Fig. 1. Here, the interference graph that represents the real network topology can be defined as G (V , E ), where V = {v1, v2, … , vN} is the set of vertices in which vi represents a wireless communication link, and E = {e1, e2, … , eM} is the set of edges in which ei represents the indication of interference between the two wireless links. If any two wireless links interfere with each other, the corresponding vertices are connected with an edge. Each communication link corresponds to a vertex v and is composed of a source node and a destination node. An interference edge consists of two vertices represented as vi and vj. Each vertex vi has a serving node svi and destination node d vi . The graph that reflects the network topology is obtained by the following rule.









(1)

vi : svi , d vi , v j : sv j , d v j .

If SNRsv , dv   c or SNRsv i

j

j

, dvi

  c , v j  vi  E and vi



 v j  E , where SNRsv , dv   svi / PLSvi , dv j i

j



N 0 ,  c is

a threshold value for sensing interference,  sv is the i

transmission power of node svi , PLSv , dv i

j

is the path loss

between source station Svi and destination station d v j , and N0 is the background noise power. Applying the above rule, we can define an interference graph G consisting of several vertices and edges.

III. Graph-Based Algorithm for Optimal Spatial Reuse Graph theory can be applied to wireless communication systems to enhance scheduling performance. In particular, we apply the well-known graph coloring problem to support the map coloring problem. This problem was proposed in the nineteenth century and an algorithm published in 1976 [33]. The chromatic number of an interference graph G is the minimum number of colors required for a proper graph coloring; that is, no two adjacent vertices share the same color. ETRI Journal, Volume 39, Number 2, April 2017 https://doi.org/10.4218/etrij.17.0116.0035

In this paper, the chromatic color is denoted by  (G ) . For an interference graph G with n vertices, the following inequality is always true. 1   (G )  n.

(2)

If each of the communication links in a network does not interfere with other communication links, the interference graph will possess no edges and chromatic number will be 1. On the other hand, if each communication link of a network interferes with all other communication links, the interference graph is a complete graph and the chromatic number is n. By applying Vizing’s Theorem [34], we can improve the upper bound of  (G ), by reformulating the bound in terms of (G ), the maximum number of interference links of any communication link. Proposition 1 is a practical application of graph theory to the spatial reuse algorithm for wireless communications systems. Variable (G ) denotes the maximum degree of any vertex. Proposition 1. For any interference graph G generated for a given network topology,  (G )  (G )  1. Proof. Select a vertex, vn, such that the vertices of G are v1 , v2 ,  , vn so that if i < j, then d (vi , vn )  d (v j , vn ). Here, d is defined as the Euclidean distance between the two vertices. Then, we can enumerate the vertices ordered by distance from vn in descending order. We use integers 1, 2, … ,  + 1 as colors, where  is the chromatic number. We first color vi with 1. Then for each vi in the order, we color vi with the smallest integer that does not violate the proper coloring requirement, and is distinct from the colors already assigned to the neighboring vertices of vi. For i < n, we claim that vi is colored with one of 1, 2, … , . For 1 < i < n, vi has at least one neighbor that is not yet colored, that is, a vertex closer to vn on a shortest path from vn to vi. Thus, the neighbors of vi use at most  – 1 colors from the colors 1, 2, … , , leaving at least one color from this list available for vi. Once v1, … , vn–1 have been colored, all neighbors of vn have been colored using the colors 1, 2, … , , so color  + 1 may be used to color vn. ■ Now, we color each vertex of the interference graph as follows. We define function F:V→N that maps a vertex to an arbitrary color in N, which is set of available colors, such that adjacent vertices cannot be assigned the same color as an adjacent vertex. Thus, if vi → vj  E, then F (vi )  F (v j ). In order to compute the maximal spatial reuse factor, we compute the chromatic number  (G ), which is defined as the minimum number of colors required to color G by applying the graph-coloring algorithm. ETRI Journal, Volume 39, Number 2, April 2017 https://doi.org/10.4218/etrij.17.0116.0035

Find maximum complete graph of G, then  (G) = k Determine k as the reuse factor Y

No more complete graph with k? N Find the next maximum complete graph with k' vertices Find a vertex whose reuse factor(n*) is already determined, then n' = 1 – (1/n*) and i++

N

All vertices are examined? Y Determine n'/(k'– i) as the reuse factor No more complete graph with k'?

Y

N k' = k' – 1 N

k' = 1? Y Complete algorithm

Fig. 2. Procedure for deciding spatial reuse factor.

Indeed, the chromatic number  (G ) of the graph G can be obtained by finding the maximal complete subgraph Gs containing the largest number of vertices from G. Thus, we can express the chromatic number as following,

 (G )  Vs , where Vs  {v|vi  Gs }.

(3)

Since  (G ) is computed, the available time slots for data transmission are divided into  (G ) classes. Thus, an arbitrary wireless link that corresponds to vertex vi can be  G 

allocated to the time slots Ti   i ,i  F ({v }) ti , where {vi } are i

defined as adjacent vertices of vi, and ti is the allocated time slot for color i. Thus, when the non-adjacent communication links use the same time slots allocated according to the interference graph, the spatial reuse can be maximized with only limited information. From the interference graph, we can compute the optimal frequency reuse factor of each wireless link. Here, the color means the frequency/time resource that is allocated to each antenna of the user. This means adjacent antennas interfering with each other cannot share the same resources. The spectral efficiency will be degraded if such adjacent links share the same resources. The chromatic number, which is defined as the number of vertices constituting a maximally complete subgraph, matches the maximum frequency reuse factor of each wireless communication link in the proposed algorithm. This is also explained by the fact that a complete graph of n vertices requires n colors as proved by Brook’s theorem [35], Ohyun Jo and Jungmin Yoon

257

2. Reuse factor 1/((1  ¼)/2) = 8/3

Scheduling procedure for dynamic resource allocation

8/3

1. Reuse factor 4

Table 1. Resource allocation to optimize the network throughput.

1 Pick an arbitrary vertex vi* from G. 2

Let n* be the spatial reuse factor of vi* determined from interference graph.

3

Allocate T /n* time slots to vi* from the available time slots. (Here, T is the total number of time slots.)

8/3 4

4

4 Update G  G  {vi* }. 5 Repeat Steps 1 through 4 until every vertex is examined. 4

4/3

4

3. Reuse factor 1/(1  ¼) = 4/3

Fig. 3. Optimal reuse factor decision based on coloring algorithm for interference graph of the sample network.

[5]. The efficient procedure for deciding the spatial reuse factor based on the interference graph is summarized in Fig. 2. The procedure begins by finding a maximally complete subgraph of G. The vertices that constitute the maximally complete subgraph must possess resources that are not allocated to any other vertex since the vertices cannot concurrently share the resource. Suppose  (G )  K , the frequency reuse factor of each vertex is computed as K in the first step. If there exists another complete subgraph with order K, the same operation is repeated. Subsequently, a complete subgraph with order K – 1 is searched. If any of the vertices constituting the complete subgraph with order K – 1 do not possess the spatial reuse factor, they are allocated an equally divided amount of resources. If some of the vertices already possess the computed spatial reuse factor from prior steps, the remaining resources are divided equally across the other vertices, excluding the pre-determined vertices. These operations are repeated until all vertices are examined and determined to possess the spatial reuse factor. Figure 3 is the interference graph for the sample network presented in Fig. 1. Each vertex is colored by applying the greedy coloring algorithm. Then, the optimal spatial reuse factor of each vertex is computed by applying the proposed algorithm. In the first iteration, the maximally complete subgraph with a degree of four is identified; thus, the spatial reuse factor of the four vertices constituting the maximally complete subgraph is determined as 4. In the second iteration, another complete subgraph of degree three is identified. The spatial reuse factor of one vertex in the complete subgraph with degree three is already determined in the prior step. Therefore, we compute the spatial reuse factor of the remaining vertices as 8/3. In the third iteration, the same procedure is repeated and another complete subgraph with degree two is identified, and 258

Ohyun Jo and Jungmin Yoon

the spatial reuse factor of the remaining vertex is computed. The major computational portion of the proposed algorithm consists of the process to calculate the chromatic number of the interference graph G. The graph coloring problem is one of Karp’s 21 problems that are NP-complete. However, the process to compute the chromatic number of a given interference graph can be completed in a finite time. This proposition is a practical application of the analysis to identify the upper bound of the computational complexity of deciding chromatic numbers in graph theory [25]–[32]. Many theorems proved in the field of graph theory and information theory are published in wireless communication and mathematics and computer science journals; these results can be applied to devise and analyze the feasibility of the proposed graph-based algorithm. This means the proposed algorithm can be implemented on a communication processor in real time. Proposition 2. The chromatic number of an interference graph can be computed in a finite amount of time [26], [27]. Proof. Suppose that an interference graph G has n vertices and m edges and assume that v and w are non-adjacent vertices in G. Denote by G + {v, w} = G + e the graph formed by adding edge e = {v, w} to G. Denote by G/e the graph in which v and w are “identified,” that is, v and w are replaced by a single vertex x adjacent to all neighbors of v and w. The number of pairs of non-adjacent vertices is specified n by  (G )     m. The proof is by induction on  . If  2

 (G )  0, then G is a complete graph and the algorithm terminates immediately. Now we note that  (G  e)  (G ) and  (G / e)  (G ): n

 (G  e)     m  1   (G )  1 2  

and  n  1   (m  c),  2 

 (G / e)  

where c is the number of neighbors of both v and w. Then, ETRI Journal, Volume 39, Number 2, April 2017 https://doi.org/10.4218/etrij.17.0116.0035

 n  1   (m  c)  2 

 (G / e)  

 n  1  mn2  2   n  1  mn2  2 

Antenna height 1 m

(4)

(n  1)(n  2)  mn22 2 n(n  1) 2(n  1)   mn 2 2 n     m  1   (G )  1. 2

If  (G )  0 then, G is not a complete graph, and therefore, there are non-adjacent vertices v and w which cannot be identified. By the induction hypothesis, the algorithm computes  (G  e) and  (G / e) correctly, and finally it computes  (G ) from these in one additional step. ■ The identification of the spatial reuse factor is followed by a scheduling process in Table 1. The proposed scheduling algorithm is implemented efficiently. According to the determined reuse factor of each vertex through the procedure previously described in Fig. 2, the resources are sequentially allocated to the wireless links to maximize the network throughput as following.

RSSI measurement (16 QAM, code rate ¾)

Fig. 4. Millimeter wave system testbed and test environments. 15 line (–6 dB) –49 dBm

32 m

20 line (11.5, 2) –68 dBm

–6 dB

16 m

(6, 2) –68 dBm

–6 dB 8m –6 dB

–6 dB

4m

2m –6 dB –19 dBm 1 m

(2.5, 2) –63 dBm

IV. Feasibility for mmWave Beamforming Systems. To demonstrate the effectiveness of the proposed algorithm within an industrial environment, we performed system level feasibility tests by using IEEE 802.11ad testbeds that operate in millimeter wave bands [36], [37]. The testbeds include a commercially available off-the-shelf 60 GHz transmitter module (SK63102). Figure 4 shows the testbed set and the test environments. The tests are performed in an indoor gym with a dimension of 45 m × 45 m. The testbed consists of a modem, ADC/DAC, RF/antenna modules, and controller PC/monitors. In the cases of the PHY operation and the 16 QAM modulation, the code-rate 3/4 error-correction code is used to realize maximum throughput of the IEEE 802.11ad single-carrier PHY [38]. The set of beams is composed of segmented 18 beam patterns, each of which have a beam width of approximately 22°. The PA output of the transmitter is fixed at 10 dBm, and the antenna gains of the transmitter and receiver are 16 dBi. The directions of the TX beam and RX beam are fixed to the bore sight. To measure the effective beam width and coverage, the location of the TX station is fixed. The received signal strength ETRI Journal, Volume 39, Number 2, April 2017 https://doi.org/10.4218/etrij.17.0116.0035

TX

Fig. 5. RSSI measurement test in millimeter wave beamforming technology.

indication (RSSI) is measured while the location of the RX station is changed. The test results illustrated in Fig. 5 show that the measured RSSI is less than −68 dBm, when the angle between the TX station and the RX station is close to 20°. In this case, multiple users can share the same resource because the interference is smaller than the clear channel assessment (CCA) carrier sensing threshold. We carried out similar experiments for various beams. Five beams in the horizontal direction, illustrated in Fig. 6, were selected from the fragmented beam set. Then, the maximum power spectral density was measured at each position. The results are illustrated in Fig. 6 and Table 2 and show the signal attenuation in the misaligned direction is significantly large so that spatial reuse can be an effective technique to increase the network throughput in directional beamforming networks using millimeter wave signals. In the case such that the angle between the TX station and the RX station exceeds 15°, the Ohyun Jo and Jungmin Yoon

259

HPBW w/antenna gain (dB), for SLS # = 0 90

0 dB

1.0

120

60 0.8 0.6

150

30 0.4 0.2

180

1

2

3

0

4

5

of interference for the generation of the interference graph. Thus, the interference graph for the IEEE 802.11ad system can be constructed to implement the proposed algorithm. In spite of the efficiency of the proposed graph model relative to its computational complexity, the pairwise model has a fundamental limitation because it does not account for aggregate interference. Thus, some centralized optimization approaches based on measurements of interference by neighbors may provide the optimal solution to this problem. However, in general, the solutions are identified from a difficult non-convex problem.

330

210

V. Performance Evaluation 240

300 270

–8.7705 dB

Fig. 6. Antenna gain patterns of the 18 segmented beams.

Table 2. Max power spectral density (dBm/Hz) with 3 MHz resolution. Horizontal angle () Beam index 1

–88.47 –88.27 –88.38 –75.86 –84.06 –85.00 –86.38

2

–86.48 –87.85 –79.09 –82.61 –86.28 –83.80 –86.41

3

–89.00 –83.53 –88.57 –88.32 –79.99 –79.78 –88.64

–45

–30

–15

0

15

30

45

4

–87.56 –78.46 –81.55 –84.08 –86.64 –88.00 –88.78

5

–83.46 –88.45 –84.72 –86.40 –88.17 –77.30 –80.84

measured maximum power density is similar to the noise power density level. To create a practical implementation of the proposed algorithm, we suggest a methodology that is applicable to the IEEE 802.11ad systems. To construct the interference graph with limited information, a simple message passing procedure can be deployed. According to the specification, multiple network allocation vector (NAV) timers are maintained to monitor multiple streams based on DBand request to send (RTS)/DBand and clear to send (CTS) frames, which form the extended version of conventional RTS/CTS frames. The DBand RTS and DBand CTS frames in the IEEE 802.11ad system contain the source address and the destination address. Using this new frame format, the stations can accurately identify the neighbor communication link that affects the station, while overcoming a hidden node problem at the same time. The network topology is obtained by reporting the multiple NAV timer durations and the address information of source/destination pairs, which result in interference with the link. One-bit quantization is sufficient to indicate the existence 260

Ohyun Jo and Jungmin Yoon

In this section, we analyze the computational complexity of the proposed algorithm. Let n be the number of wireless communication links, which are required to be scheduled for data transmission. Assume m time slots exist in a unit of time T. Then each link has an opportunity to join in each slot independently. In this situation, the complexity Cfull of the fullsearch based resource allocation is expressed as follows. Cfull  (2m ) n .

(5)

In the case of the proposed scheme, we first select k candidate vertices from the entire set of n vertices of the interference graph, and validate if the candidate vertices constitute a complete subgraph. The number of candidate vertices k can be changed from n to 1 until the maximally complete subgraph is identified. The number of possible combinations of k objects, taken from a set of n objects is expressed as follows. c( n, k ) 

n! . (n  k )!  k !

(6)

To check if the k vertices selected constitute a complete subgraph (k – 1)! comparisons are required for each case. If the k vertices are fully connected, they are regarded as a complete subgraph. Thus, the following equation approximates the complexity of the proposed algorithm Cprop based on the number of comparators. n n n ! (k  1) ! Cprop  c( n, k )  (k  1) !   k 1 k 1 ( n  k )! k ! (7) n n! n   2  1. k 1 ( n  k )!  k Figure 7 illustrates the computational complexity of each method. Our proposed method has complexity of order O(2n), and the full-search based resource allocation complexity is O((2n)n). Our proposed method outperforms the full search method due to its improved computational complexity. ETRI Journal, Volume 39, Number 2, April 2017 https://doi.org/10.4218/etrij.17.0116.0035

400 Full-search Proposed

Complexity comparison (Log2)

350 300 250 200 150 100 50 0

5

10

15 20 25 Number of links

30

35

40

Fig. 7. Complexity analysis of the proposed algorithm. 9 Beam width: 5 Beam width: 10 Beam width: 15 Beam width: 20 Beam width: 25 Beam width: 30

8

Spatial reuse efficiency

7 6

VI. Conclusions

5 4 3 2 1

5

10

15

20 25 Number of users

30

35

40

Fig. 8. Number of users vs. spatial reuse efficiency with respect to beam width.

We define the spatial reuse efficiency μ as a metric for performance evaluation.



to 30°. As shown in Fig. 8, the spatial reuse efficiency increases as the number of users increases, because the opportunity for spatial reuse increases as the number of users increases in beamforming systems. Additionally, a narrower beam width improves the system performance because the interference with other wireless communication links decreases as the beam width decreases. From these results, we are convinced that the directional beamforming technology will potentially generate large performance gains from the proposed algorithm in dense wireless communication environments by improving the possibility of spatial reuse. According to the measurements in Fig. 5, the proposed algorithm can be expected to experience a gain in network throughput of 2.3 X in practical IEEE 802.11ad systems. From the measurements, we are convinced that the proposed algorithm can be applied in many practical environments accompanied by significant improvement in network throughput.

N

1

  (v ) , i 1

(8)

In this study, we proposed a graph-based link scheduling scheme to maximize spectral efficiency while the amount of message traffic is significantly reduced. The proposed scheme is easy to implement and is very effective. We also presented a feasible methodology to apply the proposed algorithm to practical millimeter wave systems. By considering aggregate interference in the proposed graph model, the system performance can be further improved, even if the complexity of the scheduling increases. Aggregate interference is still a very important issue that should be studied further. Therefore, we plan a further study of this issue on millimeter wave systems. The performance improvement demonstrated in this study can be applied to new applications in the next generation of communication systems of the future.

i

where N is the total number of wireless communication links in the network, vi is an arbitrary communication link indexed by i,  (vi ) is the spatial reuse factor of vi that is computed by the proposed scheme. Thus, μ describes the average number of wireless communication links that successfully reuse a time slot while mitigating interference. We performed system level simulations using MATLAB to observe the performance of the proposed scheme based on the IEEE 802.11ad system. In the simulation environments, the users were assumed to be uniformly distributed within a 45 m × 45 m square space. Each user was randomly paired with another user to establish a wireless communication link. The beam width of the wireless communication link was assumed to be configurable from 5° ETRI Journal, Volume 39, Number 2, April 2017 https://doi.org/10.4218/etrij.17.0116.0035

Acknowledgement The author thanks all project members at Samsung Electronics and affiliated companies who contributed to the implementation and the tests documented herein.

References [1] G. Gronkvist and A. Hansson, “Comparison Between GraphBased and Interference-Based STDMA Scheduling,” Proc. ACM Int. Symp. Mobile Ad Hoc Netw. Comput., Long Beach, CA, USA, Oct. 4–5, 2001, pp. 255–258. [2] H. Luo, S. Lu, and V. Bharghavan, “A New Model for Packet

Ohyun Jo and Jungmin Yoon

261

[3]

[4]

[5]

[6]

[7]

[8]

[9]

[10]

[11]

[12]

[13]

[14] [15]

[16]

262

Scheduling in Multihop Wireless Networks,” Proc. Annu. Int. Conf. Mobile Comput. Netw., Boston, MA, USA, Aug. 6–11, 2000, pp. 76–86. A. Behzad and I. Rubin, “On the Performance of Graph-Based Scheduling Algorithms for Packet Radio Networks,” IEEE Global Telecommun. Conf., San Francisco, CA, USA, Dec. 1–5, 2003, pp. 3432–3436. A.D. Gore, A. Karandikar, and S. Jagabathula, “On High Spatial Reuse Link Scheduling in STDMA Wireless Ad Hoc Networks,” IEEE Global Telecommun. Conf., Washington, D.C., USA, Nov. 26–30, 2007, pp. 736–741. N. Alon, M. Krivelevich, and B. Sudakov, “Coloring Graphs with Sparse Neighborhoods,” J. Combinatorial Theory, Series B, vol. 77, no. 1, Sept. 1999, pp. 73–82. C.S. Sum et al., “Virtual Time-Slot Allocation Scheme for Throughput Enhancement in a Millimeter-Wave Multi-Gbps WPAN System,” IEEE J. Sel. Areas Commun., vol. 27, no. 8, Oct. 2009, pp. 1379–1389. H. Shokri-Ghadikolaei, L. Gkatzikis, and C. Fischione, “BeamSearching and Transmission Scheduling in Millimeter Wave Communications,” IEEE Int. Conf. Commun., London, UK, June 8–12, 2015, pp. 1292–1297. C.S. Sum and H. Harada, “Scalable Heuristic STDMA Scheduling Scheme for Practical Multi-Gbps Millimeter-Wave WPAN and WLAN Systems,” IEEE Trans. Wireless Commun., vol. 11, no. 7, July 2012, pp. 2658–2669. J. Wang, “Beam Codebook Based Beamforming Protocol for Multi-Gbps Millimeter-Wave WPAN Systems,” IEEE J. Sel. Areas Commun., vol. 27, no. 8, Oct. 2009, pp. 1390–1399. B.W. Ku, D.G. Han, and Y.S. Cho, “Efficient Beam-Training Technique for Millimeter-Wave Cellular Communications,” ETRI J., vol. 38, no. 1, Feb. 2016, pp. 81–89. Y. Niu et al., “A Survey of Millimeter Wave Communications (mmWave) for 5G: Opportunities and Challenges,” Wireless Netw., vol. 21, no. 8, Nov. 2015, pp. 2657–2676. T. Bai, A. Alkhateeb, and R.W. Heath, “Coverage and Capacity of Millimeter-Wave Cellular Networks,” IEEE Commun. Mag., vol. 52, no. 9, Sept. 2014, pp. 70–77. W. Roh et al., “Millimeter-Wave Beamforming as an Enabling Technology for 5G Cellular Communications: Theoretical Feasibility and Prototype Results,” IEEE Commun. Mag., vol. 52, no. 2, Feb. 2014, pp. 106–113. H.T. Friis, “A Note on a Simple Transmission Formula,” Proc. IRE, vol. 34, no. 5, May 1946, pp. 254–256. M. Bilal et al., “Time-Slotted Scheduling Schemes for Multi-Hop Concurrent Transmission in WPANs with Directional Antenna,” ETRI J., vol. 36, no. 3, June 2014, pp. 374–384. M. Kim, Y. Kim, and W. Lee, “Resource Allocation Scheme for Millimeter Wave–Based WPANs Using Directional Antennas,” ETRI J., vol. 36, no. 3, June 2014, pp. 385–395. Ohyun Jo and Jungmin Yoon

[17] J. Park, N. Lee, and R.W. Heath, “Cooperative Base Station Coloring for Pair-Wise Multi-Cell Coordination,” IEEE Trans. Commun., vol. 64, no. 1, Jan. 2016, pp. 402–415. [18] V. Aggarwal, A.S. Avestimehr, and A. Sabharwal, “On Achieving Local View Capacity Via Maximal Independent Graph Scheduling,” IEEE Trans. Inform. Theory, vol. 57, no. 5, May 2011, pp. 2711–2729. [19] G. Sharma, R.R. Mazumdar, and N.B. Shroff, “On the Complexity of Scheduling in Wireless Networks,” Proc. Annu. Int. Conf. Mobile Comput. Netw., Los Angeles, CA, USA, Sept. 23–29, 2006, pp. 227–238. [20] K. Zheng et al., “A Graph-Based Cooperative Scheduling Scheme for Vehicular Networks,” IEEE Trans. Veh. Technol., vol. 62, no. 4, May 2013, pp. 1450–1458. [21] D.M. Blough, G. Resta, and P. Santi, “Approximation Algorithms for Wireless Link Scheduling with SINR-Based Interference,” IEEE/ACM Trans. Netw., vol. 18, no. 6, Dec. 2010, pp. 1701– 1712. [22] G. Kim, Q. Li, and R. Negi, “A Graph-Based Algorithm for Scheduling with Sum-Interference in Wireless Networks,” IEEE Global Telecommun. Conf., Washington, D.C., USA, Nov. 26–30, 2007, pp. 5059–5063. [23] A.D. Gore, A. Karandikar, and S. Jagabathula, “On High Spatial Reuse Link Scheduling in STDMA Wireless Ad Hoc Networks,” IEEE Global Telecommun. Conf., Washington, D.C., USA, Nov. 26–30, 2007, pp. 736–741. [24] J. Qiao et al., “STDMA-Based Scheduling Algorithm for Concurrent Transmissions in Directional Millimeter Wave Networks,” IEEE Int. Conf. Commun., Ottawa, Canada, June 10– 15, 2012, pp. 5221–5225. [25] D.B. West, Introduction to Graph Theory, vol. 2, Upper Saddle River, NJ, USA: Prentice Hall, 2001. [26] B. Hajek and G. Sasaki, “Link Scheduling in Polynomial Time,” IEEE Trans. Inform. Theory, vol. 34, no. 5, Sept. 1988, pp. 910– 917. [27] J.R. Brown, “Chromatic Scheduling and the Chromatic Number Problem,” Manage. Sci., vol. 19, no. 4, Dec. 1972, pp. 456–463. [28] R. Beigel, “On the Complexity of Finding the Chromatic Number of a Recursive Graph I: the Bounded Case,” Ann. Pure Appl. Logic, vol. 45, no. 1, Nov. 1989, pp. 1–38. [29] D.G. Corneil and B. Graham, “An Algorithm for Determining the Chromatic Number of a Graph,” SIAM J. Comput., vol. 2, no. 4, 1973, pp. 311–318. [30] V. Voloshin, “On the Upper Chromatic Number of a Hypergraph,” Australasian J. Combinatorics, vol. 11, 1995, pp. 25–45. [31] U. Faigle et al., “On the Game Chromatic Number of Some Classes of Graphs,” Ars Combinatoria, vol. 35, Jan. 1993, pp. 143–150. [32] M. Jakovac and S. Klavžar, “The b-Chromatic Number of Cubic ETRI Journal, Volume 39, Number 2, April 2017 https://doi.org/10.4218/etrij.17.0116.0035

[33] [34] [35]

[36]

[37] [38]

Graphs,” Graphs Combinatorics, vol. 26, no. 1, Jan. 2010, pp. 107–118. K. Appel and W. Haken, “Every Planar Map is Four Colorable,” Bulletin Amer. Math. Soc., vol. 82, 1976, pp. 711–712. V.G. Vizing, “On an Estimate of the Chromatic Class of a pGraph,” Diskret. Analiz, vol. 3, no. 7, 1964, pp. 25–30. R.L. Brooks, “On Colouring the Nodes of a Network,” Proc. Cambridge Philosophical Soc., Math. Phys. Sci., vol. 37, 1941, pp. 194–197. O. Jo et al., “Holistic Design Consideration for Environmentally Adaptive 60 GHz Beamforming Technology,” IEEE Commun. Mag., vol. 52, no. 11, Nov. 2014, pp. 30–38. O. Jo et al., “60 GHz Wireless Communication for Future Wi-Fi,” ICT Express, vol. 1, June 2015, pp. 30–33. IEEE P802.11ad, Part 11: Wireless LAN Medium Access Control (MAC) and Physical Layer (PHY) Specifications Amendment 3: Enhancement for Very High Throughput in the 60 GHz Band, 2012.

Ohyun Jo received his BS, MS, and PhD degrees in electrical engineering from the Korean Advanced Institute of Science and Technology, Daejeon, Rep. of Korea, in 2005, 2007, and 2011, respectively. From 2011 to 2016, he was with Samsung Electronics, Suwon Rep. of Korea in charge of research and development for future wireless communication systems, applications, and services. Currently, he is a senior researcher at the ETRI, Daejeon, Rep. of Korea. He has authored and co-authored more than 20 papers, and holds more than 120 registered and filed patents. During his appointment at Samsung, He received several recognitions including the Gold Prize of the Samsung Annual Award, the Most Creative Researcher of the Year Award, the Best Mentoring Award, Major R & D Achievement Award. His research interests include millimeter wave communications, next generation WLAN/WPAN systems, 5G mobile communication systems, Internet of Things, future wireless solutions/applications/services, and embedded communications ASIC design and implementation. Jungmin Yoon received his BS degree in electrical and electronic engineering from Yonsei University, Seoul, Rep. of Korea, in 2009, and his MS degree in electrical engineering from Seoul National University, Rep. of Korea, in 2011. From Mar. 2011, he has been working with Samsung Electronics Co., Ltd., Suwon, Rep. of Korea. Since March 2016, he has been a doctoral student at Seoul National University, as a member of the Samsung Fellowship Program. He has contributed to the development of next generation wireless communication systems. His current fields of interest include research and development of mmWave wireless communication systems.

ETRI Journal, Volume 39, Number 2, April 2017 https://doi.org/10.4218/etrij.17.0116.0035

Ohyun Jo and Jungmin Yoon

263