Receiver-Initiated vs. Short-Preamble Burst MAC Approaches for Multi-channel Wireless Sensor Networks

Transcription

1 Receiver-Iniiaed vs. Shor-Preamble Burs MAC Approaches for Muli-channel Wireless Sensor Neworks Crisina Cano, Boris Bellala, and Miquel Oliver Universia Pompeu Fabra, C/ Tànger , Barcelona, Spain Absrac. Wireless Sensor Neworks (WSNs) are neworks formed by highly consrained devices ha communicae measured environmenal daa using low-power wireless ransmissions. The increase of specrum uilizaion in non-licensed bands along wih he reduced power used by hese nodes is expeced o cause high inerference problems in WSNs. Therefore, he design of new dynamic specrum access echniques specifically ailored o hese neworks plays an imporan role for heir fuure developmen. This work aims o provide insighs ino he performance of wo well-known low energy-consuming Medium Access Conrol (MAC) echniques for WSNs (he receiver-iniiaed and he shor-preamble burs), adaped o work wih muliple channels. The resuls obained show he bes working condiions for each approach. Keywords: WSNs, Muli-Channel, MAC, SA-MAC, Receiver-Iniiaed. 1 Inroducion Wireless Sensor Neworks (WSNs) are formed by small and low-capabiliy devices ha are able o sense environmenal merics and o communicae hem wirelessly o a cenral uni, known as a sink. The deploymen of dense WSNs in large, remoe and difficul-o-access areas requires keeping he size and cos of sensor nodes as low as possible. This implies ha he energy, compuaional and memory resources of he sensor nodes are usually limied. Among he differen consrains, he limiaion of he energy resources is he mos imporan one as i direcly affecs he nework lifeime. Therefore, he design of he Medium Access Conrol (MAC) layer is of crucial imporance because i conrols he mos energy consuming componen of a sensor node: he ransceiver. By puing he ransceiver in duy cycle mode, i.e., sleeping and periodically waking up, he energy consumpion can be reduced. However, a mechanism is needed o coordinae a receiver and a ransmier so as o be awake a he same insan, also known as rendezvous. Among he differen echniques o ensure a rendezvous, asynchronous MAC proocols and paricularly preamble sampling echniques [2] provide exremely low energy consumpion a low loads (he common siuaion in WSNs), have a low complexiy and a lack of synchronizaion requiremens. R. Szabó and A. Vidács (Eds.): EUNICE 2012, LNCS 7479, pp , c IFIP Inernaional Federaion for Informaion Processing 2012

2 24 C. Cano, B. Bellala, and M. Oliver Sensor nodes are normally configured o ransmi a very low power [4]. This allows hem o consume smaller amouns of energy while ransmiing bu, on he oher hand, his configuraion also makes hem suffer from exernal inerferences in a noable manner. I is expeced ha he curren level of deploymen of differen wireless neworks will make he Indusrial, Scienific and Medical (ISM) band crowded in he near fuure [10]. Therefore, i will be necessary for a sensor device o be able o selec he bes porion of he specrum o communicae wih is neighbours, hus increasing he probabiliy of a correc recepion, reducing he delay and mainaining a low energy consumpion per successfully sen packe. When here are muliple channels available, he rendezvous problem becomes more complicaed. In ha case, apar from coordinaing he nodes o be awake a he same ime, hey also need o agree on he channel o use. In his work, he adapaions o muli-channel operaion of wo well-known MAC approaches for rendezvous are evaluaed and he bes working condiions for each approach are obained. These firs resuls can be useful for he furher sudy and design of hese proocols when muliple channels are available and sensor nodes selec one in which o ransmi or lisen for daa. This aricle is organized as follows. In Secion 2 he exising asynchronous proocols for muli-channel operaion are described. Then, in Secion 3 he mulichannel echniques evaluaed in his aricle are presened. Afer ha, an analyical derivaion of hese mechanisms is explained in Secion 4. The descripion of he resuls obained is presened in Secion 5. Finally, some conclusions and fuure research direcions are oulined. 2 Relaed Work In common energy-consrained WSNs, he sensor nodes are sleeping and waking up according o heir duy cycle. Therefore, a mechanism is needed o coordinae a receiver and a ransmier o communicae, since boh have o be awake a he same insan. Techniques o achieve his rendezvous can be divided ino hree differen caegories: TDMA-like approaches, proocols wih common acive periods and asynchronous approaches [3]. Among hese, asynchronous MAC proocols and paricularly preamble sampling echniques consume exremely low energy when he nework load is low, have a low complexiy and no synchronizaion requiremens. Using preamble sampling, sensor nodes sample he channel periodically o deec aciviy and hey remain awake if aciviy is deeced. Thus, if a sensor node has somehing o ransmi, i sends a long preamble before he daa in order o wake up he receiver [7]. There are several exensions of his echnique, one of he bes known is he division of he long preamble ino a shor-preamble burs [2]. By dividing he long preamble ino shor packes, some useful informaion can be included, like, for insance, he ime a which he daa ransmission will sar, allowing he receiver o go o sleep. Receiver-iniiaed approaches [8] follow a similar idea o he one used in preamble sampling. However, in his case, when a sensor node wans o ransmi somehing, i wakes up and lisens

3 Receiver-Iniiaed vs. Shor-Preamble Burs MAC for Muli-channel WSNs 25 o he channel. Sensor nodes wake up periodically and send a beacon message indicaing hey are awake. Afer receiving he beacon message, he sender can sar he ransmission. When here are muliple channels o use, he rendezvous problem becomes more challenging. In ha case, apar from coordinaing he nodes o be awake a he same ime, hey also need o agree on he channel o use. In his work, he aenion has been placed on asynchronous approaches since, as previously saed, hey provide feaures ha are especially beneficial o WSNs. OMA [5] and ARM [6] proocols are wo asynchronous MAC proocols for muli-channel operaion. While OMA is based on he ransmission of a burs of shor preambles, ARM is based on he receiver-iniiaed approach. In hese proocols, he rendezvous is made in a common and fixed conrol channel, hen he daa ransmission is done in he channel boh receiver and ransmier agree o use. In conras, SA-MAC [1], ha is based on he ransmission of a burs of shor preambles, does no use a common conrol channel o perform he rendezvous. When a node wakes up i samples all channels rying o find a shor preamble of he receiver. The shor-preamble burs should ake ino accoun, apar from he sleep ime of he sensor nodes, he ime o sample all he available channels. No using a common conrol channel makes he proocol more robus agains inerference problems a he cos of a higher overhead. Similarly, he EM-MAC [9] proocol ensures he rendezvous beween receiver and ransmier in an asynchronous manner and wihou he use of a common conrol channel. The difference is ha his proocol is based on he receiveriniiaed echnique adaped o muli-channel operaion and combined wih a predicion of he channel and wake-up ime of he receiver based on pseudorandom numbers. This mechanism allows he ransmier o wake up righ before he receiver beacon in he seleced channel. As a drawback, each node mus sore informaion of each neighbour and exra overhead has o be included in messages o perform he predicion. In his work, he receiver-iniiaed echnique presened in [8] is adaped o muli-channel operaion by making he ransmier sample all channels o find a burs of beacons of he receiver. On he oher side, he ransmier-iniiaed echnique defined in he SA-MAC proocol is adoped. The comparison of boh echniques gives insighs ino he bes working condiions for boh approaches when muliple channels are available. 3 Muli-channel Receiver and Transmier-Iniiaed MACs In SA-MAC [1], he rendezvous based on he ransmission of a shor-preamble burs along wih he periodic channel sampling of he receiver (similar o he single-channel echnique defined in [2]) is adaped o muli-channel operaion. An example of he basic funcionaliy of his proocol wih 4 channels is shown in Fig. 1. I can be observed ha, when a node wans o ransmi a message i

4 26 C. Cano, B. Bellala, and M. Oliver sars ransmiing a burs of shor preambles in he seleced channel and afer ha, i sends he daa. The receiver, on he conrary, wakes up periodically and samples all channels. If he receiver deecs aciviy in one channel, i remains awake o receive a complee shor preamble, which includes he ime a which he daa ransmission will sar. This allows he receiver o go o sleep and wake up righ before he ransmission of he daa message. Periodic wake up of he rx Channel 1 Channel 2 Channel 3 New packe o x The ransmier finds he receiver DATA ACK Channel 4 Time o sample all channels Sleep Time Fig. 1. Example of he shor-preamble burs approach funcionaliy wih 4 channels Observe ha he rendezvous would also be possible by making he sender swich among channels and sending a shor preamble on each. The receiver could hen say lisening in one channel unil he shor preamble is received. The ransmier can eiher wai in each channel for an early-ack [2] or announce in each preamble he channel in which he message will be sen. Boh approaches require more energy wase and delay if compared o SA-MAC. The firs case requires he ransmier o lisen on each channel and he second one requires exra channel swiches. Based on he fac ha modern radios (like he CC2420 ransceiver [4]) consume less energy in ransmiing han in receiving mode, he adapaion of receiver-iniiaed approaches o muli-channel operaion seems o be a good opion, especially when he nework load is low (as i ypically occurs in usual WSN applicaions). The funcionaliy of he receiver-iniiaed proocol sudied in his work is shown in Fig. 2 for he case in which again 4 channels are available. Each node wakes up periodically and sends a burs (equal o he number of channels) of beacons o noify poenial ransmiers i is awake and ready o receive daa. As previously discussed, his behaviour is expeced o consume less energy han waking up o sample all channels in he cases in which he energy o receive is higher han he energy required o ransmi. Then, when a node wans o ransmi a packe, i sequenially and periodically scans all channels o find a beacon of he inended receiver. Afer he recepion of a complee beacon, he ransmier goes o sleep and wakes up afer he beacon burs o send he daa message (assuming he nework load o be low, no back-off has been considered).

5 Receiver-Iniiaed vs. Shor-Preamble Burs MAC for Muli-channel WSNs 27 Observe ha he ransmier should sample he channels during he sleep ime plus wice he beacon burs duraion o ensure he rendezvous. New packe o x Channel 1 Channel 2 Channel 3 Channel 4 Time o sample all channels Sleep Time The ransmier finds he receiver ACK DATA Periodic wake up of he rx Fig. 2. Example of he receiver-iniiaed approach funcionaliy wih 4 channels Observe also ha he rendezvous could alernaively be done in his case by having he receiver swiching among channels wih he ransmier waiing for he recepion of a beacon in a given channel. In his design, he receiver should eiher say awake in every channel afer he ransmission of he beacon waiing for daa or announce he channel in which i will be lisening for daa. Again, hese alernaives consume more energy or enail more delay han he one previously described. 4 Analyical Derivaion of Energy Consumpion To analyze and compare he performance of boh echniques wih muli-channel operaion we conduc an analysis of he energy consumpion. The energy consumed by each sensor during a ime T has been calculaed. Assuming ha each node generaes packes a an average rae of λ packes/s, he number of generaed messages during T is N = λ T. Then, he goal is o find he oal energy spen (e) as he sum of he energy required o ransmi hese messages (e x ), he energy spen receiving messages from oher nodes (e rx ) and he energy wased during he duy cycle (e dc ), i.e., sleeping and periodically waking up o eiher sample all channels or o send he burs of beacons: e = e x + e rx + e dc. The analyical model presened in his work considers he following assumpions: 1. I is assumed ha, as occurs on common plaforms like he MicaZ [4] nodes, each node is equipped wih only one radio ha can be uned o work in differen channels, one a a ime.

6 28 C. Cano, B. Bellala, and M. Oliver 2. The raffic load wihin he WSN is low enough o consider he collision probabiliy negligible. This is a common assumpion in WSN analyical models andisalsousedin[1]. 3. A node sends and receives he same amoun of messages and i is assumed ha here is no overhearing, i.e., nodes only receive he packes addressed o hem. 4. Ideal or random (as done in [5] and [6]) channel selecion is assumed. Since he purpose is o evaluae he behaviour of he rendezvous mechanisms, he channel selecion sraegy does no affec he comparison. 4.1 Shor-Preamble Burs Approach Le speriod be he sampling period of a sensor node, i.e., he ime beween wo channel samples, and s he ime required o sample one channel. Therefore, he sleep ime ( sleep ) can be compued as shown in Eq. 1, where n ch is he number of available channels. sleep = speriod (n ch s ) (1) Then, he ime o ransmi a message (Eq. 2) can be calculaed as he sum of he ime o ransmi he shor-preamble burs (i has o be ransmied during he speriod o ensure a correc rendezvous), he ime o ransmi he daa ( daa ) packe and he ime o receive he ACK ( ack ). T x = speriod + daa + ack (2) The energy consumpion spen o ransmi a message can be obained by muliplying he ime spen in each mode by he power drawn in he specific mode (P i,wherei can be ransmi, receive or sleep mode) as shown in Eq. 3. E x = P x ( speriod + daa )+P rx ack (3) The oal ime and energy consumpion o ransmi all messages generaed during a ime T is: x = N T x and e x = N E x. Similarly, we can compue he ime (Eq. 4) and energy spen o receive a message (Eq. 5) assuming ha each node will receive he same amoun of messages han i generaes. In his case, i has o be considered ha he receiver will find he ransmier on average in (n ch s )/2 and ha i will only receive on average 1.5 shor preamble messages [1] of duraion shor p. T rx = (n ch s ) +(1.5 shor p )+ daa + ack (4) 2 ( ) (nch s ) E rx = P rx +(1.5 shor p )+ daa +(P x ack ) (5) 2 The oal ime receiving and he oal energy spen o receive N messages is also: rx = N T rx and e rx = N E rx, respecively.

7 Receiver-Iniiaed vs. Shor-Preamble Burs MAC for Muli-channel WSNs 29 The res of he ime ( dc = T x rx ), a node will be in duy cycle, waking up o sample all channels and sleeping. The energy spen in his operaion is shown in Eq. 6. e dc = dc (P rx (n ch s )) + (P sleep sleep ) (n ch s )+ sleep (6) 4.2 Receiver-Iniiaed Approach In he receiver-iniiaed approach, he sleep ime can be defined in a similar way. However, i has o be considered ha each node sends a oal of n ch beacons each ime i wakes up: sleep = speriod (n ch beacon ) (7) To compue he ime and energy required o ransmi a message, i has been assumed ha a ransmier samples he channels during ( sleep +(2 n ch beacon ))/2 s, on average, o find a receiver and ha i will receive, also on average, 1.5 beacons. The expression o compue ha ime is described in Eq. 8 and he corresponding energy consumpion is shown in Eq. 9. T x = sleep +(2 n ch beacon ) 2 +(1.5 beacon )+ daa + ack (8) ( ) sleep +(2 n ch beacon ) E x = P rx +(1.5 beacon ) +(P x daa )+(P rx ack ) 2 (9) The ime and energy o receive a message is compued considering ha a node sends n ch beacons, receives he daa and sends he ACK as shown in Eq. 10 and Eq. 11 respecively. T rx =(n ch beacon )+ daa + ack (10) E rx =(P x n ch beacon )+(P rx daa )+(P x ack ) (11) The energy spen in duy cycle has o ake ino accoun ha when a node wakes up i sends n ch beacons (Eq. 12): e dc = dc (P x (n ch beacon )) + (P sleep sleep ) (n ch beacon )+ sleep (12) 5 Numerical Evaluaion Using he analyical models described in Secion 4, we now conduc a numerical evaluaion. Resuls varying he packe ransmission rae wih n ch se o 5, 10, 15

8 30 C. Cano, B. Bellala, and M. Oliver Table 1. Evaluaion Parameers Parameer Value Parameer Value daa 3.2 ms ack 0.32 ms beacon = shor p 15.8 ms s 15.8 ms Time (T ) 1000 s P x 46.5 mw P rx 58.9 mw P sleep 3.6 mw Energy Consumpion (J) Shor Preamble Burs (speriod = 1) Receiver Iniiaed (speriod = 1) Shor Preamble Burs (speriod = 0.5) Receiver Iniiaed (speriod = 0.5) Energy Consumpion (J) Shor Preamble Burs (speriod = 1) Receiver Iniiaed (speriod = 1) Shor Preamble Burs (speriod = 0.5) Receiver Iniiaed (speriod = 0.5) λ (packe/s) (a) n ch = λ (packe/s) (b) n ch =10 Energy Consumpion (J) Shor Preamble Burs (speriod = 1) Receiver Iniiaed (speriod = 1) Shor Preamble Burs (speriod = 0.5) Receiver Iniiaed (speriod = 0.5) Energy Consumpion (J) Shor Preamble Burs (speriod = 1) Receiver Iniiaed (speriod = 1) Shor Preamble Burs (speriod = 0.5) Receiver Iniiaed (speriod = 0.5) λ (packe/s) (c) n ch = λ (packe/s) (d) n ch =20 Fig. 3. Energy Consumpion wih differen number of channels and raffic loads and 20 are shown in Fig. 3. The defaul parameers used for he evaluaion were measured on he TelosB plaform by Ansari e al. in [1] and are shown in Table 1. In Fig. 3, i can be observed ha when he raffic of he nework is low, he receiver-iniiaed approach consumes less energy han he shor-preamble burs echnique, independenly of he speriod and he number of channels. In he shor-preamble burs approach each node periodically wakes up o sample

9 Receiver-Iniiaed vs. Shor-Preamble Burs MAC for Muli-channel WSNs 31 he channel during n ch s. In conras, in he receiver-iniiaed approach each node periodically sends beacons during n ch beacon.givenha s has been considered equal o beacon and ha i has been assumed ha he energy o receive is higher han he energy required o ransmi, he energy consumpion for he receiver-iniiaed echnique a low loads will always be lower han he one spen in he shor-preamble burs case, independenly of he number of available channels. However, as he number of channels increases, he difference is more noable. On he oher hand, noe ha for higher nework loads he receiver-iniiaed approach also consumes less energy han he shor-preamble burs approach in mos of he cases evaluaed. The difference is especially noable when he number of channels is small and sleep is long. This difference becomes smaller when he number of channels increases and sleep is reduced. The shor-preamble burs can even show a slighly lower energy consumpion han he receiver-iniiaed approach as depiced in Fig. 3(d). For each message o ransmi, he shor-preamble burs echnique sends a burs of preambles during sleep +(n ch shor p ), independenly of when he rendezvous occurs. In conras, in he receiver-iniiaed approach, he ransmier sops lisening he channels when i receives a beacon. This happens on average in a period of ( sleep +(2 n ch beacon ))/2. Therefore, when boh he number of channels increases and sleep is reduced, he energy consumpion of he shor-preamble burs mechanism can become smaller han he consumpion of he receiver-iniiaed approach. 6 Conclusions and Fuure Work The capabiliy of coordinaing low-capable sensor nodes working a differen frequency bands is sill a research challenge. In his case, muli-channel communicaion has been considered. However, i will provide higher benefis and also higher challenges o selec variable porions of he specrum in which o ransmi based on he condiions observed. Sensor nodes will, in his scenario, have o coordinae hemselves o aggree on he porion of he specrum o use. In his work, he receiver-iniiaed rendezvous approach has been adaped o work wih muliple channels and compared o he muli-channel approach based on he ransmission of a shor-preamble burs. Resuls have shown ha he energy consumed by he receiver-iniiaed approach is smaller han he energy spen by he shor-preamble burs echnique in scenarios wih low raffic load and also in cases wih high raffic load, reduced number of channels and high sleep ime. This work aims o provide he firs comparison of boh echniques for mulichannel operaion. However, a complee evaluaion in muli-hop and large scenarios is required and has been lef as a fuure work. Moreover, new approaches need o be defined in order o reduce he unnecessary energy spen o ensure he rendezvous in a muli-channel WSNs wihou he use of a common conrol channel.

10 32 C. Cano, B. Bellala, and M. Oliver Acknowledgmens. This work has been parially suppored by he Spanish Governmen under projecs TEC (Plan Nacional I+D) and CSD (Consolider-Ingenio Program), and by he Caalan Governmen (SGR2009 #00617). References 1. Ansari, J., Ang, T., Mahonen, P.: Specrum agile medium access conrol proocol for wireless sensor neworks. In: IEEE Communicaions Sociey Conference on Sensor Mesh and Ad Hoc Communicaions and Neworks (SECON), pp. 1 9 (2010) 2. Buener, M., Yee, G., Anderson, E., Han, R.: X-MAC: A Shor Preamble MAC Proocol for Duy-cycled Wireless Sensor Neworks. In: Proceedings of he 4h Inernaional Conference on Embedded Neworked Sensor Sysems (Sensys 2006), pp (2006) 3. Cano, C., Bellala, B., Sfairopoulou, A., Oliver, M.: Low energy operaion in wsns: A survey of preamble sampling mac proocols. Compuer Neworks (2011) 4. Chipcon: CC GHz IEEE / ZigBee-ready RF Transceiver 5. Li, J., Zhang, D., Guo, L.: OMA: a muli-channel mac proocol wih opporunisic media access in wireless sensor neworks. In: Inernaional Conference on Mobile Ad-hoc and Sensor Neworks (MSN), pp (2011) 6. Li, J., Zhang, D., Guo, L., Ji, S., Li, Y.: ARM: an asynchronous receiver-iniiaed muli-channel mac proocol wih duy cycling for wsns. In: Performance Compuing and Communicaions Conference, IPCCC (2010) 7. Polasre, J., Hill, J., Culler, D.: Versaile Low Power Media Access for Wireless Sensor Neworks. In: Proceedings of he 2nd Inernaional Conference on Embedded Neworked Sensor Sysems, SenSys 2004 (2004) 8. Sun, Y., Gurewiz, O., Johnson, D.: RI-MAC: A Receiver-iniiaed Asynchronous Duy Cycle MAC Proocol for Dynamic Traffic Loads in Wireless Sensor Neworks. In: Proceedings of he 6h ACM Conference on Embedded Nework Sensor Sysems (Sensys 2008), pp (2008) 9. Tang, L., Sun, Y., Gurewiz, O., Johnson, D.: EM-MAC: A dynamic mulichannel energy-efficien mac proocol for wireless sensor neworks. In: ACM Inernaional Symposium on Mobile Ad Hoc Neworking and Compuing, MobiHoc (2011) 10. Zhou, G., Sankovic, J., Son, S.: Crowded Specrum in Wireless Sensor Neworks. IEEE EmNes (2006)