A Fast and Safe Industrial WLAN Communication

Transactions of th ISCIE, Institut Vol. 29, of Systms, No. 1, pp. Control 29 39, and 216 Transactions Information Enginrs of ISCIE, Vol. 29, No. 1, pp. 29 39, 216 29 Papr A Fast and Saf Industrial WLAN Communication Protocol for Factory Automation Control Systms* Duc Khai Lam, Yasuhiro Shinozaki, Kishi Yamaguchi, Satoshi Morita, Yuhi Nagao, Masayuki Kurosaki and Hiroshi Ochi In this work, an industrial Wirlss Local Ara Ntwork (iwlan) systm usd to control industrial robots (irbs) in factory automation (FA) nvironmnts is addrssd. For fast and saf communications, w first propos a synchronous multi-usr round-robin transmission protocol. To rduc th ovrhad causd by th convntional multi-usr downlink transmission tchniqu, w thn propos a low ovrhad Packt Division Multipl Accss () transmission tchniqu. Th numrical and simulation rsults indicat that our approachs incras spd of 1% and 3% throughput compard to th convntional industrial Point Coordination Function (ipcf) and Spac Division Multipl Accss (SDMA) approachs, rspctivly. Bsids, our approachs also provid lowr systm rror rat (SER) than convntional ons. In particular, th control duration pr irb in our iwlan systm is fastr than 1 µsc. Thrfor, our proposd schms can achiv fast and saf prformanc for FA communication systms. 1. Introduction In rcnt yars, wirlss tchnology has mrgd as a promising altrnativ for industrial communication bcaus it can provid th flxibility ndd for mobil robot sttings. Th wirlss tchnology also is an advantag for th FA systms by rducing th cabl connctivity and maintnanc costs[1]. Thr ar svral solutions for FA wirlss communication. Blutooth, which is basd on IEEE 82.15.1 protocol, is widly usd for short-rang transmission applications[2]. For gratr transmission rang, Zigb, which is basd on IEEE 82.15.4 protocol, is proposd [3,4]. Although Zigb provids long-rang transmission, its slow data rat maks it only suitabl for wirlss snsor ntwork applications[5]. In ordr to improv th data rat and rliability of Zigb, th industrial communication standard WirlssHART, which is also basd on th IEEE 82.15.4, has bn rlasd[6]. Howvr, th disadvantag of WirlssHART is th poor compatibility with IEEE 82.11 WLAN dvics[7]. Th IEEE 82.11 WLAN tchnology provids highr data rat, gratr rang than th WirlssHART. In addition, it is asy to coxist with othr WLAN offic dvics. For thos rasons, th industrial communication systms basd on IEEE 82.11 WLAN tchnologis hav bn gradually dvlopd in th FA fild. In[8], th rtransmission algorithm is introducd to Manuscript Rcivd Dat: March 25, 215 Kyushu Institut of Tchnology, 68-4 Iizuka, Fukuoka 82-852, JAPAN Ky Words: industrial WLAN, factory automation control, synchronization communication,. optimiz th rtransmission numbr to incras th throughput of th IEEE 82.11g. In[9], th nhancd distributd channl accss (EDCA) is invstigatd to provid low latncy for th IEEE 82.11. Howvr, sinc both of thm oprat undr th Distributd Coordination Function (DCF) accss tchniqu, th squnc of transmissions is non-dtrministic. Th dtrminism is an important factor for saf industrial communication systms. To dal with this problm, PROFINET has offrd an industrial WLAN systm using ipcf protocol[1]. In this protocol, th Mastr (MS) controls th squnc of transmissions to irbs by using a polling mchanism. Howvr, th throughput of this protocol is slow bcaus it supports only singl usr (SU) transmissions [11,12]. In particular th numbr of irbs in FA systm is larg, th duration for on round of all irbs communications can gt so big, that it causs ral-tim control failur. In ordr to mitigat th aformntiond problms of th currnt FA systms, w propos a novl iwlan systm. Th contributions of this papr includ: (1) For fast and saf communications, w propos a synchronous multi-usr (MU) round-robin transmission protocol. (2) To rduc th critical ovrhad causd by th convntional MU - multipl inputs and multipl outputs (MIMO) downlink (DL) SDMA transmission tchniqu, w propos a low ovrhad transmission tchniqu. Th numrical and simulation rsults indicat that our approachs incras spd of 1% and 3% throughput compard to th convntional ipcf and SDMA approachs, rspctivly. Bsids, our approachs also provid lowr SER than convntional ons. In particular, th control duration 29

3 Transactions Trans. of ISCIE, of ISCIE, Vol. Vol. 29, No. 29, No. 1 (216) 1 (216) pr irb in our iwlan systm is fastr than 1 µ sc. Thrfor, our proposd schms can achiv fast and saf prformanc for FA communication systms. Th rst of this papr is organizd as follows. Sction 2. shows th FA wirlss communication systm modl with th convntional and proposd transmission protocols. In Sction 3., w show th convntional and proposd transmission tchniqus for MU- DL accss. Sction 4. compars th prformanc btwn our proposd systm with th convntional ons undr non-rror basd transmissions. In Sction 5., th prformanc of proposd systm undr rror basd transmissions is analyzd. In Sction 6., w show th numrical and simulation rsults undr rror basd transmissions. Finally, Sction 7. outlins our conclusions of this papr. 2. FA Wirlss Transmission Protocols Th FA wirlss communication systm modl is shown in Fig. 1. Th systm includs an MS and N irbs. Th communications btwn MS and irbs ar controlld by a transmission protocol at th mdia accss control () layr and a transmission tchniqu at th physical () layr. Th transmission bandwidth is 8 MHz. In FA communication systms, th control data, which ar transmittd btwn MS and irbs, ar short in lngth. Thrfor, in this papr th control data of 32 octts in lngth ar considrd. Th FA wirlss communication systm supports th dtrministic transmission fatur, in which th communications from MS to ach irb ar in qual portions and in circular ordr without priority. In this sction, w show th transmission protocols for FA wirlss systms. First, w prsnt th convntional ipcf transmission protocol, which is providd by PROFINET. Thn, w propos an iwlan transmission protocol. 2.1 Convntional ipcf Protocol Th ipcf protocol supports th dtrministic fatur. This protocol is dvlopd basd on th PCF protocol from IEEE 82.11 standard[13]. Th ipcf provids Contntion Fr Priod (CFP) for fram transfrs. Th squnc transmissions of irbs is dtrmind by a contntion fr polling mchanism. Fig. 2 shows th protocol diagram of PCF. In this protocol, MS uss a round-robin polling with SU to communicat with N irbs. During th communication with ach irb, if DL or uplink (UL) transmission rrors ar dtctd, MS will rtransmit DL fram aftr Short Intrfram Spac (SIFS) or PCF Intrfram Spac (PIFS) intrvals, rspctivly. MS announcs th nd of CFP by broadcasting th CF- End fram. Bcaus th ipcf protocol supports only Tim Division Multipl Accss (TDMA) tchniqu with SU transmissions for both DL and UL, th throughput of th ipcf systm is slow. Fig. 1 T bacon Bacon MS irb MS irb 1 irb 2 irb 3 irb N FA wirlss communication systm modl SIFS Communication with irb 1 T dl=52 us SIFS T ul=52 us SIFS DL T irb UL munication Fig. 2 Control Duration T cd Communication with irb 2 Communication with irb N If communication rror occurs during ithr DL or UL, it can b r-communicatd up to m tims Convntional ipcf protocol T cfnd CF-End 2.2 Proposd iwlan Protocol In ordr to mitigat th limitation of SU transmissions in th ipcf protocol, w propos an advancd synchronous MU round-robin iwlan transmission protocol. Bsid th dtrministic fatur, our protocol also supports MU accss for both DL and UL transmissions. Fig. 3 shows th protocol diagram of our proposd iwlan. Th transmission squnc oprations is shown in Fig. 4. Th squnc oprations of protocol includs 3 stags: Stag 1: MS starts a nw control duration (T cd ) by broadcasting a synchronization (SYNC) fram to all irbs on th bandwidth of 8 MHz. This SYNC fram carris th information of th Prcision Tim Protocol (PTP). Th duration of SYNC fram transmission is dnotd by T sync. Stag 2: MS communicats with all N irbs. This communication duration is dnotd by T com. This intrval includs th following oprations: DL transmission: Aftr th wait of SIFS duration (T sifs ), MS transmits th MU-DL fram to irbs on th bandwidth of 8 MHz. W support MU-DL fram of up to 4 usrs. Th numbr of usrs on MU- DL fram is dnotd by u. Th valus of u can b 1, 2, 3 or 4. Th duration of DL transmission is dnotd by T dl. UL transmission: If irb succssfully rcivs its data from DL fram, it will rspond to MS aftr T sifs by snding UL fram using bandwidth of 2 MHz. Up to 4 irbs ar supportd to snd thir UL frams simultanously using Frquncy Division Multipl Accss (FDMA) tchniqu. Th duration of UL transmission is dnotd by T ul. Th DL and UL transmissions ar prformd until MS finishs snding data to all of N irbs. An irb rror occurs if th irb fails to ithr rciv th DL packt from MS or transmit th UL packt to MS. Stag 3: If thr ar any irb rrors during stag 3

Lam, Shinozaki, Yamaguchi, Morita, Nagao, Kurosaki and Ochi: A Fast and Saf Industrial WLAN Communication Protocol 31 MS irbs Tsync Tsifs Tcom Trcom SYNC COMMUNICATION RE-COMMUNICATION SYNC Tdl DL {1-4} UL1 UL2 UL3 UL4 Control Duration Tcd Tsifs Tul Tsifs Tdl Tsifs Tul Tdl Tsifs Tul Tsifs Fig. 3 DL {5-8} UL5 UL6 UL7 UL8 DL {(N-3)-N} Proposd FA iwlan protocol UL(N-3) UL(N-2) UL(N-1) UL(N) 2, th MS will r-communicat with thm again. This r-communication duration is dnotd by T rcom. Th oprations during this stag ar th sam with stag 2. Stag 3 is rpatd until MS succssfully communicats with all N irbs or th T cd lapss. Whn th T cd lapss and thr is not any irb rror in this stag, a nw T cd is prformd. Th numbr of tims to prform stag 3 is dfind as th numbr of rtransmissions, dnotd by M. n m is th numbr of irb rrors occurring during th m th rtransmission (m is {1, 2,..., M }). n is th numbr of irb rrors occurring during stag 2. If thr is any irb rror aftr M rtransmissions, i.. aftr T cd lapss, th systm will b considrd to b in rror and stoppd immdiatly. Bcaus th transmission intrvals such as T sync, T dl, T ul and T cd nd to b strictly synchronizd btwn MS and irb to prsrv th opration of our iwlan protocol, th PTP tchniqu[14] is mployd to provid th sub-microscond clock synchronization. To mak th PTP compatibl with our protocol, Fig. 5 shows how th Sync, Follow Up, Dlay Rqust and Dlay Rspons mssags of PTP ar xchangd btwn MS and irb using SYNC, DL and UL frams of our protocol. For th i th synchronization priod, MS dos tim stamping to gt t i, t i 3 and irb dos tim stamping to gt t i 1, t i 2. irb rcivs t i from th DL and t i 3 from broadcastd SYNC fram of th nxt (i+1) th synchronization priod. Thn, th timing offst btwn MS and irb can b calculatd and adjustd at irb by offst = [(t i 1 t i ) (t i 3 t i 2)]/2. With th afor proposd opration, our iwlan protocol supports not only th dtrministic fatur but also th MU accss on both DL and UL transmissions for fast communications. 3. MU DL Transmission Tchniqus In th proposd iwlan transmission protocol in sction 2.2, w mploy th MU transmission tchniqu for DL accss to nhanc th throughput of th FA communication systm. Th FDMA and SDMA ar th convntional transmission tchniqus which support th MU DL accss. In th proposd iwlan protocol, bcaus th irbs rciv SYNC frams on channl bandwidth of 8 MHz, thy cannot rciv th DL frams on diffrnt channl bandwidth by using th convntional FDMA tchniqu. In this sction, w firstly prsnt th convntional MU-MIMO Stag1: SYNC Start nw T cd at MS SYNC fram transmission Stag2: COMMUNICATION at MS at irb Stag3: RECOMMUNICATION n m >? Ys m=m+1 at MS No n m-1 >? Ys SIFS count at irb Fig. 4 t i t i 3 t i+1 N >? Ys SIFS count No MU(u)-DL fram transmission N = N - u SIFS count UL fram transmit m= No MU(u)-DL fram transmission n m-1 = n m-1 - u SIFS count UL fram transmit Wait T cd timout End T cd Transmission squnc oprations of proposd FA iwlan protocol MS Clock Fig. 5 irb Clock offst t i 1 t i 2 t i+1 1 PTP frams xchang on iwlan protocol DL SDMA tchniqu, thn w show our proposd MU- DL tchniqu. 3.1 Convntional MU-MIMO SDMA DL Tchniqu For MU DL transmissions, th MU-MIMO SDMA is an mrging tchniqu in th 82.11ac WLAN systms[15] bcaus accss point (AP) can simultanously transmit indpndnt strams to MU on th sam channl by applying crosstalk intrfrnc (CTI) minimizing prcoding vctors[16]. Th MU-MIMO SDMA DL tchniqu is shown in Fig. 6. At AP, th Srvic Data Units (MSDUs) of four usrs, MSDU (1 4), ar prcodd into four transmission frams SDMA (1 4). Thn, ths four frams ar transmittd through four diffrnt antnna strams on th sam channl of 8 MHz. At ach usr, its MSDU is dcodd from th rcivd intrfrnc SDMA {1,2,3,4}. In ordr to hlp ach usr to dcod its data, th 31

32 Transactions Trans. of ISCIE, of ISCIE, Vol. Vol. 29, No. 29, No. 1 (216) 1 (216) MSDU 1 AP SDMA1 8 MHz channl Tx 1 SDMA {1,2,3,4} MSDU 1 Dcoding Usr1 MSDU 1 AP 8 MHz channl MSDU 1 Extraction Usr1 MSDU 2 MSDU 3 MU- MIMO SDMA Prcoding SDMA2 SDMA3 MSDU 4 SDMA4 Tx 2 Tx 3 Tx 4 Sounding procdur SDMA {1,2,3,4} SDMA {1,2,3,4} SDMA {1,2,3,4} MSDU 2 Dcoding MSDU 3 Dcoding MSDU 4 Dcoding Usr2 Usr3 Usr4 MSDU 2 MSDU3 MSDU4 MU- Forming Tx MSDU 2 Extraction MSDU 3 Extraction MSDU 4 Extraction Usr2 Usr3 Usr4 Fig. 6 MU-MIMO SDMA DL tchniqu Fig. 8 Proposd MU- DL tchniqu Sounding priod Sounding Procdur MU-MIMO transmissions Sounding Procdur MU-MIMO transmissions MU- DL fram format MU-MPDU Hadr Hadr MU-MSDU CRC SIFS SIFS SIFS SIFS SIFS (42.8+.26xN)us 4us 193us 43us 193us AP NDPA NDP BFRP Usr1 CB SIFS 43us BFRP SIFS 193us SFH1 MSDU1 SFH2 MSDU2 SFH3 MSDU3 SFH4 MSDU4 Usr2 Usr N Fig. 7 CB Sounding procdur in MU-MIMO SDMA DL tchniqu AP rquirs th channl stat information at transmittr (CSIT) for th pr-coding procss. To gt th CSIT, it taks a critical ovrhad duration to prform th sounding procdur bfor th data transmissions (MU-MIMO transmissions), as shown in Fig. 7. Th sounding priod includs th sounding procdur and th MU-MIMO transmissions. Th sounding procdur xchangs th mssags of th Null Data Packt Announcmnt (NDPA), Null Data Packt (NDP), Comprssd Bamforming (CB) and Bamforming Rport Poll (BFRP) btwn AP and usrs. Th CB duration (193 µsc) and BFRP duration (43 µsc) mak th sounding ovrhad of ach usr approximat 236 µsc[15]. Th MU-MIMO transmissions xchang th control data btwn AP and usrs. In WLAN systms with th MU-MIMO SDMA DL tchniqu, th data transmission frams ar long and th sounding priods ar also long, approximatly N 2 milisconds[17], thn th ovrhad of sounding procdur is only a small proportion of th sounding priod. Thrfor, this ovrhad is not critical for WLAN systms. Manwhil, in th FA communication systm, th data transmission frams ar short and th targt control duration, T cd, is also short, approximatly N 1 microsconds, thn th ovrhad of sounding procdur bcoms xtrmly critical. To mitigat th critical ovrhad for MU DL transmissions, w propos a low ovrhad MU- DL tchniqu. 3.2 Proposd MU- DL Tchniqu In ordr to support th MU DL accss with low ovrhad for FA communication systms, th MU- DL transmission tchniqu is proposd, as shown in Fig. 8. At accss point (AP), th MSDU data of four usrs, MSDU (1 4), ar mrgd into on CB Usr Addrss MSDU Lngth UL Channl 6 octts 1 octt 1 octt Fig. 9 Proposd MU- DL fram format transmission fram, calld MU-. Thn, this MU- fram is transmittd to all usrs through on antnna stram on th sam channl of 8 MHz. This MU- fram can also b transmittd to all usrs through four antnna strams to gt th divrsity gain. At ach usr, its MSDU is xtractd from th rcivd MU- fram. Th format of MU- is shown in Fig. 9. To mak th our FA systm compatibl with th WLAN systm, th format of MU- fram adhrs to th IEEE 82.11 standard, that is comprisd of th hadr, hadr, MU-MSDU and cyclic rdundancy chck (CRC) data. Th MU- Protocol Data Unit (MU-MPDU), which is comprisd of th hadr, MU-MSDU and CRC data, is constructd on layr. Thn, th MU- fram, which is comprisd of th hadr and MU-MPDU data, is constructd on layr. To dlimit th data of ach usr in th MU- MSDU, a Sub-Fram Hadr (SFH) is mbddd btwn ach MSDU. Th SFH includs: (1) Usr Addrss fild (6 octts) that indicats th dstination addrss of its following MSDU, (2) MSDU Lngth fild (1 octts) that indicats th lngth of that MSDU, (3) UL Channl fild (1 octts) that dfins th channl allocatd for UL transmission. At rcivr sid, th usr gts its MSDU basd on th Usr Addrss and MSDU lngth filds, thn it prforms th UL transmission on th channl indicatd by th UL Channl fild. Th UL fram of ach usr is transmittd on th channl of 2 MHz. Instad of using th critical ovrhad sounding procdur, th MU- DL tchniqu uss only th small ovrhad SFHs (8 octts) to dcod th data, thrfor in FA systms, this tchniqu provids highr throughput than th MU-MIMO DL tchniqu. 32

Lam, Shinozaki, Yamaguchi, Morita, Nagao, Kurosaki and Ochi: A Fast and Saf Industrial WLAN Communication Protocol 33 Tabl 1 Systm configuration paramtrs Paramtrs Conv.1 Conv.2 Prop. Protocol ipcf iwlan iwlan DL Accss Schm TDMA SDMA Fram Format VHT -mixd VHT -mixd VHT -mixd Protocol ipcf iwlan iwlan UL Accss Schm TDMA FDMA FDMA Fram Format Tabl 2 Paramtrs VHT -mixd Lgacy Lgacy Transmission paramtrs Valus Bandwith (MHz) 8 Modulation QPSK Coding Rat 1/2 Numbr of irbs (N) 4 MSDU Lngth/iRB (Octts) 32 SFH (Octts) 8 T sifs (µsc) 16 4. Prformanc Comparisons Undr Non-Error Basd Transmissions In ordr to undrstand th improvmnts of our approachs, in this sction w show th numrical prformanc comparisons btwn our proposd systm with two convntional systms undr non-rror basd transmissions, in which no packt rrors occur during th transmissions btwn irbs and MS, i.. M = or T rcom =. Th configuration paramtrs of thr systms ar shown in Tabl 1. Th proposd systm (Prop.) mploys th iwlan protocol with th proposd MU- DL accss schm. Th first convntional systm (Conv.1) mploys th convntional ipcf protocol with th convntional SU- TDMA accss schm. Th scond convntional systm (Conv.2) mploys th iwlan protocol with th convntional MU-MIMO SDMA DL accss schm. For fair comparisons, thr systms mploy th sam transmission paramtrs, which ar shown in Tabl 2. For saf transmissions, w mploy th Quadratur Phas Shift Kying (QPSK) modulation schm. To mak th our systm compatibl with th WLAN systm, th of transmission frams adhr to th IEEE 82.11 standard. Th fram of DL and UL transmissions of th Conv.1, Conv.2 and Prop. systms ar shown in Fig. 1, Fig. 11 and Fig. 12, rspctivly. For th transmission frams using 2 MHz bandwidth, and hadrs mploy th format of th Lgacy Physical Layr Convrgnc Protocol (PLCP) Protocol Data Unit (PPDU) DL fram UL fram Fig. 1 DL fram UL fram Fig. 11 DL fram UL fram Fig. 12 36 octts 32 octts 4 octts Hadr Hadr MSDU CRC 4 us 12 us 8MHz VHT-mixd format PPDU, QPSK modulation (52 us) 36 octts 32 octts 4 octts Hadr Hadr MSDU CRC 4 us 12 us 8MHz VHT-mixd format PPDU, QPSK modulation (52 us) DL and UL fram of Conv.1 systm 36 octts 32 octts 4 octts Hadr Hadr MSDU CRC 4 us 12 us 8MHz VHT-mixd format PPDU, QPSK modulation (52 us) 3 octts 32 octts 4 octts Hadr Hadr MSDU CRC 2 us 48 us 2MHz Lgacy format PPDU, QPSK modulation (68 us) DL and UL fram of Conv.2 systm 8MHz VHT-mixd format PPDU, QPSK modulation (68 us) 4 us 28 us 36 octts 16 octts 4 octts Hadr Hadr SFH1 8 octts MSDU1 32 octts MU-MSDU SFH4 8 octts 3 octts 32 octts 4 octts Hadr Hadr MSDU CRC 2 us 48 us 2 MHz Lgacy format PPDU, QPSK modulation (68 us) CRC MSDU4 32 octts DL and UL fram of Prop. systm fram[13]. For th transmission frams using 8 MHz bandwidth, and hadrs mploy th format of th Vry High Throughput (VHT) mixd PPDU fram[15]. Th MSDU lngth of ach irb quals th control data lngth of ach irb (32 octts). Th MU- MSDU of th MU- DL fram of th Prop. systm is comprisd of th data (SFH + MSDU) of four irbs, thn th lngth of MU-MSDU is 16 octts. To obtain th CSIT information th sounding procdur, which is shown in Fig. 7, nds to b prformd on th Conv.2 systm. From aformntiond configuration paramtrs and transmission paramtrs, th numrical valus of T dl and T ul for ach systm ar calculatd and shown in Tabl 3. Th sounding ovrhad of ach irb, T sounding, is obtaind from sction 3.1. For th Conv.1 systm, th T cd of 4 irbs, which quals 4 (T dl +T sifs +T ul + T sifs ), is 544 µsc. For th Conv.2 systm, th T cd of 4 irbs, which quals (T dl + T sifs + T ul + T sifs )+4 T sounding, is 196 µsc. For th Prop. systm, th T cd of 4 irbs, which quals (T dl +T sifs +T ul +T sifs ), is 168 µsc. From ths rsults, th valus of control duration pr irb (T irb ), which ar calculatd by T cd /4, of th Conv.1, Conv.2 and Prop. systms ar 136, 247 and 42 µsc, rspctivly. Ths rsults indicat that th throughput of th Prop. systm is fastr than that of th Conv.1 systm. Th small transmission ovrhad is th rason for this improvmnt. In FA communication systms, 33

34 Transactions Trans. of ISCIE, of ISCIE, Vol. Vol. 29, No. 29, No. 1 (216) 1 (216) Tabl 3 Prformanc comparisons undr non-rror basd transmissions Paramtrs Conv.1 Conv.2 Prop. T dl (µsc) 52 52 68 T ul (µsc) 52 68 68 MU-DL support no ys ys MU-UL support no ys ys T sounding pr irb (µsc) 236 T irb (µsc) 136 274 42 sinc MS and irbs mploy only short MSDU data to transmit to ach othr, th transmission ovrhad which is causd by and hadrs (T hadrs ) bcoms a larg proportion of th transmission PPDU packt. Our iwlan protocol supports MU accss of 4 irbs for both DL and UL, thrfor th ovrhad of ach irb quals T hadrs /4. Manwhil th ipcf protocol supports only SU accss, thus th ovrhad of ach irb quals T hadrs. Ths rsults also indicat that th throughput of th Prop. systm is fastr than that of th Conv.2 systm. Th larg sounding ovrhad, T sounding, maks th throughput of th Conv.2 systm slowr than that of Prop. systm. 5. Prformanc Formulation Undr Error Basd Transmissions In our iwlan protocol, th T cd includs th durations of T sync, T com and T rcom. T rcom is th duration of M r-communications. If th valu of M is larg, th communication systm is saf but th valu of T cd is big, that maks its throughput slow. Thrfor, in this sction th iwlan transmission protocol is analyzd to find th bst valu of M to satisfy both fast and saf rquirmnts undr rror basd transmissions, in which th packt rrors may occur during th transmissions btwn irbs and MS. To analyz th iwlan transmission protocol, th following dfinitions ar usd, irb rror: occurs if th irb fails to ithr rciv th DL packt from MS or transmit th UL packt to MS. Systm rror: occurs if thr ar still irb rrors in th systm aftr M r-communications ar alrady don, i.. aftr T cd lapss. P M : th probability that irb rrors occur aftr M r-communications ar alrady don. P cd : th probability that a systm rror occurs aftr T cd lapss. Thus, P cd = P M (1) Ps cd : th probability that no systm rrors occur aftr T cd lapss. Thn, Ps cd =1 P cd (2) T sd : th working duration that is dfind by T sd = k T cd (3) Ps sd : th probability that no systm rrors occur aftr T sd lapss. Thn, Ps sd = (1 P cd ) k (4) Sinc P cd 1, by using th Taylor-Maclaurin sris, (4) is simplifid to Ps sd =1 k P cd (5) P sd : th probability that th systm rrors occur aftr T sd lapss. Thn, P sd =1 Ps sd (6) From (5) and (6), w obtain, P sd = k P cd (7) From (3) and (7), w obtain, T sd P sd = T cd P cd Thus, w hav P sd (8) = P cd T sd (9) T cd Th iwlan systm is considrd safty during th T sd if th valu of P sd is lowr than an xpctd thrshold P th, i.. P cd T sd P th (1) T cd From (1) and (1), w obtain, P M T sd T cd P th (11) Now, w will find th valu of M to satisfy (11). In ordr to find th valus of P M and T cd, w considr th communications btwn MS and n t irbs. Th PER dl and PER ul ar dfind as th packt rror rat of DL and UL transmissions, rspctivly. Thn, th probability that an irb rror occurs is obtaind by PER dl +(1 PER dl ) PER ul. Th n dl is dfind as th numbr of th irbs to which MS transmits th DL frams, thus w hav n dl = n t. Th n d is dfind as th numbr of th irbs which fail to rciv th DL frams from MS. Th probability that n d irb rrors occur whn MS transmits th DL frams to n t irbs is thus Pn t,n d = n t Cn d (PER dl ) n d (1 PER dl ) n t n d (12) whr n C k is th k-combination from a st of n lmnts. Th n ds is dfind as th numbr of th irbs which succssfully rciv th DL frams, thus n ds = 34

Lam, Shinozaki, Yamaguchi, Morita, Nagao, Kurosaki and Ochi: A Fast and Saf Industrial WLAN Communication Protocol 35 COM (m=) 1 st RE-COM (m=1) 2 nd RE-COM (m=2) M th RE-COM (m=m) Com PN,N N irbs N irb rrors N irb rrors N irb rrors in TN in TN in TN in TN PN,(N-1) (N-1) irb rrors in TN-1 PN,1 P(N-1),(N-1) (N-1) irb rrors in TN-1 P(N-1),1 (N-1) irb rrors in TN-1 T m rcom = N n n =1n 1 =1... n m 2 n m 1 =1 [P N,n Pn,n 1... Pn m 2,n m 1 (Tn +Tn 1 +...+Tn m 1 )]/P m 1 (16) Fig. 13 PN, 1 irb rror in T1 irb rror P1,1 P1, 1 irb rror in T1 P(N-1), irb rror 1 irb rror in T1 irb rror Communication chain with transition probabilitis of iwlan systm whr Tn m is th duration which MS r-communicats with n irb rrors at th m-th r-communication. Th valus of Tn m ar pr-dfind basd on th valu of n and th transmission paramtrs in Tabl 1 and Tabl 2. Thn, th T cd aftr m r-communications, dnotd by Tcd m, is obtaind by n t - n d. Th n ul is dfind as th numbr of th irbs which transmit th UL frams to MS. Sinc only th n ds irbs can transmit th UL frams to MS, thn n ul = n t - n d. Th n u is dfind as th numbr of th irbs which fail to transmit th UL frams to MS. Th probability that n u irb rrors occur whn n ul irbs transmit th UL frams to MS is thus Pn ul,n u = n ul Cn u (PER ul ) n u (1 PER ul ) n t n (13) Th n is dfind as th numbr of th irb rrors occurring whn MS communicats with n t irbs, thn n = n d + n u. Th probability that n rrors occur whn MS communicats with n t irbs is givn by n Pn t,n = Pn t,n d Pn ul,n u (14) n d = Our communication protocol with on MS and N irbs can b xprssd by th communication chain shown in Fig. 13. Each nod prsnts for th communications or r-communications btwn MS and n irbs (n = 1,..., N ), and th duration of ths communications or r-communications,.g. Nod [ N irb rrors in T N,1 st RE-COM (m=1)] prsnts for th first r-communications btwn MS and N irb rrors, which occur in prvious nod (m=), and th duration of ths r-communications, T N. Th transition path from a sourc nod to a dstination nod is lablld by Pn t,n, that prsnts for th probability that n irb rrors occur whn MS communicats with n t irbs in th sourc nod. MS r-communicats with ths n irb rrors in th dstination nod. From th communication chain, th probability that th irb rrors occur aftr (m-1) r-communications, dnotd by P m 1, is calculatd by N n P m 1 = n =1n 1 =1... n m 2 n m 1 =1 P N,n Pn,n 1... Pn m 2,n m 1 (15) Thn, th duration of th total m r-communications, dnotd by T m rcom, is givn by T m cd = T sync +T sifs +T com +T m rcom (17) Aftr m r-communications, th valus of P m and Tcd m ar obtaind. Th bst valu of m is found by running th following algorithm, 1) Initializ m = 1 and dfin th xpctd valus of T sd and P th. 2) Calculat P m and Tcd m by (15) and (17), rspctivly. 3) Rplac P M = P m and T cd = Tcd m, thn chck th condition (11), if it is satisfid, th final valu of m is approachd, othrwis m=m+1and go back to 2. Aftr th final valu of m is approachd, th valus of M and T cd of systm ar st by th valus of m and Tcd m, rspctivly. Th valu of th control duration pr irb can b drivd by T irb = T cd /N (18) 6. Prformanc Evaluation Undr Error Basd Transmissions In this sction, th communication prformancs undr rror basd transmissions ar valuatd. Firstly, w show th numrical formulation rsults of th T irb for our proposd systm. Thn, by using th Matlab simulations w show th prformanc comparison rsults of th proposd systm with two convntional systms. For ths valuations, in addition mploying th paramtrs in Tabl 1 and Tabl 2, w also add othr transmission paramtrs, which ar shown in Tabl 4. Th FA working nvironmnt is modlld by th TGac channl D[18], which indicats th indoor nvironmnt. Th Signal-to-Nois Ratio (SNR) is varid from 16 db to 22 db. Th numbr of irbs in th systm is varid from 4 to 32. For fair comparisons, th transmission(tx)/rcption() antnnas of MS and irb ar 4/4 and 1/1, rspctivly, for thr systms. Th valu of T sync is usd for th iwlan protocol. Th valus of Bacon transmission duration (T bacon ), CF End transmission duration (T cfnd ) and T pifs ar usd for th ipcf protocol. 35

36 Transactions Trans. of ISCIE, of ISCIE, Vol. Vol. 29, No. 29, No. 1 (216) 1 (216) Tabl 4 Additional paramtrs for th prformanc valuation undr rror basd transmissions Paramtrs Valus Packt Error Rat (PER) TGac channl modl D (Indoor nvironmnt) SNR (db) 16-22 Tx/ Antnnas of MS 4/4 Tx/ Antnnas of irb 1/1 T sync (µsc) 56 T bacon (µsc) 56 T cfnd (µsc) 52 T pifs (µsc) 25 Numbr of irbs (N) 4-32 T sd P th 1 1-1 1-2 1-3 1 yar 1, 1 1, 1 2, 1 3 PER of UL transmission PER of DL transmission Control Duration pr irb [usc] 6 5 4 3 2 1 SNR16 SNR17 SNR18 SNR2 SNR22 Expctd Control Duration pr irb is 1 microsconds at N > 32 4 8 12 16 2 24 28 32 Numbr of irbs (N) Fig. 15 T irb numrical rsults of th proposd systm at P th = 1 and diffrnt SNR valus Control Duration pr irb [usc] 6 5 4 3 2 1 P thrshold =.1 P thrshold =.1 P thrshold =.1 P thrshold = 1 1-4 16 17 18 19 2 21 22 SNR [db] Fig. 14 Packt Error Rat (PER) simulation rsults for UL and DL of th proposd systm 6.1 Prformanc Numrical Rsults For th numrical valuation, T sd is dfind to b 1 yar and P th is dfind to b 1, 1 1, 1 2 and 1 3. Th paramtr T irb is masurd. In ordr to prform th numrical valuation, which is prsntd in Sction 5., th valus of PER ul and PER dl of th proposd systm nd to b obtaind. By using th layr simulator with th paramtrs in th third column of Tabl 1 and in Tabl 2, th valus of PER ul and PER dl ar obtaind as Fig. 14. W can s that, for high SNR th PER ul is bttr than PER dl bcaus th UL fram is shortr than th MU- DL fram. Howvr, for th low SNR th PER dl is bttr than PER ul bcaus of th divrsity gain from multipl antnnas at MS sid. Aftr th valus of T sync, T dl, T ul, PER ul and PER dl ar obtaind, th valu of T irb of th proposd systm can b found by running th algorithm in Sction 5.. Fig. 15 shows th T irb rsults at P th = 1 and diffrnt SNR valus. Th rsults show that if th SNR is ovr 17 db, with N biggr than 32, th 4 8 12 16 2 24 28 32 Numbr of irbs (N) Fig. 16 T irb numrical rsults of th proposd systm at SNR = 17 db and diffrnt P th valus T irb is fastr than 1 µsc. Th prformanc is incrasd with th incras of N bcaus th valu of M grows vry slightly whn N is incrasd. Fig. 16 shows th T irb rsults at SNR = 17 db and diffrnt P th valus. Th rsults show that th T irb is fastr than 1 µsc for th FA systm with N mor than 32 and P th of 1 3. Ths rsults indicat our proposd systm can achiv th fast and saf rquirmnts. 6.2 Prformanc Simulation Rsults For th simulation valuations, w build th simulators of thr systms on Matlab platform. Th simulators includs th and layrs. Th transmission protocols ar controlld by layrs and th transmission tchniqus ar controlld by layrs. In ach itration of simulations, MS communicats with N irbs. If thr ar any irb rrors, MS will r-communicat to thos irbs. Th valu of T cd is obtaind whn MS communicats succssfully to all irbs, thn th valu of T irb is calculatd by (18). In ordr to masur prformanc prcisly, w hav dvlopd a bit-basd layr simulator, in 36

Lam, Shinozaki, Yamaguchi, Morita, Nagao, Kurosaki and Ochi: A Fast and Saf Industrial WLAN Communication Protocol 37 Infor Fram Infor Channl Infor Control Duration pr irb [usc] 6 5 4 3 2 Simulator: PER Evaluation PER Fig. 17 Transmission Indication Array Gnration Random Transmission Error Indication 1: Error : Succssful Simulator: Prformanc Evaluation Abstract simulation modl Prformanc Valus Convntional systm 1 - Bit-basd modl Convntional systm 1 - Abstractd-basd modl Convntional systm 2 - Bit-basd modl Convntional systm 2 - Abstractd-basd modl Proposd systm - Bit-basd modl Proposd systm - Abstractd-basd modl Control Duration pr irb [usc] 6 5 4 3 2 1 Convntional systm 1 - Simulation rsults Convntional systm 2 - Simulation rsults Proposd systm - Simulation rsults Proposd systm - Numrical rsults 4 8 12 16 2 24 28 32 Numbr of irbs (N) Fig. 19 T irb simulation rsults at SNR = 17 1 17 18 19 2 21 22 SNR [db] Fig. 18 T irb simulation comparison btwn Bit-basd and Abstractd-basd modls at N = 32 which th PER of fram transmission is basd on th bit rror rat (BER) of ach transmittd fram[19]. Howvr, this simulation is so xtrmly tim-consuming that it is not suitabl for valuating a systm during a long opration obsrvation, i.. a larg numbr of simulation itrations. Thrfor, an abstract- modl is dvlopd to b abl to provid th accurat prformanc for such long obsrvation tim. Th abstract modl is shown in Fig. 17. First, th PER ul and PER dl of thr systms ar masurd by simulators. Thn, basing on ths valus of PER, th transmission indication array, in which 1 indicats an rror transmission and indicats a succssful transmission, is gnratd. For xampl, an array including 3 lmnts of 1 and 997 lmnts of will b gnratd if th PER is 3 1 3. Finally, prformanc of th systm can b valuatd by th simulator, in which DL and UL transmission statuss ar indicatd by randomly picking an lmnt from th transmission indication array that is rprsntd for th random transmissions. Fig. 18 shows th simulation rsults by using th bit-basd and th abstractd-basd modls with th itration numbr of 3. It indicats that th simulation rsults btwn two modls ar slightly diffrnt for all thr systms, howvr th simulation tim using th abstractd-basd modl is fastr by 14 tims than that of using th bit-basd on. Thus, th abstractd-basd modl is mployd for our simulations to valuat th prformanc of thr systms with th itration numbr of 3,,, which is considrd as a long obsrvation tim. Fig. 19 and Fig. 2 show th T irb comparison Control Duration pr irb [usc] 6 5 4 3 2 1 Convntional systm 1 - Simulation rsults Convntional systm 2 - Simulation rsults Proposd systm - Simulation rsults Proposd systm - Numrical rsults 17 18 19 2 21 22 SNR [db] Fig. 2 T irb simulation rsults at N = 32 simulation rsults among thr systms. W can s that th Prop. systm with MU accss achivs 1% fastr throughput than th Conv.1 systm with only SU accss. Our systm with th low ovrhad MU- DL schm also achivs 3% fastr throughput than th Conv.2 systm with th critical ovrhad MU-MIMO DL schm. Th rsults also indicat that th simulation and numrical rsults provid th sam prformanc undr th sam duration of obsrvation. Fig. 21 shows th systm rror rat (SER) simulation rsults of thr systms with 32 irbs. Th SER is masurd by th numbr of systm rrors ovr th total numbr of itrations of simulation. For fair SER comparisons, w ignor th sounding procdur in Conv.2 systm and w dfin th valus of T cd of thr systms ar th sam and qual N 136 µsc. Th valu of 136 µsc is th T irb of Conv.1 systm undr non-rror basd transmissions. Th rsults show that th SER of th Prop. systm is lowr than that of th Conv. systms,.g. at SNR of 14 db, th SER of Prop. systm is 3 1 2, manwhil th SER valus of th Conv. systm 1 and Conv. systm 2 ar 4 1 1 and 1, rspctivly. From ths rsults, 37

38 Transactions Trans. of ISCIE, of ISCIE, Vol. Vol. 29, No. 29, No. 1 (216) 1 (216) Systm Error Rat (SER) Probability 1 1-1 Convntional systm 1 Convntional systm 2 Proposd systm 1-2 12 13 14 15 SNR [db] 1 1-1 1-2 1-3 1-4 1-5 Fig. 21 Systm Error Rat (SER) at N = 32 SNR=17 [db] SNR=18 [db] SNR=19 [db] SNR=2 [db] SNR=21 [db] SNR=22 [db] 1-6 1 2 3 4 5 6 Numbr of R-transmission Fig. 22 Th probability of rtransmission numbr (M ) of th proposd systm at diffrnt SNR valus w can stat that th our proposals can provid mor safty than th convntional ons. Fig. 22 shows th probabilitis of th valus of M of th proposd systm. Th valu of M is obtaind whn MS communicats succssfully to all irbs. Whn SNR is 17 db, th systm nds to rtransmit up to 5 tims to mak all communications succssful. This numbr is rducd to 3 if th SNR is gratr than 19 db. This valu is usd to configur th rtransmission numbr to improv th throughput of th systm whn th SNR is alrady known. 7. Conclusion A vry high throughput FA wirlss communication systm for robot controls has bn prsntd in this papr. W hav proposd a synchronous MU round-robin iwlan transmission protocol. This protocol is applicabl for FA communications bcaus it supports th dtrministic fatur. Our protocol achivs 1% highr prformanc than th convntional ipcf protocol bcaus it can support th MU transmissions for both UL and DL. W hav also proposd th low ovrhad transmission tchniqu for th MU-DL accss. In FA transmissions, this tchniqu has much lss ovrhad than th MU- MIMO DL tchniqu in SDMA systm. Indicatd by th simulation rsults, our systm achivs 3% highr throughput than th convntional SDMA systm undr th sam working constraints. Bsids, our approachs also provid lowr systm rror rat than convntional ons. In particular, th control duration pr irb in th iwlan systm is fastr than 1 µsc. Thrfor, th proposd schms can achiv fast and saf prformanc for FA communication systms. Acknowldgmnts This work was supportd by a rsarch grant from th Adaptabl & Samlss Tchnology Transfr Program through Targt-drivn R&D (A-STEP) of Japan Scinc and Tchnology Agncy. Rfrncs [1] A. Willig:Rcnt and mrging topics in wirlss industrial communications: A slction;ieee Transactions on Industrial Informatics, pp. 12 124 (28) [2] L. Lo Bllo and O. Mirablla:Communication tchniqus and architcturs for Blutooth ntworks in industrial scnarios;ieee Confrnc on Emrging Tchnologis & Factory Automation (ETFA), pp. 52 61 (25) [3] D. Egan:Th mrgnc of ZigB in building automation and industrial control;computing and Control Enginring Journal, Vol. 16, No. 2, pp. 14 19 (25) [4] N. Bakr:ZigB and Blutooth strngths and waknsss for industrial applications;computing and Control Enginring Journal, Vol. 16, No. 2, pp. 2 25 (25) [5] J. S. L, Y. W. Su and C. C. Shn:A comparativ study of wirlss protocols: Blutooth, UWB, ZigB, and Wi-Fi;Annual Confrnc of th IEEE on Industrial Elctronics Socity (IECON), pp. 46 51 (27) [6] HART Communication Foundation:HART Fild Communication Protocol Spcification, Rvision 7. (27) [7] S. Ptrsn and S. Carlsn:Prformanc valuation of WirlssHART for factory automation;ieee Confrnc on Emrging Tchnologis & Factory Automation (ETFA), pp. 1 9 (29) [8] I. Dominguz-Jaims, L. Wisniwski, H. Trsk and J. Jasprnit:Link-layr rtransmissions in IEEE 82.11g basd industrial ntworks;ieee Intrnational Workshop on Factory Communication Systms (WFCS), pp. 83 86 (21) [9] G. Cna, L. Sno, A. Valnzano and C. Zunino:On th prformanc of IEEE 82.11 wirlss infrastructurs for soft-ral-tim industrial applications;ieee Transactions on Industrial Informatics, Vol. 6, No. 3, pp. 425 437 (21) [1] G. Santandra:A PROFINET IO application implmntd on Wirlss LAN;IEEE Intrnational Workshop on Factory Communication Systms (WFCS) 38

Lam, Shinozaki, Yamaguchi, Morita, Nagao, Kurosaki and Ochi: A Fast and Saf Industrial WLAN Communication Protocol 39 (26) [11] I. A. Candian:PROFIsaf and PROFINET via IWLAN;Simns, Automation & Drivs PICC Mting (28) [12] Simns Corporation:Undrstanding IWLAN vn allowing for Safty;PROFIBUS-PROFINET Usr Confrnc (21) [13] IEEE:IEEE Std 82.11-212, Part 11: Wirlss LAN Mdium Accss Control () and Physical Layr () Spcifications (212) [14] IEEE:IEEE Standard for a Prcision Clock Synchronization Protocol for Ntworkd Masurmnt and Control Systms, IEEE Standard (28) [15] IEEE:IEEE Std 82.11ac/D7., Part 11: Wirlss LAN Mdium Accss Control () and Physical Layr () Spcifications (213) [16] E. H. Ong, J. Knckt, O. Alann, Z. Chang, T. Houvinn and T. Nihtila: IEEE 82.11ac: Enhancmnts for vry high throughput WLANs; IEEE Intrnational Symposium on Prsonal Indoor and Mobil Radio Communications (PIMRC), pp. 849 853 (211) [17] G. Rditab, L. Cariou and P.Christin: + channl sounding intrval analysis for IEEE 82.11ac MU-MIMO; IEEE Intrnational Symposium on Wirlss Communication Systms (ISWCS), pp. 154 158 (212) [18] G. Brit, H. Sampath, S. Vrmani, t al.: TGac channl modl addndum; IEEE 82.11-9/38r12 (21) [19] D. K. Lam, Y. Shinozaki, K. Yamaguchi, S. Morita, Y. Nagao, M. Kurosaki and H. Ochi:A fast and safty industrial WLAN protocol for factory communication systms;ieee Intrnational Confrnc on Industrial Tchnology (ICIT), pp. 188 1813 (215) Authors Duc Khai Lam (Studnt Mmbr) Rcivd B.E. (26) and M.S. (211) from Univrsity of Scinc, Vitnam. H is currntly pursuing his Ph.D. studnt at Kyushu Institut of Tchnology. His rsarch intrst is th industrial wirlss LAN systm. H is a studnt mmbr of th IEEE. Yasuhiro Shinozaki Rcivd B.E. (213), M.E. (215) from Kyushu Institut of Tchnology (Kyutch). H thn joind NTT Communication Corporation. His rsarch intrsts ar industrial wirlss LAN systm. H is a studnt mmbr of th IEICE. Kishi Yamaguchi Rcivd B.E. (214) from Kyushu Institut of Tchnology (Kyutch). H is currntly pursuing his M.E. studnt at Kyutch. His rsarch intrsts ar industrial wirlss LAN systm. H is a studnt mmbr of th IEICE. Satoshi Morita Rcivd B.E. (213), M.E. (215) from Kyushu Institut of Tchnology (Kyutch). H thn joind KDDI Corporation. His rsarch intrsts ar industrial wirlss LAN systm. H is a studnt mmbr of th IEICE. Yuhi Nagao Rcivd M.I. (26) and Ph.D. (29) from Kyushu Institut of Tchnology. Sinc 29 h has bn with Kyushu Institut of Tchnology as Rsarchr. His rsarch intrst is th wirlss communication. H is a mmbr of th IEEE. Masayuki Kurosaki Rcivd B.E. (2), M.E. (22) and Ph.D. (25) from Tokyo Mtropolitan Univrsity. H was with Kyushu Institut of Tchnology (KIT) from 25 to 211 as an Assistant Profssor. Sinc 211, h has bn with Kyushu Institut of Tchnology (KIT) as an Associat Profssor. His rsarch intrsts ar imag procssing and wirlss communication for multimdia. H is a mmbr of th IEEE. Hiroshi Ochi Rcivd B.S. (1981), M.S. (1984) Ph.D. (1991) all in lctronics nginring. H was with Univrsity of th Ryukyus from 1986 to 1999 as Associat profssor. From 1999, h is with Kyushu Institut of Tchnology (Kyutch) as a profssor in computr and lctronics nginring dpartmnt. H got MBA (27) dgr from Kyushu Univrsity. H also organizs Radrix co.ltd as a CTO. (25). His rsarch intrsts ar signal procssing, wirlss communication, VLSI chip dsign and MOT ducation, tc,. H is a mmbr of th IEEE. 39