Journal of Communcatons Vol., No., November 26 Dstrbuted Adaptve Channel Allocaton n Mult-Rado Wreless Sensor Networks We Peng, Dongyan Chen, Wenhu Sun, and Guqng Zhang2,3 School of Control Scence and Engneerng, Shandong Unversty, Jnan, 254, Chna School of Infromaton and Electrcal Engneerng, Shandong Janzhu Unversty, Jnan, 25, Chna 3 The Key Laboratory of Intellgent Buldngs Technology of Shandong Provnce, Shandong Janzhu Unversty, Jnan 25, Chna Emal: pengwe928@63.com; dchen@sdu.edu.cn; sunwenhu23@63.com; qqzhang@sdjzu.edu.cn 2 Heurstc frequency assgnment n mult-rado and mult-channel FW access networks were proposed n [4]. In [5] lnear programmng s used to solve channel and power assgnment problems n mult-power mult-rado WSNs. RF transcever channel bnd mechansm was proposed n [6]-[9], whch bnd each rado nterface to a channel permanently or for a long tme. The statcally bndng channel allocaton scheme can be easly acheved, but s lack of flexblty for channel s change. Besdes that, t never consders the dfference and varaton between channels n channel bndng procedure. Mult-rado mult-channel meda access control protocol for mult-rado mult-channel WSNs was proposed n []-[3]. Those protocols were desgned based on control channel negotaton. Ths scheme has some drawbacks. Frst, the control channel competton and data channel swtch nevtably lead to data transmsson delay. Second, the control channel may become a bottleneck n communcaton. When the control channel s nterfered or can not communcate properly, t wll nevtably lead to the degradaton of entre network s performance. In [4], both recever-based and lnk-based dstrbuted channel allocaton protocols were proposed. However, frstly, the protocol s stll based on sngle rado. It may not be sutable for mult-rado. Secondly, the dstrbuted channel allocaton algorthm s stll a statc allocaton. It dd not solve the channel update problem that s cased by channel dynamcally change. Channel allocaton based on Galos Feld and channel-hoppng scheme n mult-rado mult-hop wreless networks were proposed n [5]. Although ths mechansm can solve mult-rado channel allocaton problem n mult-rado mult-hop wreless networks, there are some problems to be solved. Frst, t requres strct tme synchronzaton between nodes. Ths s sometmes dffcult n practce. Second, node needs to swtch the channel n each channel-hop slot. Ths causes addtonal communcaton delay. Last, the channel allocaton procedure does not consder the dfference between channel qualtes. At any tme slot, f nodes use bad channel, ths wll nevtably cause many problems, such as too many retransmssons, lower transmsson success rate, channel congeston or even can not communcaton. The worst s that the problem Abstract To solve the problem of channel allocaton n mult-rado Wreless Sensor Networks (WSNs), ths paper proposed dstrbuted channel allocaton and adaptve update scheme based on channel utlzaton and Data Transmsson Success Rate (DTSR). Frst, node perodcally calculates the channel utlzaton and DTSR on each data channel. Second, node can adaptvely adjust the statstcal perod and update some data channels based on the calculaton result. Ths can effectvely avod unnecessary channel swtch and keep each RF transcever workng on relatve optmal channel. Last, we do some experments to verfy the effectveness of the proposed scheme and gve some performance comparson (e.g., delay an delvery rato) wth other channel allocaton protocols. Experments result show that our proposed scheme can ensure each RF transcever work on the relatve optmal channel based on adaptve update mechansm and effectvely estmate the channel qualty based on DTSR. Comparson results show that the proposed scheme can mprove some performances. Index T erms WSNs, mult-rado, channel allocaton, channel qualty estmaton I. INTRODUCTION The rapdly development of mcroelectroncs technology dramatcally decreases the cost and mproves the performance of wreless RF transcever and makes t s possble to equp multple RF transcevers on one node. The technology of mult-rado provdes a new way to solve network performance degradaton ssue caused by channel competton and channel qualty fluctuaton. The advantages of mult-rado technology were verfed n []-[3]. In fact, there are multple non-overlappng, orthogonal channels whch can be avalable n WSNs. In [4], channel allocaton had been proved to be a NP-hard problem n WSNs. Thus, studyng relable and effcent channel allocaton mechansm n mult-rado WSNs has mportant theoretcal and practcal sgnfcance. Manuscrpt receved July 3, 26; revsed November 2, 26. Ths work was supported by the Natonal Natural Scence Foundaton of Chna under Grant NO. 6573225. Correspondng author emal: pengwe928@63.com do:.272/jcm...984-99 26 Journal of Communcatons 984
Journal of Communcatons Vol., No., November 26 tme and end tme. Channel utlzaton can be derved as follows. Accu CU Tcu wll always exst n each channel-hoppng cycle. To solve above mentoned problems, we proposed mult-rado WSNs adaptve dstrbuted channel allocaton algorthm. Ths algorthm not only consders the balance of channel allocaton, but also desgns adaptve channel update mechansm based on channel utlzaton and channel qualty estmaton. To the mutaton of channel utlzaton and channel qualty, ths algorthm can quckly fnd and update the channel to ensure each RF transcevers equpped on one node workng on the optmal non-overlappng, orthogonal channel. The rest of the paper s as follows. Secton II presents the nformaton statstcs of channels. Secton III provdes the steps of channel allocaton and adaptve update scheme. Secton IV evaluates our scheme and ts performance. At last, we conclude the paper n secton V. (2) where Tcu s statstcal perod, Accu s accumulatve used tme of channel wthn the same Tcu, CU s channel utlzaton of channel wthn the same Tcu. The statstcal perod s denoted as Tcu, whch s adaptve. Tcu (Tcu _ mn, Tcu _ max ), Tcu _ mn and Tcu _ max are mnmum and maxmum values. Nodes exchange channel utlzaton nformaton through control channel. Table I shows the statstcal channel utlzaton nformaton. TABLE I: STATISTICAL CHANNEL UTILIZATION INFORMATON II. CHANNEL INFORMATION STATISTICS Ths paper mostly focuses on mult-rado WSNs adaptve dstrbuted channel allocaton. We make some reasonable assumptons as follows: ) All the M half-duplex RF transcevers equpped to one node have the same transmt power. Ths can effectvely avod hdden termnal problem caused by dfferent transmt power between nodes. 2) All Nodes have sustaned and stable power supply, so we don t need to consder low power problem. 3) Channels are dvded nto N (N>M) orthogonal, non-overlappng channels. Each channel has the same parameters. One channel s fxed as control channel to dscover neghbors and exchange topology nformaton and channel nformaton. Remaned N- channels are avalable as data channels. Data channels are used for allocaton and data communcaton. 4) Every node n the same network has a unque ID. The ID s generated by node s MAC address and called Node_ID. Besde, each RF trancever equpped to one node has correspondng ID. The ID s unque wthn the same node and called RF_ID. To ensure M RF transcevers work on the optmal channels and avod frequently channel swtch, node perodcally calculate channel utlzaton, DTSR and the number of N- data channels allocaton. Channel No Ch Start Tme St End Accu CU Ch j St j End j Accu j CU j Chk Stk Endk Accuk CU k End Tme Channel Utlzaton B. Channel Qualty Estmaton Each packet transferred by the same RF transcever wthn the same Tcu s ndvdually marked wth a contnuous and monotoncally ncreasng sequence number. If the transmsson s faled, the retransmtted packets are re-marked accordng to ncreasng sequence order. By contnuous statstcs, the total number of transmtted packets and the numbers of successfully transmtted packets can be obtaned. Then, node calculates DTSR accordng to statstc result and estmates channel qualty based on currently DTSR. We use the number of extra expected transmssons to represent channel qualty. So, the lower estmated result represents the better channel qualty, vce versa. DTSR and Channel Qualty Estmaton (ChQltyEst) can be derved as follows. DTSR ( k ) DTSR ( k ) ( ) A. Channel Utlzaton Each node records start tme, end tme and calculate accumulated used tme of all the data transmsson (sendng or recevng). The start tme, end tme and accumulatve used tme of channel are denoted as St, ExtTransEst (k ) DTS ( k ) ( DTSR (k ) ) (4) DTSR (k ) ( ) ExtTransEst ( k ) follows: (3) DTT (k ) ChQltyEst (k ) ChQltyEst (k ) End and Accu. Accu whch can be derved as Accu End ( k ) St ( k ) Accumulatve Tme (5) where s called smooth factor. Its value s from to n (). DTT ( k ) and DTS (k ) are the total number of transmtted packets and the number of successfully where k s the used number of a RF transcever wthn transmtted packets wthn the k th Tcu of channel. the same Tcu. St ( k ) and End (k ) are the k th start DTSR (k ) s data successful transmsson rate wthn k 26 Journal of Communcatons 985
Journal of Communcatons Vol., No., November 26 the k th T cu of channel. ExtTransEst ( k) s the requred number of extra transmssons of channel n currently DTSR ( k ). ChQltyEst ( ) k s the value of estmated channel qualty wthn the k th T cu of channel. Node updates local nformatons ncludng channel utlzaton, DTSR and ChQltyEst, then broadcasts them to neghbors through control channel. Neghbors receve the broadcasted nformaton and update local nformatons, then repeat the above procedures untl all nodes complete the updates. III. CAHNNEL ALLOCATION AND UPDATE A. Data Channels Allocaton In the dstrbuted channel allocaton algorthm, node mantans three data structures: Neghbor Informaton (NI) table whch records neghbor s and some nformaton about RF transcevers; Channel Allocaton (CA) table whch records the number of Channel Allocaton threshold (ChAlloc_th) and Channel Allocaton Counter (CAC), when the data channel s allocated one tme, the correspondng counter ncreases one. If the value of counter exceeds the ChAlloc_th, the correspondng channel wll no longer be allocated. Ths can prevent some data channels from beng over-allocated. Channel Usage Informaton Statstcs (CUIS) table, records the channel usage nformaton whch nclude channel utlzaton and DTSR. We use DTSR to estmate channel qualty. Accordng to the CUIS and the result of ChQltyEst, node decdes whether t needs to update the data channel whch t s usng. The procedures of dstrbuted channel ntally allocaton manly conssts of the followng steps: ) Node generates a random number that s used as random delay. Meanwhle, t ntalzes CA table and CUIS table. Channel utlzaton s ntalzed to be zero. DTSR and ChQltyEst are ntalzed to be one. 2) Nodes exchange nformaton through control channel, construct NI table and update CA table. 3) When the random delay tmeout, node randomly selects a data channel from N- data channels. Then, t checks the channel accordng to the followng condtons. () The channel has never been allocated n the NI table. (2) The channel has been allocated, but the value of counter not exceeds the ChAlloc_th. If the selected channel meets one of the above condtons, node allocates the channel to RF transcever. Otherwse, node randomly selects another data channel agan and repeats the above procedure, untl successfully allocates channel to the RF transcever. 4) Node updates CA table and broadcasts the channel allocaton nformaton to ts neghbors. 5) Nodes repeat the steps 2)-4) accordng the RF_ID sequences, untl all of M RF transcevers are successfully allocated wth data channels. Algorthm provdes pseudocode whch descrbes the ntally allocated procedures of data channel. Algorthm : data channels ntally allocated : Intalzes CA and CUIS table, generates: Random_delay; 2: Communcate wth other nodes, update NI table and CA table; 3: f(random_delay s tmeout) then 4: for(=;<m) do 5: temp_ch = Randomly select a channel; 6: f(temp_ch not exceeds thresholds) then 7: rado[] = temp_ch; 8: updates CA table and broadcasts the allocate nformaton 9: ++; : contnue; : else 2: contnue; 3: end f 4: end for 5: end f In the step 3), ChAlloc_th s decded by the number of neghbor nodes and the number of data channels. We can balance N- data channels allocaton based on NI table and CA table. Ths process makes sure M RF transcevers can work on the optmal orthogonal, non-overlappng channels wthn the nterference range. B. Data Channels Adaptve Update Channels are usually subject to electromagnetc nterference and co-nterference between nodes. It nevtably leads to the degradaton of network s performance, such as greater communcaton delay, lower successful communcaton rate, lower throughput and so on. So, when ntal allocaton s fnshed, node must dynamcally update data channels accordng to the change of channel parameters. To acheve adaptve update of data channels, node needs to mantan three thresholds. They are ChAlloc_th, Channel Utlzaton Threshold (ChUtl_th) and Channel Qualty Threshold (ChQlty_th). ChAlloc_th s the maxmum number that each data channel can be allocated. It can prevent some data channels from beng over-allocated. ChUtl_th s the maxmum value of channel utlzaton. It can prevent some data channels from beng hot pont. ChQlty_th s the mnmum value of channel qualty. If a channel s qualty s greater than ChQlty_th, t can be allocated and used. Otherwse, t can t. Ths can avod some bad channels to be allocated. If one of the followng condtons s fulflled, node wll update some data channels on whch some RF transcevers work. ) The allocaton counter of channel s greater than ChAlloc_th, namely CAC ChAlloc _ th. Ths means that the channel s over-allocated. 2) The channel utlzaton of channel s greater than ChUtl_th, namely CU ChUtl _ th. Ths means that the channel s over-used. 3) The ChQltyEst of channel s less than ChQlty_th, namely ChQltyEst ( k) ChQlty _ th. Ths means that he channel qualty s too bad to be used. Adaptve update procedure of data channels manly 26 Journal of Communcatons 986
Journal of Communcatons Vol., No., November 26 conssts of the followng steps: IV. EVALUATION ) Node sets the statstcal perod Tcu Tcu _ mn and We evaluate our adaptve dstrbuted channel allocaton algorthms on the mult-rado WSNs testbed whch we desgned and constructed. Fg. shows the mult-rado wrless sensor node. We can see each node s equpped wth four RF transcevers. Besde, each node has two mn USB nterfaces through whch we can easly observe the debug nformaton. We ported the Real-Tme Operaton System FreeRTOS [6] to our testbed by whch the multple tasks coordnate work s acheved, such as the nter-nodes communcaton tasks, the channel utlzaton statstcs task, the channel qualty estmaton task, and so on. We deployed the testbed over our laboratory floor. The testbed s conssted by 5 mult-rado nodes. Fg. 2 shows the testbed deployment and topology. All mult-rado nodes are made wthn one-hop range. They are neghbors from each other. ① ~ ⑬ are calculates CAC, channel utlzaton and ChQltyEst, when the ntal allocaton s completed. 2) After contnuous ten tmes Tcu, f there s no data channel meets any one of the above condtons, then node sets the statstcal perod Tcu Tcu + cu and repeats the above statstcal process, untl Tcu Tcu _ max. cu s an ncremental step. 3) After step 2), node uses Tcu _ max as statstcal perod and repeats the above statstcal process. Once a channel meets one of the above channel update condtons, node mmedately adjusts Tcu to Tcu _ mn and quckly calculates CAC, channel utlzaton and ChQltyEst. After contnuous ten tmes Tcu, f ths data channel stll meets the channel update condtons, node wll update the data channel, broadcast the result of update and repeat steps )-3). Algorthm 2 provdes adaptve channel update pseudocode. mult-rado nodes. s the snk node, also called root node. s the nterference node. To collect the nformaton of network, s connected to a PC va the mn USB nterface. s connected to another PC va mn USB nterface. The PC to whch s connected can assgn any data channel, the destnaton of packets, transmsson frequency, duraton and the marked number of packet to. Through ths testbed, we can smulate dfferent channel utlzaton and channel qualty, and verfy the effectveness of adaptve channel update algorthm whch we proposed. Algorthm 2: channels adaptve update temp_ch Select the optmal canddate channel; 2 f(temp_ch don t fulflls any update condtons) then 3 rado_ch = temp_ch; 4 else 5 select another channel and repeat the process; 6 end f 7 Update CA table and broadcast channel allocated nformaton C. Complexty Analyss Accordng to the data channel adaptve update process, to N- data channels the channel allocaton counter, the statstcal process of channel utlzaton and DTSR, and the process of compare wth each threshold, those process can fnsh wthn O() tme. So, the tme complexty of the whole algorthm s O(N-). In ths algorthm, every node needs to record the nformaton of each RF transcevers whch are equped on ts neghbor s node. We use two-dmensonal lnear table to record the nformaton and provde O() data query servce. So, f the number of neghbors s h, the space complexty of ths algorthm s O(hm). Fg. 2. Testbed deployment and topology The PC to whch s connected runs web-based montorng software to acheve real tme data collecton, analyss and statstcs. Based on ths, we can verfy the effectveness and adaptaton of the algorthm that we proposed. By collected nformaton, the montorng software can track the whole network s nformaton. A. Balance of Channel Allocaton We set sx channels n our deployed testbed. They are numbered as Ch~Ch6. We fxed Ch6 as control channel. Other channels are used as data channels. To verfy the balance of channel allocaton, we frstly remove the Fg.. Mult-Rado node 26 Journal of Communcatons 987
Journal of Communcatons Vol., No., November 26 nterference node I from the network. ChAlloc_th s frstly setted to be xa, ChUtl_th to be one and ChQlty_th to be zero. Table II shows the part result of ntally data channel allocaton n mult-rado WSNs. In RF column, R=~R=3 represent three RF transcevers equpped wth one node. In Chx columns, one represents the channel s allocated, zero represents the channel s not allocated. Fg. 3 shows the fnal results of channel allocaton of the whole network. Results show that Ch~Ch5 are relatvely balanced allocated. utlzaton, DTSR and estmate channel qualty wthn mnutes after the networks stable. Fg. 4 and Fg. 5 show the statstcal results of channel utlzaton and DTSR. From Fg. 4 and Fg. 5, we can observe that channel utlzaton s no greater than 4% and the DTSR s no less than 93%. Channel utlzaton and DTSR are relatvely stable, after the channel allocaton was completed and the network was stable. All of the results verfy the effectveness of the proposed dstrbuted channel allocaton algorthm. TABLE II: RESULT OF DATA CHANNELS INITIALLY ALLOCATION Node RF Ch Ch2 Ch3 Ch4 Ch5 3 8 2 R= R=2 R=3 R= R=2 R=3 R= R=2 R=3 Fg. 5. Results of DTSR B. Channel Adaptve Update ) Interference node I contnuously sends packets to node 3 on Ch2. Adjust transmsson frequency to smulate dfferent channel utlzaton and verfy channel adaptve update process wth dfferent channel utlzaton. Fg. 3. Fnal results of channel allocaton Fg. 6. Interferece of channel utlzaton (short tme) Fg. 4. Results of channel utlzaton Nodes randomly select neghbor and send packets after completng the channel allocaton. In ths process, node calculates channel utlzaton and DTSR, estmates channel qualty and sends the statstcal result to PC va snk node. We contnuously calculate channel Fg. 7. Results of channel update (short tme) 26 Journal of Communcatons 988
Journal of Communcatons Vol., No., November 26 Once the channel utlzaton exceeds the threshold ChUtl_th, the node mmedately makes Tcu Tcu _ mn and quckly calculates CAC, channel utlzaton and ChQltyEst. Although the channel utlzaton exceeds the threshold ChUtl_th, the duraton s less than ten tmes Tcu _ mn. That means there are short tme and bursty data transmsson on the channel. To mantan the network stable, node doesn t need to update the channel, as shown n Fg. 6 and Fg. 7. When the duraton s greater than ten tmes Tcu _ mn, Fg.. Interferece of DTSR (short tme) node wll start the adaptve channel update process to reduce the number of ths channel allocaton and exchange the update result wth ts neghbors. So, neghbor nodes can avod usng ths channel, as shown n Fg. 8 and Fg. 9. Fg.. Results of channel update (short tme) Fg. 8. Interferece of channel utlzaton (long tme) Fg. 2. Interferece of DTSR (long tme) Fg. 9. Results of channel update (long tme) In summary, when channel utlzaton s greater than ChUtlz_th and duraton s larger than ten tmes Tcu _ mn, node wll start adaptve channel update process to avod usng the channel. Ths can balance the channel utlzaton, avod congeston and ensure the real-tme data transmsson n network. 2) Interference node adjusts the marked number of packets to reduce DTSR and verfes adaptve channel update process when the DTSR s less than the threshold ChQlty_th. Fg. ~ Fg. 3 shows the results of channel update wth lower DTSR. 26 Journal of Communcatons Fg. 3. Results of channel update (long tme) DTSR s less than the threshold ChQlty_th, and duraton s less than ten tmes Tcu _ mn. That means channel qualty happened ephemeral fluctuaton. To keep the stablty of network, node wll not update channels, as shown n Fg. and Fg.. 989
Journal of Communcatons Vol., No., November 26 When duraton s larger than ten tmes T cu _ mn, node wll start channel update process to avod usng the channel, as shown n Fg. 2 and Fg. 3. Ths can effectvely reduce retransmsson whch s caused by bad channel qualty. Fg. 2 shows the channel utlzaton frstly ncreases then decreases. The reason s bad channel qualty cause many retransmssons. So, the channel utlzaton ncreases. After the node completed channel update and reduced the number of ths channel allocaton, the channel utlzaton decreased. In summary, when DTSR s less than threshold and duraton s greater than ten tmes T cu _ mn, node wll start channel adaptve update process to avod usng the bad channel. Ths can ensure each RF transcevers work on the optmal data channel. C. Performance Evaluaton From Subsecton IV-A and Subsecton IV-B, we verfed that the proposed algorthm can effectvely adjust the data channel adaptvely wth dfferent type nterference. Now, we evaluate the performance of ths algorthm. We select nodes 3 and 2 as sender and recever. Form the Table II, we know node 3 use Ch3, Ch2 and Ch5 as data channels and node 2 use Ch3, Ch4 and Ch. The sender try ts best to send a 64byte data packet every 5ms, the recever try ts best to receve all the data. The nterference node I s frstly removed from the networks, then jon n and generate nterference on two data channels (Ch4 and Ch) whch the recever used respectvely every ms. We mplement the proposed channel allocaton mechansm and compare ts performance wth the Channel Bnd Mechansm (CBM) proposed n [6]-[9] and the Control Channel Negotaton (CCN) proposed n []. Fg. 4 and Fg. 5 show the results of packet delvery delay and packet delvery rato. the sender need more tme to negotate an dle channel. Whle for CBM, t s hard to fnd a channel that s not nterfered and can be used to transmt data packets. Fg. 5. Packet delvery rato From Fg. 5, we can observe clearly that our mechansm can effectvely avod usng nterfered channel and mprove the delvery rato. Especally, n the worst case, our mechansm can keep the delvery rato greater than 98%. It s much better than CCN and CBM. V. CONCLUSIONS The technology of mult-rado n WSNs can obvously mprove some performances. One of the most mportant problems n mult-rado WSNs s how to effectvely allocate the wreless channel. To solev the above problem, a dstrbuted adaptve channel allocaton mechansm based on channel utlzaton, DTSR and the number of channel allocaton was proposed. Frstly, node calculates the channel utlzaton, DTSR and estmates the channel qualty by DTSR. Secondly, node adaptvely adjusts the statstcal perod and channel update process based on channel qualty and utlzaton. Fusng the above mentoned nformatons, the scheme proposed n ths paper can ensure each RF transcever equped on one node workng on relatve optmal channel. We evaluated some performances of our algorthm through experments. Results verfed the effectveness of the mechansm we proposed. In the future work, we wll further optmze the complexty of ths algorthm and evaluate the performance of proposed algorthm based on some other topologes (e.g., random topology). ACKNOWLEDGMENT Fg. 4. Packet delvery delay From Fg. 4, t s clear that the delvery delay of our mechansm s much less than CCN and CBM, even though the two data channels are nterfered. Because, the node whch uses our mechansm can quckly fnds nterference and adaptvely adjust the data channels. In the worst case (two data channels are nterfered), for CCN, Ths work was supported by the Natonal Natural Scence Foundaton of Chna under Grant NO. 6573225. The authors would lke to thank the anonymous revewers for ther nsghtful suggestons that wll help us to mprove the qualty of ths paper. REFERENCES [] A. Safullah, Y. Xu, C. Lu, and Y. Chen, End-to-End 26 Journal of Communcatons 99
Journal of Communcatons Vol., No., November 26 [2] [3] [4] [5] [6] [7] [8] [9] [] [] [2] [3] [4] vol. 25, no. 9, pp. 2264-2274, September 24. [5] T. Y. Ln, K. R. Wu, and G. C. Yn, Channel-Hoppng scheme and channel-dverse routng n statc mult-rado mult-hop wreless networks, IEEE Transactons on Computers, vol. 64, no., pp. 7-86, Jan. 25. [6] FreeRTOS. [Onlne]. Avaable: http://www.freertos.org communcaton delay analyss n ndustral wreless networks, IEEE Transactons on Computers, vol. 64, no. 5,pp. 36-374, May 25. L. H. Yen and K. W. Huang, Lnk-Preservng nterference-mnmzaton channel assgnment n mult-rado wreless mesh networks, Internatonal Journal of Ad Hoc and Ubqutous Computng, vol. 8, no. 4, 25. T. F. Wykret, L. H. A. Correa, D. F. Macedo, J. C. Gacomn, and L. T. Andrade, Evaluaton and avodance of nterference n WSN_ A mult-rado node prototype usng dynamc spectrum allocaton, n Proc. IFIP Conf, Valenca, 23, pp. -3. S. Sabno, N. Correa, and A. Barradas, Frequency assgnment n mult-channel and mult-rado FW access networks, n Proc. 9th Iberan Conference on Informaton Systems and Technologes, Barcelona, 24, pp. -6. J. L, X. Guo, and L. Guo, Jont routng, schedulng and channel assgnment n mult-power mult-rado wreless sensor networks, n Proc. IEEE 3th Internatonal IPCCC Conf, Orlando, 2, pp. -8. M. Alcherry, R. Bhata, and L. E. L, Jont channel assgnment and routng for throughput optmzaton n multrado wreless mesh networks, IEEE Journal on Selected Areas n Communcatons, vol. 24, no,, pp. 96-97, November 26. K. Das, R. Vjayakumar, and S. Roy, WLC3-4: Statc channel assgnment n mult-rado mult-channel 82. wreless mesh networks: Issues, metrcs and algorthms, n Proc. Global Telecommuncatons Conference, San Francsco, 26, pp. -6. T. Y. Ln, W. H. Tam, K. L. Fan, and Y. C. Tseng, Resource plannng and packet forwardng n mult-rado, mult-mode, mult-channel, mult-rate (M4) wreless mesh networks, Elsever Computer Communcatons, vol. 3, no. 7, pp. 329-342, May 28. A. Ranwala and T. C Chueh, Archtecture and Algorthms for an IEEE 82.-based Mult-Channel wreless mesh network, n Proc. IEEE INFOCOM Conf, March 25, pp. 2223-2234. Z. Lu and W. Wu, A dynamc mult-rado mult-channel MAC protocol for wreless sensor networks, n Proc. IEEE Internatonal Conference on Communcaton Software and Networks, Sngapore, 2, pp. 5-9. M. D. Felce, G. Zhu, and L. Bonon, Future channel reservaton medum access control (FCR-MAC) protocol for mult-rado mult-channel wreless mesh networks, n Proc. ACM PE-WASUN, October 28. R. Huang, H. Zha, C. Zhang, and Y. Fang, SAM-MAC: An effcent channel assgnment scheme for mult-channel ad hoc networks, Elsever Computer Networks, vol. 52, no. 8, pp. 634-646, June 28. S. L. Wu, C. Y. Ln, Y. C. Tseng, and J. P. Sheu, A new mult-channel mac protocol wth on-demand channel assgnment for mult-hop moble ad hoc networks, n Proc. Int l Symposum on Parallel Archtectures, Algorthms and Networks, Dallas 2, pp. 232-237. A. Safullah, Y. Xu, C. Lu, and Y. Chen, Dstrbuted channel allocaton protocols for wreless sensor networks, IEEE Transactons on Parallel and Dstrbuted Systems, 26 Journal of Communcatons We Peng was born n Shandong Provnce, Chna, n 986. He receved the B.S. degree from Guangx Unversty of Scence and Technology, Guangx, n 2 and the M.S. degree from Shandong Janzhu Unversty, Shandong, n 23. He s currently pursung the Ph.D. degree wth School of Control Scence and Engneerng, Shandong Unversty. Hs research nterests nclude WSN protocol and Industral Wreless Networks. Dongyan Chen receved the B.Eng. degree n electrcal engneerng from the Southeast Unversty, Nanjng, Chna, n 995, the M.S. degree n communcaton engneerng from Nanyang Technologcal Unversty, Sngapore, n 999, and the Ph.D degree from Duke Unversty, Durham, NC, n 23. He s currently a Full Professor wth School of Control Scence and Engneerng, Shandong Unversty. Hs man research nterests nclude WSN protocol development and applcaton, Industral Hgh Speed Wreless Networks. Wenhu Sun was born n Shandong Provnce, Chna, n 989. She receved the B.S. degree from Qlu Unversty of Technology, Jnan, n 2 and the M.S. degree from Donghua Unversty, Shangha, n 24. She s currently pursung the Ph.D. degree wth School of Control Scence and Engneerng, Shandong Unversty. Her research nterests nclude Bg Data analyss and applcaton. Guqng Zhang receved the B.E. degress form Shandong Janzhu Uversty, Jnan, Chna, n 986 and the Ph.D. degree from the X an Jaotong Unversty, X an, Chna, n 22. He s currently a Full Professor of the School of Informaton and Electrcal Engneerng, Shandong Janzhu Unversty. Hs research nterests nclude ntellgent control methods, ntellgent buldngs, smart home, and Internet of thngs. He has publshed more than 4 journal and conference papers and holds more than ssued paten tmests. He has served as the Program Commttee Member of several nternatonal conferences and s a Revewer for several nternatonal journals and conferences. 99