Comparison of Lateral Control in a Reconfigurable Driving Simulator

Transcription

1 Comparson of Lateral Control n a Reconfgurable Drvng mulator Daha Guo Ph.D. tudent Department of Computer Engneerng chool of Electrcal Engneerng and Computer cence Unversty of Central Florda, Orlando, FL, 386 Phone: (407) Fax: (407) Emal: dgu@bruce.engr.ucf.edu Harold I. Klee, Ph.D. Assocate Professor Department of Computer Engneerng chool of Electrcal Engneerng and Computer cence Unversty of Central Florda, Orlando, FL, 386 Phone: (407) Fax: (407) Emal: klee@mal.ucf.edu Essam Radwan, Ph.D., P.E. Professor and Char Department of Cvl and Envronment Engneerng Unversty of Central Florda, Orlando, FL, 386 Phone: (407) Fax: (407) Emal: aeradwan@mal.ucf.edu ubmsson Date: eptember, 003

2 ABRAC he UCF drvng smulator was desgned to be reconfgurable so that ether a commercal truck cab or entre passenger vehcle (aturn wth engne and transmsson removed) can be mounted on a moton platform. everal vehcle dynamcs models are avalable for varous truck/traler combnatons as well as dfferent passenger vehcles. A study was performed to assess whether there are sgnfcant dfferences n drvng patterns, specfcally wth respect to lateral control for both class of vehcles. ubjects were nstructed to drve through a vsual database n the truck and aturn. Drver nputs, gas, brake and steerng wheel along wth vehcle poston was logged at 30 Hz. A test secton of two lane road approxmately 350 meters n length wth several bends was used to compare the lateral dsplacement from the centerlne n the truck and aturn. ests of hypotheses nvolvng the dfferences n populaton parameters (mean and varance) for each of 5 drver's lateral poston whle drvng n the truck and aturn were conducted. he drver's steerng behavor for both types of vehcles was also compared usng frequency doman analyss. Results are presented n the paper. HE MARK II YEM FROM GE-DD he Mark II system s manufactured by GE-DD, whch has been makng drvng smulators for many years. he truck vehcle cab s a md-szed, late-model manufacturers cab shown n Fg.. FIGURE he Vehcle Cab Of he Mark II ystem he cab s nstalled on a 6-degree-of-freedom moton base. hree channels of vson are projected on 3 flat whte walls n front of the vehcle cab as shown n Fg.. wo channels of vson are dedcated for rear vew mrrors whch are actually two LCD montors. he mage generaton software s runnng on 6 personal computers. One of them nstalls the vehcle dynamcs model and calculates smulator poston data and moton of the cab. he other fve computers generate mage sequences for the fve channels. A 7 th PC s equpped wth the operaton console and s for loadng the vehcle dynamc models and the exstng traffc scenaros. he cenaro Edtor s used to desgn new drvng scenaros.

3 FIGURE he Drvng Envronment GE-DD also delvered and nstalled a aturn re-confgurable cab on the moton base (see Fg.3). FIGURE 3 he aturn On he Moton Base EXPERIMENAL DEIGN We dd not have any pror knowledge about how the lateral control of trucks compares the lateral control of passenger cars, but accordng to the observatons n our prevous study wth the smulator, t was easy to tell that the subjects normally perform dffe rently n keepng the smulator on the center of the lane when drvng the truck and the passenger car. otally, ffteen subjects partcpated n ths study, ten of whch were males and fve of whch were females. Every one of them had Florda drver lcense. he dstrbuton of age wll be revewed later n ths paper. No smulator sckness happened to any of the subjects. he subjects were nstructed to drve on a segment of road whose approxmate geometrc shape s shown n Fg.4. Each drver was nstructed to observe the posted speed lmt and mantan approprate lateral control n both the aturn vehcle and truck. For each subject, the two experments, n the truck and n the aturn were separated by at least two months. Besdes, each of the subjects was gven a short orentaton, actng as the tranng about the smulator before the real smulaton. From vehcle dynamcs pont of vew, we used two models. he truck s dmensons are 7.09-meter long and.4-meter wdth. he wheel wdth s.79-meter wde. Another one s passenger car. Its dmensons are 4.97-meter long,.8-meter wde. Its wheel wdth s.63-meter wde.

4 FIGURE 4 est egment of Road he arrows ndcate the drecton of travel and the two black bars ndcate the startng pont and endng pont of data loggng, whch conssted of the D vehcle coordnates and the followng drver nputs: brake, steerng wheel angle and accelerator. Data loggng was performed at a rate of 30 Hz. Each smulaton requred from mnute to mnutes, dependng on dfferent drvers speeds. he amount of raw data vared from 800 to 3600 numercal values for each logged varable. DAA ANALYI o compare lateral control of the aturn and truck, devatons from the center of the lane n the smulaton world were calculated. he lane wdth was feet. he devaton profles are a measure of how well the drvers could mantan lateral control of the vehcles. Excessve devatons are an ndcaton of possble flaws n the lateral dynamc response of the smulator. Fg. 5 s the hstogram of the lateral devaton of a subject, drvng the passenger car. It s easy to tell that ths subject kept the car near the center of the lane most of tme, but the large lateral devaton dd happen. hs should be due to people s normal drvng habt.

5 FIGURE 5 he Hstogram of Lateral Devaton (count vs. lateral devaton n meters) he followng table ntroduces the varables used n the statstcal analyss of the calculated centerlne lane devaton data. ymbol m j n n c c Explanaton Number of subjects, who drove the same segment of road usng both the truck and aturn. ubscrpt denotng a specfc drver uperscrpt denotng the truck Number of raw data ponts for subject n the truck uperscrpt denotng the aturn. Number of raw data ponts for subject n the truck Number of raw data ponts for subject n the aturn he autocorrelaton nterval for subject n the truck he autocorrelaton nterval for subject n the aturn r Reduced sze of data for subject n the ruck r d, j d, j Reduced sze of data for subject n the aturn Devaton from the center of the lane to the poston of the smulator (ruck) Devaton from the center of the lane to the poston of the smulator (aturn) µ Mean of the devaton of for subject n the truck

6 µ Mean of the devaton of for subject n the aturn d d ample mean devaton of fo r subject n the truck ample mean devaton of for subject n the aturn t he t value accordng to the threshold for subject and the t dstrbuton, t he actual t value for subject ABLE he Defnton of the Varables Because the data loggng occurred at 30 Hz, adjacent data values n the raw data are hghly correlated. In order for the sample data to be ndependent observatons from the same populaton, the autocorrelaton functon must be known. Fg.5 shows one of the emprcally determned autocorrelaton functons. he nterval between ndependent observatons s selected to be the tme of the frst local mnmum or the frst autocorrelaton less than Consequently, the orgnal correlated sample s reduced n sze by a sgnfcant amount. FIGURE 6 Autocorrelaton Functon for ubject 5 n the ruck Autocorrelaton for subject and the truck wth samplng nterval k s n k j= n ( d j= ( d )( d + k Autocorrelaton for subject and the aturn wth samplng nterval k s ) )

7 n k j= n ( d j= ( d )( d he followng table contans the samplng ntervals, whch cause ether the frst local mnmum or the frst autocorrelaton value less than 0.05, and the reduced sample szes. ) + k ) ubject Length of correlated nterval # of ndependent observatons c c r r ABLE Autocorrelaton Functon Results None of the subject reduced sample szes was greater than 30 n the truck or the aturn. As a result, the t- dstrbuton was chosen for use n the statstcal analyss of the reduced sze ndependent samples of lane devatons. [] he Null Hypothess assumes that the mean lateral devaton n the truck s dentcal wth the mean lateral devaton n the aturn for each subject drver. hat s, Null Hypothess H 0 : Alternatve: µ µ µ = µ, =,,..., m he null hypothess s accepted f t t, [ t,, t, hypothess s accepted or not for each drver. ubject <, otherwse t s rejected. he confdence nterval s ]. able 3 contans the degrees of freedom for dfferent drvers, t, d.o.f. t t, t,, and whether the Null, crtcal value H 0 Accepted No No Yes

8 No Yes No No Yes Yes No No No No Yes No ABLE 3 Null Hypothess est Results he hypotheses are accepted for 5 subjects and rejected for 0 subjects. By observaton of the t values n able 3, t follows that the lateral devatons of drvers are less n the aturn than n the truck. PECRAL ANALYI OF HE EERING INPU Nne out of ten tmes when the Null hypothess was rejected, larger devatons from the center of the lane were observed n the truck than n the passenger car. However, results from a post smulaton survey found that most subjects were satsfed wth the response of the truck smulator and complaned about the performance of the aturn, especally wth respect to the brake and steerng wheel. Prevalng opnon was that the aturn s brake dd not feel lke a real brake. Wth respect to the aturn s steerng wheel, subjects commented that the feedback was too loose and the wheel would not center when released. he brake s performance s unrelated to controllng the lateral poston of the vehcle, however the steerng wheel s performance does play an mportant role. he spectral propertes of steerng wheel dsplacements n the aturn and truck s summarzed by ther Fourer transforms. o be more specfc, the FF (Fast Fourer ransform) of each drver s steerng nput n the truck and n the aturn s requred. For each drver there are two FFs, one for the truck and one for the aturn. hey are denoted by FF and FF. o help decde whether the dfference n the steerng wheel s performance contrbuted to acceptng or rejectng the hypotheses, the sum of the FF power was computed for all fve drvers where the hypotheses were accepted. In addton, the same was done for fve out of the ten drvers where the hypotheses were 5 ( FF ) and ( FF = rejected. he result s to calculate = where the Null hypotheses were accepted and the second set of fve drvers for whch t was rejected. he four FF power plots are shown n Fgure 7 and 8. 5 ) for both sets of fve drvers, the frst

9 5 x ruck aturn FIGURE 7 teerng Input Power (Null Hypothess Accepted) 3.5 x ruck aturn FIGURE 8 teerng Input Power (Null Hypothess Rejected) By observaton of Fgures 7 and 8, t appears that drvers turned the steerng wheel more frequently n the truck than they dd n the aturn, ndependent of acceptance or rejecton of the Null hypothess. One explanaton s that the steerng n the aturn was easer (looser) compared to the truck. he actual reason

10 why the spectral powers are dfferent s not mportant. What s mportant s ths dfference s not a factor n decdng whether the Null hypothess was accepted or rejected because smlar dfferences between aturn and truck) exsted n both cases (acceptance and rejecton of Null hypothess ). We conclude that the dfference n lateral poston when drvng the truck and the passenger car can be attrbuted to a combnaton of the subjects drvng habts and the nherent dfferences n the aturn and truck vehcle dynamcs. On another aspect, the null hypothess accepted group put a lot more effort nto steerng both vehcles. hat s because ths group of drvers ntended to keep a good lateral poston, so they kept adjustng the steerng wheel, whch results n hgh spectral power. FUURE WORK able 4 presents the facts of ages, average speeds and standard devatons of speeds of the ffteen subjects. Number Gender/Age Average peed (mph) td Devaton of peed H 0 Accepted ruck/aturn (mph) ruck/aturn Male/7 49.4/ /4.5 No Female/7 37.8/ /7.0 No 3 Male/39 6.6/ /0.4 Yes 4 Male/ 4.4/ /5.3 No 5 Female/9 43.9/ /9. Yes 6 Male/ 4./59.5.3/7.5 No 7 Male/8 50.9/ /0. No 8 Male/4 38.3/ /9. Yes 9 Male/4 33.6/39..0/.0 Yes 0 Male/3 3./ /8. No Female/37 4./ /.7 No Female/ / /7.6 No 3 Male/ 33./ /9.7 No 4 Male/ / /4.6 Yes 5 Female/ / /8.4 No ABLE 4 he Facts of Age and peed For those sx male hypothess -accepted subjects, the average age s 39.5 year-old. For those four male hypothess -rejected subjects, the average age s 3.3 year-old. For those four female hypothess -rejected subjects, the average age s 33.8 year-old. here s only one female hypothess -accepted subject, whose age s 7 year-old. here exsts somewhat dfference between the average ages, but ths could be caused by small sample sze. In the future, we are gong to nvte more subjects wth a certan age dstrbuton and repeat what we dd n ths study. hen we wll be able to put the factor of age nto concern. Addtonally, we could analyze the causal relatonshp between the speed and the lateral poston. From the data above, some subjects average speeds n the truck and the aturn vary sgnfcantly, lke No. 3 and 6. ome of those vary not much, lke No. 5 and 7. What s nterestng s that the hypotheses are rejected n No. 3 and 5 and accepted n No. 6 and 7. herefore, we need more subjects and analyze the relatonshp between speed or the standard devaton of speed and lateral control. Besdes, the vehcle dynamcs may also play an mportant role n controllng the lateral poston. herefore n the future, we also can compare the dynamcs of these two models. In ths study, we used a curve road. In the future, we also plane to conduct smlar lateral control studes on dfferent road segments to see whether the results are repeatable,.e. ndependent of road geometry.

11 Concluson In ths study, 5 subjects drove along a segment of road n a drvng smulator n both a aturn passenger car and a truck. Based on statstcal analyss of lateral poston, two thrds of the drvers responded dfferently n the aturn compared to the truck. Furthermore, the devaton from the center of the lane n the truck was greater than n the passenger car. Reference: [] Mendenhall W. and ncch. tatstcs for Engneerng and the cences. 4 th Edton, Prentce-Hall, Inc. New Jersey, 995.