USING INERTIAL MEASUREMENT TO SENSE CRASH-TEST DUMMY KINEMATICS

Transcription

1 USING INERTIAL MEASUREMENT TO SENSE CRASH-TEST DUMMY KINEMATICS Sangam Rdka, Azona Stat Unvsty; Tom Suga, Azona Stat Unvsty; Anshuman Razdan, Azona Stat Unvsty; Ujwal Konu, Azona Stat Unvsty; Bll Dllad, Achangl Systms; Kathk Naayanan, Achangl Systms Abstact In ths study, th authos psnt a novl fuzzy-logc sgnal-pocssng and snso-fuson algothm wth quatnon mplmntaton to comput dummy knmatc paamts n a vhcl cash vnt usng ntal snsng. Ths algothm s calld Quatnon Fuzzy Logc Adaptv Sgnal Pocssng fo Bomchancs (QFLASP-B). Ths algothm s ffcnt and uss 3 ats obtand usng gyoscops and 3 acclatons obtand usng acclomts (on gyo and acclomt pa p axs) to comput knmatc paamts n a cash vnt. In ths study, ths QFLASP-B algothm was valdatd usng MSC-ADAMS and Lf-Mod smulaton softwa. In vtual smulatons of cash tstng, th poblm of fowad knmatcs was solvd usng MSC-ADAMS and Lf-Mod to obtan body acclatons and body angula vlocts. Th nvs knmatc poblm of computng ntal soluton usng body ats and acclatons was solvd usng QFLASP-B. Th sults of ths two analyss w thn compad. Th sults vald clos smlats. In th xpmntal valdaton, th soluton obtand fom th Nn Acclomt Packag (NAP) was compad wth th soluton obtand fom th th gyos and th acclomts, o Intal Masumnt Unt (IMU), usng th QFLASP-B algothm fo had ontaton computaton. Ths sults w also closly algnd. Th QFLASP-B algothm s computatonally ffcnt and vsatl. It s capabl of vy hgh data ats nablng al-tm soluton computaton and knmatc paamts dtmnaton. Th adaptv fltng n QFLASP-B nabls ngns to us low-cost MEMS gyoscops and acclomts, about $30 ach, whch a typcally nosy and show sgnfcant tmpatu dpndnc and bas dft both shot-tm and long-tm to obtan manngful and accuat sults wth th supo sgnal-pocssng and snso-fuson algothm. It s antcpatd that ths ntal tackng/snsng appoach wll povd an nxpnsv altnatv fo ngns ntstd n masung knmatc paamts n a cash vnt. Intoducton Moto vhcl accdnts sult n mo than 40,000 fatalts and th mllon njus ach ya n th Untd Stats. To ncas occupant safty, t s mpotant to study vhcl occupant knmatcs and th mchansms that gnat th focs that nju vhcl occupants dung cashs. Rsachs hav studd ths poblm fom thotcal and pactcal aspcts [1], [2]. Cash tstng s outnly cad out to valuat cashwothnss. Cash tstng of dumms (Hybd-II o III) and th knmatcs plays a sgnfcant ol n undstandng occupant/pdstan moton n cashs. Th a vaous tchnqus cuntly usd to cod and undstand vhcl occupant o vhcl-pdstan ntacton. It s ctcal to know th postons and ontatons of vaous body sgmnts of a cash-tst dummy n a typcal cash vnt. Ths data s utlzd to undstand njuy mchansm, svty of njuy, ffctvnss of satblts o abags n od to dtmn th ovall safty atng of th vhcl. Th a vaous snsng tchnqus usd to tack th moton of a dummy n th cash scnao. Som of th wdly usd tchnqus fo moton captu and snsng a hgh-spd vdo, acclomty Nn Acclomt Packag (NAP) and ntal snsos [1]-[4]. Smla tchnqus a also usd n Augmntd Ralty (AR) and Vtual Ralty (VR) applcatons [5]-[6] fo moton snsng. Th objctv of ths study was to psnt a novl softwa sgnal-pocssng algothm and ts applcatons to comput dummy knmatc paamts usng ntal snsng. Cuntly, th snsos usd fo cash tstng (lk ATA at-snsos gyoscops and Endvco acclomts) a vy xpnsv, bulky and hav stct pow-condtonng and mountng qumnts. In oth wods, usng a snso sut mad up of cuntly-avalabl hadwa s not only xpnsv but also tm consumng. Fotunatly, cnt advancs n Mco Elcto-Mchancal Systms (MEMS) tchnology hav bought sold-stat, ntgatd low-cost MEMS acclomts and gyoscops to makt, whch can thotcally b usd fo snsng applcatons. Ths MEMS snsos, dspt th low cost, suff fom dft, scal-facto nonlnaty, nos and coss-axs os [7]. It s almost mpossbl to us ths snsos dctly fo cash tstng applcatons. In ths study, th authos psnt a smat Fuzzy Logc Adaptv Sgnal Pocssng (FLASP) algothm that would nabl ngns to us ths low-cost MEMs snsos fo accuat ntal snsng of knmatc paamts. Som of th commcally-avalabl IMUs a shown n Fgu 1 [8]-[10]. In ths study, Achangl's IMU, known as USING INERTIAL MEASUREMENT TO SENSE CRASH TEST DUMMY KINEMATICS 17

2 IM3 (Intal Masumnt Cub), was usd to mplmnt th QFLASP-B algothm. IM3 s a sx-axs Intal Masumnt Unt (IMU) systm n a sngl ¾ cub. Ths cub masus and thmally compnsats acclatons n 3 othogonal axs (local X, Y and Z) and otatonal vlocts n th othogonal axs (about local X, Y and X) and computs ontatons, postons and vlocts va an onboad DSP. mnmz systmc os. Typcally, fo IMUs, sgnalpocssng algothms basd on Kalman fltng a usd [11]-[13] fo snso fuson. Unfotunatly, snso-fuson algothms usng Kalman fltng nvolv numous matx nvsons and cannot b mplmntd on a low-cost DSP platfom whn hgh-updat ats (100 Hz o mo) a ndd. Th QFLASP-B algothm poposd h dos not nvolv any matx nvsons and can b mplmntd on a low-cost DSP wth computng solutons at fquncs of 100Hz o mo. Ths algothm s psntd n scton 2, followd by ts mbddd mplmntaton, smulaton, and tstng sults n scton 3. Fnally, scton 4 summazs th wok. An Algothm fo Intal Snsng a) Achangl IM3 b) Spakfun IMU c) Analog Dvcs IMU Fgu 1. Intal Masumnt Unts Th algothm fo ntal snsng (QFLASP-B) psntd h uns on a low-cost DSP such as a DsPIC usd by IM3, uss nosy masumnts fom MEMs snsos and poducs qually accuat solutons obtand by hgh-cost pcson snsos. Ths algothm mplmnts snso-o modls that Th motvaton fo QFLASP can b xpland by consdng atttud-stmaton poblms wth stap-down snsos, usually cast as a two-vcto Wahba poblm [15]. Gvn masumnts of two non-co-lna vctos n a fxd-body fam, and wth knowldg of th vctos n a fnc fam, Wahba poposd an stmat of atttud by ducng th o btwn th fnc vcto st and th otatd vcto st fom th fxd-body fam. Th us of quatnon psntaton lmnats th sngulaty ssus assocatd wth Eul-angl psntatons [16], [17]. Th algothm uss masumnts fom a fxd-body tad of gyos and acclomts. b Th stmaton poblm qus two fams a fxdbody fam and a non-otatng ntal-fnc fam. Lt a and b psnt th fnc fam and th fxd-body fam, spctvly. Th atttud can b psntd by a squnc of th ght-handd otatons fom th fnc fam to th body fam. If ψ, θ and φ psnt th otatons about th z axs and th ntmdat y and x axs, spctvly, a vcto u n th fnc fam can b a tansfomd to a vcto u wth th otaton matx C b/ a. b u = Ca/ b u (1) Th otaton matx s gvn by cθ cψ cθ sψ sθ C a/ b = cφ sψ sφ sθ cψ cφ cψ sφ sθ sψ sφcθ + + (2) sφ sψ + cφ sθ cψ sφcψ + cφ sθ sψ cφ cθ wh c and s = cos and sn, spctvly. Fo a otaton squnc, Eul-angl psntaton π has a sngulaty atθ =, wh th oll and yaw angls a 2 undfnd. An altnat psntaton of atttud s wth a a 18 INTERNATIONAL JOURNAL OF MODERN ENGINEERING VOLUME 10, NUMBER 2, SPRING/SUMMER 2010

3 quatnon. Quatnons a gnalzatons of complx numbs n th dmnsons, and a psntd by [ ] T q0 q q = (3) wh q 0 and q a th al pat and th vcto pat of th quatnon. If th nom of th quatnon s unty, t s fd to as a unt quatnon. Just lk th otaton matx, a unt quatnon (o any quatnon n gnal) can b usd to otat a vcto fom a fnc fam to a fxd-body fam. Th otaton quaton n tms of quatnon s xpssd as b a u = q u q q q0 = q wh q s th nvs of th unt quatnon q and q s th unt quatnon that otats a vcto fom systm a to systm b [17]. Th atttud quatnon can also b psntd as a poduct of componnt quatnons n th axs. q [ cos( / 2) sn( / 2) 0 0] T φ = φ φ (5) [ ] (4) q cos( / 2) 0 sn( / 2) 0 T θ = θ θ (6) [ ] q cos( / 2) 0 0 sn( / 2) T ψ = ψ ψ (7) Th quatnon q can b dvd fom th componnt quatnons as q = qψ qθ q φ (8) Posson s knmatc quaton n quatnon fom that lats th at of chang of th atttud quatnon to th angula at of th body fam wth spct to ntal fam [17] s gvn by q& = 0.5 q ω (9) b/ a b/ a b/ a If q 1 and q 2 a two quatnons, th latv ontaton o th o quatnon btwn th two s q = q1 q 2 (10) Th two fams, whos atttud quatnons wth spct to a fnc fam, a q 1 and q2 and concd only f δ q = 1 and wh δ q and δ q = 0 (11) δ q a th al and vcto pats of th quatnon o. δ q = 0 s a suffcnt condton fo th two fams to concd. Algothm Dscpton Th algothm psntd h uss masumnts fom th axs gyos and acclomts. If ω T s th tu angula at and ω s th masud angula at, thn ω = ω + ε + ε + η (12) T B wh ε B s a tm-vayng bas, η s nos, and ε s oth os. If gyo bas can b captud wth an actv bas stmaton schm, thn th stmat of th tu angula at s gvn by ω = ω ε ε (13) B wh ε B s th cunt stmat of th gyo bas and ε ω s an angula at cocton dvd fom th atttud o. Th stmat of angula at s usd to comput an atttud stmat fom th gyos by th ntgaton of quaton (9). Expssd n matx fom, quaton (9) s gvn by q& 0 0 ωx ω y ωz q 0 q& ω 0 1 x ωz ω y q 1 = q 2 ωy ωz 0 ω (14) & x q2 q& 3 ωz ωy ωx 0 q 3 wh ωx, ωy, ω z a stmats of th tu angula ats n th x, y and z axs. Gvn an ntal atttud stmat, quaton (14) can b ntgatd to obtan th latst atttud stmat, q 0. A fnc atttud stmat s obtand fom acclomts. Th acclomt masumnts a gvn by v& = v& + ω v + G (15) ω I B B wh vb, v I a body and ntal vlocts, spctvly, G s th gavty vcto componnt n th body coodnats gvn by G = [ g snθ g cosθ snφ g cosθ cos φ] T (16) If a masu of fowad vlocty n th body fam s not avalabl, th oll and ptch angls obtand fom acclomts a couptd by th lna acclaton and th coss poduct tms nvolvng angula at and lna vlocty. If th fnc atttud quatnon s q l, th atttud o usng quaton (11) s gvn as l 0 q = q q (17) If th atttud o s small, q 0 1 and [ q q q ] v = can b assumd to b th os n oll, ptch and yaw atttuds. Th quatnon o can b usd to gnat angula at coctons usng ε ω = k v (18) wh k s an stmato gan. Th angula at cocton s usd as a fdback cocton as gvn n quaton (13). Th man fatus of QFLASP a 1. Adaptv Swtchng/Fltng: Th data flow s altd at untm. Thus, ctan flts a actvatd o dactvatd basd on qualty, consstncy, and chaactstcs of data. Th swtchng s mplmntd by mans of Fuzzy Logc USING INERTIAL MEASUREMENT TO SENSE CRASH TEST DUMMY KINEMATICS 19

4 (dscussd n nxt scton). Th Fuzzy stmato conssts of a fuzzfcaton pocss, an nfnc mchansm, a Rul Bas and a dfuzzfcaton pocss. Th fuzzfcaton pocss assgns a dg of mmbshp to th nputs ov th Unvs of Dscous. As th o changs, th dg of ctanty changs and oth masus of µ hav non-zo valus. Thus, th os a ncodd by th dg of ctanty that a btwn ctan o bounds. Th valus fo th o bounds (E 1, E 2 ) can b dtmnd usng cnt clustng tchnqus on actual cash-tst xpmntal data. Lkws, nput mmbshp functons a dtmnd fo th chang n o. Ths pocss s xpland n dtal n th nxt scton. 2. Adaptv Gan Tunng: Th gans of th body-at o, ntal-at o and dlay a tund dung untm. Thus, th algothm tuns tslf to povd an optmum soluton. In fact, th QFLASP-B output accuacy mpovs wth pod of us. Th sdual dft n th gyos and acclomts s movd by mans of fd-fowad flts and a mplmntd n an o-cocton loop. nn loop to dtmn knmatc paamts.g., oll, ptch, yaw, ntal vlocts, and postons va ntgaton. Quatnon Fuzzy Logc QFLASP s a novl appoach fo movng snso os. FLASP, lk Fuzzy Logc fom whch t s dvd, s a mo ntutv pocss than Kalman Fltng [16], [17]. As an xampl, th quatnon os and gyo bass a calculatd by ths algothm and usd n an adaptv loop to mov th ffcts. Th Fuzzy stmato conssts of a fuzzfcaton pocss, an nfnc mchansm, a Rul Bas and a dfuzzfcaton pocss. Th fuzzfcaton pocss assgns a dg of mmbshp to th nputs ov th Unvs of Dscous. Rfng to Fgu 2, f th o () n Eul angl k s zo, th dg of ctanty, µ 0 (cnt mmbshp functon), s 1 and all oths a zo. As th o changs, th dg of ctanty changs and oth masus of µ hav non-zo valus. Thus, th os a ncodd by th dg of ctanty of th o bounds. 3. Gavty Compnsaton: Acclomts masu spcfc foc,.., th acclomt dos not masu gavty but ath th componnt of total acclaton mnus gavty along ts nput axs. Th gavty-compnsaton functon n QFLASP acqus data fom acclomts, basd on th chaactstcs and valdty of data, usng appopat fltng and passs slavng nfomaton to a snso-fuson algothm. Th govnng quaton fo IMU dynamcs a gvn as & θ = ω cosφ ω snφ y & φ = ω + ω snφ tanθ + ω cosφ tanθ ψ& = ( ω snφ + ω cos φ) scθ z x y z y Th acclaton quatons a gvn as axcg = U& + ω W ω V + g sn θ ( a) y x z aycg = V& + ωzu ωxw g cosθ sn φ ( b) azcg = W& + ω V ω U g cosθ cos φ ( c) y z (19) (20) wh ωx, ωy and ω z a body-angula vlocts about x, y, z dctons, spctvly, and θ, φ and ψ a ptch, oll and yaw (ntal) angls. U, V, W a body vlocts n th x, y, z dctons, spctvly, and ax cg, aycg and azcg a ntal acclatons. It should b notd that quaton (19) s a coupld quaton n angula vlocts but dos not nvolv any acclaton tm. Equaton (20) nvolvs acclatons as wll as body ats. Thus, dsct vson of quaton (19) s solvd n an out loop and quaton (20) s solvd n th Fgu 2. Mmbshp Functons fo Fuzzy Logc Tabl 1. Rul Tabl fo Fuzzy Logc Th valus fo th o bounds (E 1, E 2 ) can b dtmnd usng th cnt-clustng tchnqus on cash data. Lkws, nput mmbshp functons a dtmnd fo th chang n o. Fo fv o nput mmbshp functons and fv chang-n-o nput mmbshp functons, twnty fv uls sult as sn n Tabl 1. Any mmbshp functon wth a non-zo dg of ctanty s sad to b on and th cospondng ul s also 20 INTERNATIONAL JOURNAL OF MODERN ENGINEERING VOLUME 10, NUMBER 2, SPRING/SUMMER 2010

5 actv [16], [17]. If both th o and chang n o w small nough to b wthn th smallst o bounds (-E 1 to + E 1 n Fgu 2), th lngustc ul s consdd as follows: If s zo and chang n s zo thn, cocton s zo. Th ctanty of th pms,, s gvn by: (1) µ = mn( µ 0, µ 0) In gnal, th uls a gvn as: (2) If µ% s j A and % µ s ε = g and ε& = h ( ) thn ( ) k A l, Th symbol smply ndcats th AND agumnt. In QFLASP-B, th quatnon o s fst ducd to th o n Eul angls: ε { ε, ε, ε } (21) q φ θ ϕ Th uls of Tabl 1 a thn appld to ach Eul angl o. Th output cocton fo ach Eul angl and Eul at s calculatd usng a cnt-of-gavty mthod: ˆ ε R = 1 = 1 ulangl =, ˆ ε ulangl = R R g µ h µ µ µ = 1 = 1 R (22) Kalman Flt by th matx nvson, whch s absnt n th QFLASP-B. Smulaton and Expmntal Rsults H, th authos tstd th QFALSP-B algothm though ADAMS-LfMOD smulatons. In ths tst cass, QFLASP-B pfomanc fo had-ontaton computaton and toso-ontaton computaton n a fontal cash stuaton was nvstgatd. Th fowad knmatcs poblm was posd and solvd n ADAMS [18] that gvs body ats and acclatons. Ths body ats and acclatons w fd nto QFLASP n od to solv th nvs knmatcs poblm. Ths smulatons wll hlp to nsu that: 1. QFLASP-B swtchs a wokng poply. Th s no tm dlay n opatng swtchs. A dlay n swtchng would sult n naccuat o ncoct solutons. 2. QFALSP dos not ncount sngulats. Th advantag of QFLASP-B ov FLASP-Eul angl fomulaton s that t can handl 90 ptch stuatons. Unfotunatly, th computaton s tm-consumng and may ntoduc goup dlay that would coupt th soluton. 3. At ths plmnay stag, t s not possbl conduct to xtnsv lab tstng that would valdat QFLASP-B fo all possbl cash scnaos. It s antcpatd that ths smulatons wll val poblms wth QFLASP- B and hlp tun th algothm btt. Coctons to th body ats can thn b dtmnd. To apply quatnon coctons, th stmatd o quatnon must b constuctd: ε { ε, ε, ε } q φ θ ϕ (23) Ths coctons a thn appld to th quatnon to mov quatnon o. Onc th atttud s dtmnd, angls and at can b substtutd nto quaton (20). Equaton (20) can b ntgatd to calculat th vlocty and poston. It s notd that th Fuzzy logc appoach dscussd h s gnc and can b xtndd to mnmz acclomt systmc os. Whl smla sults to th QFLASP-B can b obtand usng a Kalman Flt, th opatonal softwa ovhad s consdabl. In ou own tsts, th Kalman Flt took 3.5 ms to un p taton, whl QFLASP-B took und 1 ms p taton on a Txas Instumnt C33 DSP wth a clock spd of 60 MHz. Smlaly, th Kalman Flt cod qud mmoy of naly 10,000 wods, whl th QFLASP-B was und 3,000 wods. Both qumnts w dvn n th Fgu 3. ADAMS-LfMOD Stup fo vtual cash tstng It was notd that th fowad dynamcs data can b couptd usng gyo, acclomt-bas modls so that ths USING INERTIAL MEASUREMENT TO SENSE CRASH TEST DUMMY KINEMATICS 21

6 datast would b vy clos to xpmntal data. Ths couptd data can b fd nto QFLASP-B to valuat th ffctvnss of th algothm. Fo ths sampl smulaton, th dummy was postud as an occupant dvng th ca wth 3- pont sat blts attachd (f to Fgu 3). A tanslatonal jont was catd btwn th gound and ca sat to smulat mpact (standad SAE shock puls). An qulbum analyss was cad out so that th modl sttls und th acton of gavty. An appopat coodnat tansfomaton matx was usd to ppocss th had angula at (shown n Fgu 4) and had acclaton data obtand fom ADAMS smulatons bfo fdng thm to QFLASP-B. Had angula vlocty n body coodnats s shown n Fgu 4. It s notd that ω x achs a maxmum valu of 1500 dg/s. Th had ontatons computd by QFLASP-B fom aw acclomt and gyo data a shown n Fgu 5. It can b obsvd n Fgu 5 that th fowad ADAMS atttud soluton (obtand fom MSC-ADAMS-LfMod ntal mak) matchs qut closly wth th nvs QFLASP soluton. Fgu 5. Had Intal Fowad Knmatcs- ADAMS soluton (ndcatd by dottd lns) and Invs Knmatcs-QFLASP Soluton (ndcatd by sold lns). Fgu 4. Had Angula Vlocty n body fam In th scond smulaton, toso spons s valuatd n a fontal cash. As bfo, th dummy was constand by a 3- pont lap should blt and a standad SAE shock puls was appld. Cntal toso body acclatons about th x, y and x axs a plottd n Fgu 6(a). Cntal Toso body angula vlocts a plottd n Fgu 6(b). Th atttud soluton s psntd n Fgu 6(c). It should b notd that du to lapshould blts, th cntal toso moton s stctd. Th body acclatons and body ats obtand fom ADAMS- LfMod smulatons w fd nto QFLASP-B to gt toso ontatons n ntal coodnats. a) Cntal Toso Body Acclatons vs. tm (sconds) 22 INTERNATIONAL JOURNAL OF MODERN ENGINEERING VOLUME 10, NUMBER 2, SPRING/SUMMER 2010

7 It should also b notd that th ADAMS soluton snsd by a mak n an ntal fam matchs qut closly to th atttud solutons computd by QFLASP-B, as shown n Fgu 6(c). Expmntal Valdaton b) Cntal Toso Body Angula Vlocty vs. tm (sconds), (R1-otaton about x, R2-otaton about y, R3-otaton about z axs) In plmnay studs, th algothm was tstd fo had knmatcs. Th pupos of ths tst was to valuat hadstant sponss. Th was about 20 msc of p-cash data (about 250 ponts), whch was usd fo bas captu. Th tst was don on a Hybd III 50th pcntl dummy wth NAP and angula snsos. Th advantag of such a confguaton s that t s possbl to comput had ontatons usng th NAP algothm [1] and also acclaton and at data can b fd nto QFLASP-B. Thus, had ontatons can b computd wth two dffnt tchnqus. Th stup wth snso-mountng locatons s shown n Fgu 6. Th dummy was subjctd to standad SAE cash pulss and th aw snso data n body coodnat systm s shown n Fgu 7. It s mpotant to not that fo manngful sults, th IMU coodnat systm (at th CG of th had) should b mappd to th SAE coodnat systm. c) Cntal Toso Intal Fowad Knmatcs- ADAMS soluton (dottd ln) and Invs Knmatcs-QFLASP Soluton (sold ln). Fgu 6. Cntal Toso Acclatons, Body Angula vlocty and ntal soluton IMU Coodnats) Fgu 6. Stup fo Had Knmatcs Tst (snso mountngs locatons shown n gn) USING INERTIAL MEASUREMENT TO SENSE CRASH TEST DUMMY KINEMATICS 23

8 wth th fnc NAP soluton shown n Fgu 8. Howv, du to lss computaton ovhad, th spd of xcuton of QFLASP-B s much hgh and QFLASP-B uss body acclatons fo slavng (o to coct atttud soluton usng acclomt data). Thfo, unlk th NAP pocssng algothm, QFLASP-B can b opatd n untm fo a long duaton o on a much chap DSP platfom, f qud. Oth knmatc paamts such as angula acclatons, lna acclatons and ntal angula vlocty can b asly dvd usng ths appoach, whch can b usd to comput njuy masumnts. It can b sn that ths ntal masumnt allows fo computng latv ontatons of vaous body pats, such as had otaton wth spct to nck, n a fxd fnc fam. Fgu 7. Raw Rat Snso Data n Body Coodnats An appopat coodnat tansfomaton matx was usd to ppocss th data bfo fdng thm nto QFLASP-B. In Fgu 7, t can b obsvd that ats about th y axs a vy hgh at about 800 dg/s. Th ontatons computd by QFLASP-B a shown n Fgu 8. It can b notd that ptch vas fom +20 to -30 (bound moton). Howv, oll and yaw a wthn 5. Ths sults compa favoably Fgu 8. Atttud Soluton n Had Rstant Tst NAP Soluton (ndcatd by dottd lns) and QFLASP-B Soluton (ndcatd by sold lns) Dscusson and Conclusons In ths study, th authos psntd a novl appoach fo snsng dummy knmatcs n cash vnts usng ntal masumnt va Fuzzy logc. Snsng of dummy knmatc paamts n a cash vnt s cucal fo valuatng th cashwothnss of vhcls. Ths knmatc paamts w usd to comput vaous njuy paamts, to undstand th svty of njus, to study th ffctvnss of sat blts and abags, and oth occupant safty dvcs. Th ntal-snsng appoach dscussd h s basd on snsng ats and acclatons n 3 mutually ppndcula dctons and usng a Fuzzy-Logc-basd algothm fo computng nvs knmatc ntal solutons. Ths supo sgnal-pocssng algothm compnsats fo snso os lk nos and dft psnt n typcal low-cost MEMs snsos and povds qually accuat solutons nomally obtand by xpnsv tstng mthods/snso suts. Ths quatnon-basd appoach s f fom sngulats at 90 and, unlk th Kalman flt, dos not nvolv matx nvsons. Th hadwa and softwa aspcts of ntal snsng along wth smulatons and plmnay xpmntal sults w also dscussd. In smulatons, fowad-knmatcs poblms w posd and solvd n MSC -ADAMS to obtan body ats and acclatons. Ths acclatons and ats w fd nto QFLASP to obtan an nvs knmatc ntal soluton. Two smulaton cass, had-ontaton calculaton and toso-ontaton calculaton, w psntd. In both cass, th MSC-ADAMS soluton obtand va a mak n an ntal fam matchs vy closly wth th QFLASP soluton. In plmnay xpmntal tsts, th QFLASP algothm was tstd to comput had ontaton and compad aganst th soluton computd by a standad NAP snso sut. Th NAP soluton and QFLASP soluton matchd qut wll. Cuntly, ffots a undway to tst QFLASP n a vaty of cash stuatons. QFLASP can b mplmntd on 24 INTERNATIONAL JOURNAL OF MODERN ENGINEERING VOLUME 10, NUMBER 2, SPRING/SUMMER 2010

9 a low-cost DSP at much hgh updat ats. It s antcpatd that ths Intal Tackng/Masumnt appoach wll nabl tst ngns to us low-cost MEMs snsos fo cash tstng and povd an nxpnsv altnatv fo masung knmatc paamts n a cash vnt. Acknowldgmnts Th authos thank D. Mchal Gn and M. Vcto Tnt of Achangl Systms fo th hlp n ths pojct. Suppot fom MSC Softwa and th US Dpatmnt of Tanspotaton s also gatfully acknowldgmnt. Rfncs [1] J. Hll, M. Rgan, R. Adzn, and L. Esnfld, Systm fo Rcodng th Bowl Sounds of Pmatu Infants, ASME Bomd 2008 Confnc, Jun [2] A. J. Padgaonka,K. W. Kg and A. I. Kng, "Masumnt of angula acclaton of a gd body usng lna acclomts.",jounal of Appld Mchancs 42, pp , 1975 [3] JRW Mos" Acclomty - A tchnqu fo th masumnt of human body movmnts." Jounal of Bomchancs, 6, pp , [4] R. E. Mayagota, P. H. Vltnk," Acclomt and at gyoscop masumnt of knmatcs: an nxpnsv altnatv to optcal moton analyss systms". Jounal of Bomchancs, 35(4), pp , 2002 [5] A. J. van dn Bogt, L. Rad and B. M. Ngg, "A mthod fo nvs dynamc analyss usng acclomty". Jounal of Bomchancs. 29(7), pp ,1996. [6] K. Amnan K. and B. Najaf, "Captung human moton usng body-fxd snsos: outdoo masumnt and clncal applcatons.", Comput Anmaton and Vtual Wolds, 15, pp.79 94, [7] E.B. Bachmann "Intal and magntc tackng of lmb sgmnt ontaton fo nstng humans nto synthtc nvonmnts". PhD Thss, Naval Postgaduat School, [8] H. J. Lung "Intal snsng of human movmnt". PhD Thss, Unvsty of Twnt, [9] [10] [11] [12] E. Foxln "Intal had-tack snso fuson by a complmntay spaat bas Kalman flt". In Pocdngs of VRAIS 96, pp , [13] R. G. Bown, Intoducton to Random Sgnals and Appld Kalman Fltng, Wly Publshng, [14] M. S. Gwal, L. R. Wll and A. P. Andws,Global postonng systms, ntal navgaton, and ntgaton, Wly-Intscnc, 2000 [15] Wahba, G., "A Last-Squas Estmat of Spaccaft Atttud". SIAM Rvw, 7, 3, pp , 1965 [16] Gn M. and Tnt V., Softwa algothms n a data atttud hadng fnc systms, Acaft Eng. And Aospac Tch., Wst Yoksh, Emald Pub., 75(5), pp , [17] K. Naayanan and Gn M., " A Unt Quatonon and Fuzzy Logc Appoach to Atttud Estmaton, In th pocdngs of ION NTM 2007 pp [18] Bogaphs SANGRAM REDKAR s an Assstant Pofsso n Engnng Tchnology at Azona Stat Unvsty (ASU) D. Rdka may b achd at sangam.dka@asu.du TOM SUGAR s an Assocat Pofsso n th Engnng Dpatmnt at ASU. D. Suga may b achd at thomas.suga@asu.du ANSHUMAN RAZDAN s an Assocat Pofsso n th Engnng Dpatmnt at ASU. D. Razdan may b achd at a@asu.du UJWAL KONERU s a gaduat studnt n th Dpatmnt of Comput Scnc at ASU. M. Konu may b achd at ujwal.konu@asu.du BILL DILLARD s th Dcto of Emgng Tchnologs at Achangl Systms, Aubun, AL. H can b achd at bll@achangl.com KARTHIK NARAYANAN s th lad softwa ngn at Achangl Systm. H can b achd at kathk@achangl.com USING INERTIAL MEASUREMENT TO SENSE CRASH TEST DUMMY KINEMATICS 25