Professional Documents
Culture Documents
Li Zhitang1, Fan Jun1, Nie Wei2, Wang Li1,Chen Yuan1 School of Computer Science and Technology, Huazhong University of Science and Technology, Wuhan, Hubei, P.R.China 2 School of Automation Engineering, University of Electronic Science and Technology, Chengdu, Sichuan, P.R.China leeying@mail.hust.edu.cn; odbc@tom.com; nwei888@sina.com; wtwl@hust.edu.cn; ychen@smail.hust.edu.cn Abstract
Issues such as the hidden and exposed terminal problems and limited traffic carrying capability still remain unsolved in wireless ad hoc networks. In this paper, we present a novel MAC protocol employing power control. Our design mainly focuses on alleviating the problems of the hidden and exposed terminals and enhancing spatial reuse. We introduce an effective power control scheme to facilitate the power negotiation between the source node and the receiver node. Our protocol can collaborate well with high layer protocols such as AODV in the multi-hop scenarios. We implement our protocol in OPNET environment and study the parameters to optimize its performance. Our preliminary simulation results indicate that, with properly chosen parameters, our protocol works well and achieves considerably high performance in both sparse and dense ad hoc networks. In this paper, we introduce a novel MAC protocol that employs power control for wireless ad hoc networks. The core contribution of our study is as followings. On one hand, we propose an effective power control scheme to facilitate the negotiation of transmission power between the source node and receiver node. On the other hand, we introduce a new scheme to alleviate the problems of the hidden and exposed terminals. As illustrated by the simulation results, our protocol can greatly improve the traffic carrying capacity and is suitable for both sparse and dense ad hoc networks. The rest of the paper is organized as follows. In section 2, we summarize some related works in the area of MAC protocol using power control for ad hoc networks. In section 3, we describe our protocol in detail in all aspects. In section 4, we present the preliminary simulation results and compare with IEEE 802.11 MAC protocol. Finally in section 5, we conclude the paper and highlight some open problems and future research directions.
1. Introduction
Wireless ad hoc networks are formed by autonomous system of mobile hosts connected by wireless links. Without fixed infrastructure, communication in ad hoc networks takes place through the intermediate nodes, acting as hosts and routers in a store-and-forward manner [1]. The goal of medium access control is to enable efficient and fair sharing of the common wireless channel among nodes in ad hoc networks. However, the hidden and exposed terminals are introduced by the distributed nature of wireless communication, making the design of MAC protocol a rather challenging issue. Besides addressing the hidden and exposed terminal problems, medium access control protocol must exercise power control to improve spatial reuse and save energy.
2. Related works
In recent years, quite a few power control MAC protocols have been proposed, which are almost the adaptation of CSMA/CA (especially IEEE 802.11 [2]) for use over power control. The main idea of the proposed protocols is to use the maximum transmission power for RTS-CTS and the minimum necessary transmission power for DATA-ACK [3] [4]. Although multiple power levels can improve spatial reuse, it brings asymmetric links and more collisions. Even worse, nodes with higher power may overwhelm nodes with low power in accessing channel resources and consequently causes unfairness. In PCM [5], the source node periodically transmits the DATA packet at the maximum power level, so that
2007 International Conference on Multimedia and Ubiquitous Engineering(MUE'07) 0-7695-2777-9/07 $20.00 2007
nodes in the carrier sensing range can sense it. However, the drawbacks are the interference caused by such DATA packets and the difficulty in implementing frequent changes in the transmission power levels. PCMA [6] uses two channels, one for data and control packets and the other for busy tones. By sensing the strength of the busy tone signal, a sender node determines a maximum power level that would not interfere with ongoing communication. Although this scheme shows encouraging performance, the additional receiver hardware as well as radio channel causes relative complexity in order to implement it. As analyzed by [6] [7], CSMA/CA is inherently incompatible with power control. In order to guarantee collision-free floor acquisition, RTS-CTS must be transmitted with the largest power. Adversely, sending RTS-CTS at the maximum power yields little benefit over no power control. We think that this dilemma can only be solved by separating the RTS-CTS and DATAACK in different channels, or using a synchronous scheme so as to eliminate the interference between RTS-CTS and DATA-ACK packets, as described in this paper.
which are composed of mini-slots for the assertion signal of CRS (Collision Resolution Signaling), RTS slot, CTS slot and SBT (Sender Busy Tone) slot. We divide the data period into DATA cycle and ACK cycle. In order to avoid the overlap of the control and the data packets, we restrict the transmission and reception of the assertion signal, RTS, CTS, DATA and ACK packets in the responding time slots.
Figure 1. Frame and timeslot organization In order to reduce the overhead, we need to make time interval of time slots as short as possible. However, we must include the synchronization accuracy, propagation delay, detection time, and receive-to-transmit transition time to prevent ambiguity.
2007 International Conference on Multimedia and Ubiquitous Engineering(MUE'07) 0-7695-2777-9/07 $20.00 2007
IEEE 802.11 RTS, except that the two-byte field containing duration is replaced with PMax. Receiving the RTS, the receiver node R gets PMax and computes the PMin (the minimal transmission power should be used by the source node so that the receiver node can decode successfully, as described in the power control sub-section). If PMin < Pm ax , the transmission solicitation from node S is acceptable. For this transmission will not destroy the registered senderreceiver pairs around the source node. If PMin Pm ax , or the node R is an exposed terminal, or node R is busy, the transmission solicitation from the source node is rejected. We use two types of CTS: the CTS-A (CTSAcceptance) and the CTS-R (CTS-Rejection). The CTS-A informs the source node and neighboring nodes that a new sender-receiver pair has been registered. The format of the CTS-A is similar to the IEEE 802.11 CTS, except the two-byte field containing the duration is replaced with PMin. The CTS-R informs only the source node that the transmission solicitation has been rejected. CTS will be transmitted with Pone-hop in the CTS slot. SBT is a burst similar to the assertion signal, and is transmitted by the source node with power PMin in the SBT cycle. If SBT is received, a node (except the receiver) marks it as an exposed terminal that should not receiver in current transmission slot. 3.2.2. The repeated CRS/RTS/CTS/SBT handshake Since the distance between CRS survivors is twohop radius, the sender-receiver pairs produced in one contention cycle, or rather one CRS/RTS/CTS/SBT handshake, are quite sparse. In order to improve spatial reuse, we have multiple contention cycles in the contention period. In the next contention cycle, except the nodes in the busy status, other nodes can contend for the access of medium in current transmission slot. 3.2.3. DATA/ACK exchange As a result of multiple contention cycles, multiple sender-receiver pairs that do not interfere with each other can be registered. In the last step, the DATAACK exchange of all these sender-receiver pairs will take place simultaneously. The design consideration for SPC-MH is an optimization problem. On one hand, each contention cycle consumes the wireless channel for some amount of time. Therefore it is desirable to minimize the number of contention cycles. On the other hand, the overhead of contention cycles is amortized over the multiple data packets transmitted in data period. We need to balance the length of contention period and the data period.
There are mainly two points in our power control scheme. On one hand, different transmit power level is assigned for the assertion signal, RTS, CTS, SBT, DATA and ACK. On the other hand, a source node decides PMax, while the receiver node decides PMin. From [10], Formula (1) holds. The source node transmits a signal with power Pt, the receiver node receives the signal with power Pr, is the carrier wavelength, d is the distance between transmitter and receiver, n is the path loss coefficient, and gt/gr is the antenna gain, which is 0dB since omni-directional antenna is used in our protocol. n t r (1) Pr = Pt gg 4 d Firstly, we make the assertion signal transmitted with power 8Pone-hop (we assume the path loss coefficient equals to 3), and RTS-CTS transmitted with power Pone-hop. As result, the range of the assertion signaling is two-hop radius. Consequently, the distance between the CRS survivors is two-hop radius, too. Secondly, receiving the RTS, the receiver node computes PMin as Formula (2). The Prcv-rts is the signal strength of the RTS measured by the receiver node, and PD is the power threshold that can satisfy the correct decoding. PMin will be included in the CTS-A. The SBT, DATA and ACK will be transmitted with power PMin.
PM in = Pone h op PD Prcv rts
(2)
Thirdly, we limit the transmit power of the source node so as to protect the registered sender-receiver pairs. If CTS-A is overheard, PMax is computes as Formula (3). The Prcv-cts is the signal strength of the CTS-A measured. If multiple CTS-A are overheard, the PMax should be updated with the smallest value. If no CTS-A is overheard, the default PMax is Pone-hop.
PMa x = Pone hop PD Prcv cts
(3)
2007 International Conference on Multimedia and Ubiquitous Engineering(MUE'07) 0-7695-2777-9/07 $20.00 2007
and RTS-CTS handshake can succeed with very high possibility. As illustrated in Figure 2, we assume that the source node S survives the CRS, while other nodes such as A, B, C and R fail. In the following RTS slot, except S, other nodes in the coverage of signaling area ASIG keep silent. Thus the collision of the RTS at receiver node R can be avoided, even in the case that D is a CRS survivor. Similarly, only node R can transmit in the CTS slot in the coverage of signaling area ASIG. It is evident that modified CRS can eliminate the collision of RTS-CTS caused by hidden terminals, and RTS-CTS exchange can succeed with very high possibility.
As illustrated in Figure 4, we assume the source node S and the receiver node R register a senderreceiver pair first. The dotted circles illustrate the onehop range of RTS-CTS, while the shadowed circle illustrates the range of DATA-ACK exchange. Node N3 receives the SBT from node R and recognizes that it is an exposed terminal. As a result, node N3 (the exposed terminal) is not allowed to receive in this transmission slot. However, node N3 can transmit to node N4 in condition that other registered senderreceiver pairs in the vicinity of node N3 will not be interfered.
Figure 4. Alleviating the exposed terminal problem Figure 2. RTS/CTS handshake of CRS survivor 3.4.2. Alleviating the hidden terminal problem in DATA/ACK As illustrated in Figure 3, we assume that the source node S and the receiver node R register a senderreceiver pair first. The dotted circles illustrate one-hop range of RTS/CTS/SBT, while the shadowed circle illustrates range of DATA-ACK exchange. Node N1 receives CTS-A from node R, recognizing it is a hidden terminal and computing PMax. In this transmission slot, node N1s transmit power should not go beyond PMax. By comparing PMax and PMin required by node N2, node N1 (the sender node) decides whether to cancel the transmission solicitation or not. Thus node N1 (the hidden terminal) will not impact the sender-receiver pair of node S and R.
Figure 3. Alleviating the hidden terminal problem 3.4.3. Alleviating the exposed terminal problem in DATA/ACK
2007 International Conference on Multimedia and Ubiquitous Engineering(MUE'07) 0-7695-2777-9/07 $20.00 2007
the page-limit of this paper, we can only present the results in 4 figures. For the convenience of reading, we summarize the results of simulations in Table 2. From the simulation results, we find some trends as follows. Table 2. Network layer performance
Node Density (nodes/km 2 ) 16 25 36 49 Throughput (Kbits/s) IEEE 802.11 340 280 250 240 SPC-MH 480 520 560 670 Delay (s) IEEE 802.11 0.4 1.7 2.1 3.4 SPC-MH 0.4 0.6 0.7 0.8
We implemented our first version of SPC-MH model in OPNET Modeler environment. Our main objective is to investigate the impact of manipulating three parameters, including node density, traffic generation rate and the number of contention cycles, on the network layer performance in term of throughput and delay in multi-hop scenarios (AODV is used in network layer). In order to check the performance of our protocol, we compare the performance of an ad hoc network employing SPC-MH and IEEE 802.11b (Direct Sequence) DCF protocols in the MAC layer in a fair base.
Firstly, we find from Figure 5~8 that, as traffic generation rate increases, the throughput increases while the delay degrades using SPC-MH and IEEE 802.11 in MAC layer. This is due to the queue delay brought by jam of the wireless channel. Secondly, we find that the optimal number of contention cycles for best performance is different at different node density. CC3~4 is suitable for low node density, CC5~6 is suitable for moderate node density, and CC7~8 is suitable for high node density. Thirdly, node density impacts the network performance significantly. As node density and traffic generation rate increases, with properly chosen parameters, the throughput using SPC-MH is 480 Kbits/s, 520 Kbits/s, 560 Kbits/s and 670 Kbits/s, while the throughput using IEEE 802.11 is 340 Kbits/s, 280 Kbits/s, 250 Kbits/s and 240 Kbits/s. The throughput using SPC-MH is about 141%, 185%, 224% and 279% of that using IEEE 802.11 at node density of 16, 25, 36 and 49 nodes/km2. It is obvious that SPM-MH can achieve much higher performance in term of delay at moderate and high node density. As can be seen from the above simulation results, SPC-MH significantly improves the network layer performance in dense ad hoc network. We believe that the enhancement in the performance is attributed to two points. On the one hand, SPC-MH alleviates the hidden terminals, which is the main cause of collisions. On the other hand, there is much more parallelism in the packets transmissions with the alleviation of the exposed terminal problem and the employment of power control.
2007 International Conference on Multimedia and Ubiquitous Engineering(MUE'07) 0-7695-2777-9/07 $20.00 2007
control, study of the performance of SPC-MH implementing more routing protocols and extension to SPC-MH by employing directional antennas.
References
[1] IETF MANET Working Group,
5. Conclusion
In this paper, we introduce a MAC protocol that employs power control for wireless ad hoc networks. Our protocol can not only alleviate the hidden and exposed terminal problems, but also greatly improve the spatial reuse. We also propose an effective power control scheme to facilitate transmission power negotiation between the source node and the receiver node. With this feature, although complex, our protocol provides efficient access and achieves high traffic carrying capacity. We evaluate the performance of multi-hop ad hoc network employing SPC-MH and IEEE 802.11 via extensive simulations. Our preliminary performance simulations show that SPC-MH is suitable for both sparse and dense ad hoc network. The results we obtained are promising, and the extent of this performance improvement is even larger in dense network. Future works would include comparison of SPC-MH with the proposed MAC protocol with power
2007 International Conference on Multimedia and Ubiquitous Engineering(MUE'07) 0-7695-2777-9/07 $20.00 2007