Enhanced Time Slotted Scheduling Scheme for Multi-hop Concurrent Transmission in WPANs with Directional Antenna

Muhammad Bilal

Abstract— A high speed wireless personal area networks (WPANs) using millimeter wave (mmWave) with directional antenna is gaining increased interests. Due to some special characteristics of mmWave and use of multiple directional antennas make it possible to find non interfering transmissions in a localized region. To fully utilize high speed link, a multi-hop concurrent transmission scheme were proposed. In this paper, we have proposed an enhanced multi-hop concurrent transmission scheme and also made minor change in way of finding relaying nodes.

Index Terms—high speed WPANs, Millimeter wave communication, concurrent transmission, transmission scheduling, Throughput

I. INTRODUCTION

For ultra high speed short range communication networks, 60 GHz with 7~9GHz of bandwidth is most acceptable candidate among all mmWave communication bands. It has especial characteristics of oxygen absorption, which limits the transmission range of mmWave communication around 10 meters and the link speed is drastically decreased over distance. Also by using steering beams via directional antennas, increases the space reusability and increases the probability of concurrent transmission in a localized region. In mmWave communication the line-of-sight (LOS) communication is important to keep high data rate because in a high frequency band reflection is more dominant than diffraction at receivers.

In [1] a multi-hop concurrent transmission (MHCT) scheduling algorithm for WPAN is proposed. In order to keep a shorter distance and LOS, instead of direct communication, traverse the traffic on intermediate and special relay nodes. The piconet controller (PNC) selects a proper number of intermediate relay nodes, based on a novel hop selection metric reflecting the degree of distribution of the traffic load across the network. To further improve the network performance, a multi-hop concurrent transmission (MHCT) scheme used to allow non-interfering transmissions to transmit concurrently.

Through the analysis, we point out there is the margin of the algorithm to be improved in the classification of concurrent transmission grouping. The MHCT only considered avoiding the transmission interferences within a group does not consider avoiding the transmission interferences between groups. By properly adjusting the transmission times of groups, we can modify the classification method defined by the MHCT.

The remainder of the papers is organized as follows. In Section II, we describe the system design considerations. In section III the marginal improvement of MHCT is explained

and an enhanced version of MHCT with simulation result is presented. Section VI ends with conclusions.

II. SYSTEM DESIGN

An indoor WPAN usually consists of several wireless terminals (WTs) and a single PNC. Each of them is equipped with electrically steerable directional antenna, which means transmitters and receivers direct their beam toward each other for data transmission. We assume the PNC has topology information of each WT and itself.

A.mmWave Communication

According to Shannon theory and Friis transmission equation, the achievable data rate is calculated as the following equation,

Where W is the system bandwidth, N0 and I are the one side power spectral density of white Gaussian noise and interference respectively. Pt is transmission power, Gr and Gt

are the antenna gain of receiver and the antenna gain of transmitter respectively, λ is the wavelength and r is the transmission distance between the transmitter and the receiver, n is the path loss exponent whose value is usually between 2 and 6 for indoor environment [1]. According to (1), we can easily understand R is very sensitive to r. bigger r, smaller R.

B.Antenna Model

For the antenna gain, considered the following equation,

where Δφ = 2π/N is the antenna beamwidth when every node is equipped with an antenna with N beams, each of which spans an angle of 2π/N radians. Thus, if a transmitter and a receiver are directed within the antenna beamwidth each other (|φ|≤Δφ/2), the antenna gains of transmitters and receivers, Gt = Gr = 12dBi [3][4] and Gt = Gr = 0 outside.

C. Directional MAC Structure

The IEEE 802.15.3 superframe structure in Fig. 1 is divided into three parts; Beacon Period (BP), Random Access Period RAP) and Transmission Period. In BP PNC broadcast control and synchronization information. In RA period WTs sends transmission requests to PNC and in TP all transmissions are done according to the scheduling information distributed during BP.

Figure 1. IEEE 802.15.3 MAC.

D. Multi-hop Conversion Scheme

During random access period, the transmission requests of WT1 and WT4 are arrived at the PNC. Then, PNC first tries to construct the directional complete graph of the network based on the following hop selection metric.

Where, and are the average link length square and average traffic loads of each WT respectively. d2(i, j) is the link length of link (ij) and F(j) is the traffic load of WTj.

Once the cost of each link is calculated, PNC converts direct transmission requests into Multihop transmissions (MHT) by selecting the proper path between source and destination nodes, if ∑Ti < T, where Ti represent the time slots required by ith hop in MHT and T is time slot requirement for direct transmission. The non-colliding multihop transmissions then scheduled for concurrent transmissions

III. ENHANCED TIME SLOT ALLOCATION

In this section, we analyze the time slot allocation process of the original MHCT. By observing the process, we point out a margin to be improved and propose an extension of time slot allocation for multi-hop concurrent transmission and we also present our enhanced version of algorithm (EMHCT).

A.Time Slot Allocation Process

If we assume m transmission requests are arrived at the PNC during the random access period. Let Ri denote each transmission request with its arrival order. Let {hRi

k , k = 1, 2, …, m} denote the ordered sequence representing the multi-hop transmission for Ri. m will be a number between 1 and n(n-1)/2 when n is the number of WTs. For example, the ordered sequence for R1 is {hR1

1 , hR12 , …, hR1

m1}. Initially, we receive multiple transmission requests R1 to Rk. Each Ri is then converted to multiple hop transmissions. The set of TREQs{hR1

1 , hR21 , …, hRk

1} represent the first hop transmission requests of multi-hop transmissions. All transmission requests in TREQ are sorted in decreasing order and checked for concurrency and form a subset Group (G1) of TREQ which contains all first hop transmission requests that can be scheduled concurrently. In next step the set of TREQs{hR1

2 , hR22 , …, hRk

2} representing the second hops transmission requests of multi-hop transmissions, along with remaining transmission requests from first hops, will checked for concurrency and form next group of concurrent multi-hop transmissions, G2. This is a recursive process which continues until one of the two conditions satisfied.

1- ∑Ti ≥ SF (Super Frame) or

2- All requests are scheduled. Where Ti denotes the maximum number of time slots

consumed by i-th group. When condition “1” satisfied, the next hop to the latest successful scheduled hop of Ri is consider as initial hop and recalculate multi-hop transmissions to final node.

After finishing the scheduling of the last transmission request, we may obtain the transmission scheduling map such as Fig.5.

Figure 2. An example of time slot allocation map of transmission period for concurrent transmission

The allocation map made by MHCT Algorithm indicates which concurrent hop transmissions belong to each group and how many time slots each group consumes. Then the total time slots allocated will be

Therefore, the bandwidth efficiency is determined by (4) and higher bandwidth efficiency will be achieved with smaller (4). If we carefully observe the map, we can find a margin to improve the efficiency with a little change to the allocation map.

B.Span Overlapping of Each Group

MHCT Algorithm does not consider inter-collisions among groups but only considers intra-collisions within a group. In other words, the hop transmissions in a group are guaranteed to have no collision but hop transmissions between groups are not checked if they are interfered or not. For example, hop transmission of R2 and R4 in G3 are interfering R3 and R4 in G2, but both have no interference with R1 and R2 of G2. In this case R2 and R4 in G3 can be transferred to G2 and can be scheduled at non colliding position to R3 and R4 of G2. So, the hop transmissions of R2

and R4 in G3 do not need to wait for the end of the hop transmission of R3 or R4 in G2. Thus, if we move the hop transmission of R2 and R4 in G3 at non colliding position such as Fig. 6 total time slots of all the groups will be reduce, which leads to bandwidth efficiency increase. We can perform more generally the span overlapping of each group by following the next descriptions.

STEP 1: Execute MHCT Algorithm and obtain the time slot allocation map

STEP 2: Sort the hop transmissions in every group with the number of allocated time slots in descending order

STEP 3: We start span overlapping process from G2

against G1. After finishing the span overlapping between G2 and G1, we apply the same procedure to G3 against G2 and so on. In case of G2’s span

overlapping, we start it from the first hop transmi-ssion of G2, Check if this hop transmission or the span overlapping candidate causes a collision with the hop transmissions in G1 one by one until meeting the hop transmission having a collision with each other or the hop transmission belonging to the same transmission request.

STEP 4: if the lookup of STEP 3 finds the specific hop transmission satisfying the conditions, move the allocated time slots of the span overlapping candi-date back to back at the end of the hop transmission before the specific hop transmission. Then, the next hop transmission of G2 performs the same procedure of STEP 3. After finishing span overlapping of all hop transmissions in G2, the hop transmissions of G3

starts the span overlapping procedure as described in STEP 3 and STEP 4. The span overlapping procedure will continue until finishing the last group’s span overlapping.

Figure 3. An example of the span overlapping considering inter-collisioin between groups

C. Performance Comparision

It is clear the number of the total time slots allocated with span overlapping is less than or at least equal to that without span overlapping. Therefore, the multi-hop concurrent transmission with span overlapping is better than the original multi-hop concurrent transmission without span overlapping.

Fig. 4 shows the result of our simulation of both algorithms. Enhanced version of algorithm performs well for high number of traffic flows.

IV. CONCLUSION AND FUTURE WORK

In this paper, we analyzed the process of multi-hop concurrent transmission proposed in [1] and pointed out there is the margin to be improved if we consider the collision relations between hop transmissions in the adjacent groups. Thus, for better bandwidth efficiency, we have proposed a span overlapping EMHCT scheme to lessen the total number of allocated time slots in transmission period for a given transmission requests. Also we implicitly show span overlapping is beneficial. Currently we are making the simulation code to explicitly investigate bandwidth efficiency difference.

BIOGRAPHY

Muhammad Bilal has received his BS degree in Computer Systems Engineering from University of Engineering and

Technology (UET) Peshawar Pakistan and MS in Computer Engineering from Chosun University Korea. Currently he is Ph.D student at University of Science and Technology (UST) Korea in Electronics and Telecommunication Research Institute

(ETRI) campus. His research interests include Design and Analysis of Network Protocols, Network Architecture and Future Internet.

REFERENCES

[1] J. Qiao, L.X. Cai, X. Shen, “Multi-Hop Concurrent Transmission in Millimeter Wave WPANs with Directional Antenna,” in Proc. IEEE ICC’10, May 2010, pp.1-5.

[2] C. Kärnfelt, , P. Hallbjörner, H. Zirath, and A. Alping, “High Gain Active Microstrip Antenna for 60-GHz WLAN/WPAN Applications,” in IEEE Transactions on microwave theory and techniques, June 2006, VOL. 54, NO.6

[3] R. Mudumbai, S. Singh, U. Madhow, “Medium Access Control for 60 GHz Outdoor Mesh Networks with Highly Directional Links,” in Proc.IEEE INFOCOM’09, April 2009, pp. 2871-2875.

[4] G. Karnfelt, P. Hallbjorner, H. Zirath and A. Alping,” High Gain Active Microstrip Antenna for 60-GHz WLAN/WPAN Applications”, IEEE Transactions on Microwave Theory and Technology, VOL. 54, NO. 6, Jan. 2006.

Figure 4. Network throughput MHCT Vs EMHCT

APPENDIX

Algorithm EMHCT______________________________________________________________________________________ BEGIN:

1: PNC receives a request for n (I, J) time lots

2: for all non-empty group (Gb! = Null) do 3: for all non-empty group (Gi! = Null) ,{i=1,2….b-1} do

4: if ’s beams conflict with few of existing hops in Gi or have shared nodes with other hops in Gi then

5: Gc = Identify beam conflicting and shared nodes, where Gc ⊆ Gi

6: n(c) = Maximum n in Gc

7: for all in Gc do

8: if n (I, J) ≤ n(i) - n(c)

9: Update Gi = Gi ⋃ i;{ }, position at n(c) and Go to END;

10: end if11: end for12: end if13: end for

14: if ’s beam does not conflict with those of all existing hops in Gb then

15: if does not have shared nodes with other hops in Gb then

16: if requires extra slots, n (I, J) − n (b) > 0 then

18: if Available slots N ≥ n (I, J) − n(b) then

19: Schedule in Group Gb;

20: Update Gb = GbU{ };

21: Update the available slots N = N − [n (I, J) −n (b)];22: Update n(b) = n (I, J);

23: Update the allocated slots for ;

24: Sort all hops in the decreasing order of allocated slots.25: go to END;26: else27: go to line 38;28: end if29: else

30: Schedule in Group Gb;

31: Update Gb = Gb ⋃b;{ };

32: Update the allocated slots for ;

33: Sort all hops in the decreasing order of allocated slots.34: Go to END;35: end if36: end if37: end if 38: Next Group;39: end for