a new approach to modeling driver reach

7/31/2019 A New Approach to Modeling Driver Reach

1/12

400 Commonwealth Drive, Warrendale, PA 15096-0001 U.S.A. Tel: (724) 776-4841 Fax: (724) 776-5760 Web: www.sae.org

SAE TECHNICAL

PAPER SERIES 2003-01-0587

A New Approach to Modeling Driver Reach

Matthew P. Reed, Matthew B. Parkinson and Don B. ChaffinUniversity of Michigan

Reprinted From: Human Factors in Driving, Seating, & Vision(SP-1772)

2003 SAE World CongressDetroit, Michigan

March 3-6, 2003


2/12

All rights reserved. No part of this publication may be reproduced, stored in a retrieval system, ortransmitted, in any form or by any means, electronic, mechanical, photocopying, recording, or otherwise,without the prior written permission of SAE.

For permission and licensing requests contact:

SAE Permissions400 Commonwealth Drive

Warrendale, PA 15096-0001-USAEmail: [email protected]

Fax: 724-772-4028Tel: 724-772-4891

For multiple print copies contact:

SAE Customer ServiceTel: 877-606-7323 (inside USA and Canada)

Tel: 724-776-4970 (outside USA)Fax: 724-776-1615Email: [email protected]

ISSN 0148-7191

Copyright 2003 SAE International

Positions and opinions advanced in this paper are those of the author(s) and not necessarily those of SAE.The author is solely responsible for the content of the paper. A process is available by which discussionswill be printed with the paper if it is published in SAE Transactions.

Persons wishing to submit papers to be considered for presentation or publication by SAE should send the

manuscript or a 300 word abstract of a proposed manuscript to: Secretary, Engineering Meetings Board, SAE.

Printed in USA


3/12

2003-01-0587

A New Approach to Modeling Driver Reach

Matthew P. Reed, Matthew B. Parkinson and Don B. ChaffinUniversity of Michigan

Copyright 2003 SAE International

ABSTRACT

The reach capability of drivers is currently represented invehicle design practice in two ways. The SAERecommended Practice J287 presents maximum reachcapability surfaces for selected percentiles of a genericdriving population. Driver reach is also simulated usingdigital human figure models. In typical applications, afamily of figure models that span a large range of thetarget driver population with respect to body dimensionsis positioned within a digital mockup of the driver'sworkstation. The articulated segments of the figuremodel are exercised to simulate reaching motions anddriver capabilities are calculated from the constraints ofthe kinematic model.

Both of these current methods for representing driverreach are substantially limited. The J287 surfaces arenot configurable for population characteristics, do notprovide the user with the ability to adjust accommodationpercentiles, and do not provide any guidance on thedifficulty of reaches that are attainable. The figure model

method is strongly dependent on the quality of themodels used for posturing and range of motion, and, inany case, cannot reliably generate populationdistributions of either reach capability or difficulty.

A new method of modeling driver reach capability ispresented. The method is based on a unified model ofreach difficulty and capability in which a maximum reachis a maximally difficult reach. The new approach ismade possible by new measurement methods that allowdetailed and efficient sampling of an individuals reach-difficulty function. This paper summarizes theexperimental approach and presents the structure of the

new integrated model of population reach difficulty andcapability.

INTRODUCTION

SAE Recommended Practice J287

Reach analysis is one of the primary activities inassessing driver workstation layout. Vehicle designersand engineers must verify that the primary andsecondary controls can be reached and manipulated byan acceptable percentage of the driver population. The

most commonly used tool for representing driver reachcapabilities is the model embodied in SAERecommended Practice J287 (SAE 2002). The J287surfaces are parameterized by a packaging factor thatcombines vehicle interior dimensions, including seaheight, fore-aft and vertical steering wheel position, andsteering wheel diameter, into a single "G" scoreSurfaces are available for several different values of G.

In the laboratory study on which J287 is based(Hammond and Roe 1972), maximum reach data wereobtained from 120 men and women in three vehiclepackage configurations. Figure 1 shows the tesapparatus. Drivers grasped the ends of themeasurement rods and pushed them forward as far aspossible. The rod rack was moved laterally to span thespace in front of the driver. No data were collected fovertical or lateral reaches. Testing was conducted withlap belts only and with lap and fixed-length torso beltsThe data were analyzed to produce reach surfaces, anexample of which is shown in Figure 2. The reachsurfaces are interpreted with respect to population

capability, as opposed to the capability of individualswith a particular body size. For example, 95 percent odrivers from a 50/50 male/female U.S. driving populationare expected to be able to reach to push-button targetsthat are located aft of the surface depicted in Figure 2Note that this is notthe reach capability of a male driverwho is 95

thpercentile by stature, or of any other

particular anthropometric category.

Figure 1. Test fixture used for the SAE driver reach study(Hammond and Roe 1972).


4/12

2

Figure 2. Surface generated from J287 (Roe 1993).

The reach surfaces in J287 are positioned relative to thevehicle seating reference point (SgRP), whichcorresponds to a seat H-point location obtained with theseat near the rear of the seat adjustment path.

J287 presents reach surfaces for two restraintconditions: lap belt only and lab belt plus fixed-length

shoulder belt. The emergency locking retractor usedalmost universally on modern vehicles, which allowsspooling of the belt with low belt tension, was not yet inwidespread use when the original study was conducted.Although the lap-belt-only condition would seem to bemore representative of the restraint condition in modernvehicles, industry practitioners more commonly use thelap-plus-torso-belt surfaces to develop design guidelinesand assess designs. This practice is motivated by theobservation that controls located within the more-restrictive surfaces obtained with torso restraint arealmost certainly reachable by a large percentage of thedriving population. In effect, practitioners use the lap-

plus-torso-belt maximum-capability surface as anacceptable-difficulty surface for drivers with modernretracting belts.

One of the strengths of the J287 approach is that itincorporates an implicit seat-position model. Themaximum reach of a driver relative to the vehicle is afunction of the drivers capabilities (including thosedetermined by body dimensions) and the drivers seatposition and torso posture. The locating functions inJ287 take into account the effects of vehicle packagegeometry and body dimension distributions on driver-selected seat position. However, the seat-position

model implicit in J287 cannot be separated from thereach model, and hence the accuracy of the reachcapability assessment is dependent on the extent towhich the G-factor equation accurately reflects theeffects of package variables on driver-selected seatposition. Recent research leading to the development ofa new seat-position model (Flannagan et al. 1998)suggests that the range of package variables used in theoriginal study may have been insufficient to yield anaccurate seat position model.

There is also no capability within J287 to adjust for adifferent population (taller, shorter, or more obese, foexample). Most critically, J287 provides no informationon how difficult an attainable reach will be. If the modeis accurate, a control located just inside the 95th-percentile surface (for example) would be unreachableby only 5 percent of the driving population. Howeverthe distribution of reach difficulty for the 95 percent whoare capable of reaching the control is unknown. Hencethe practice provides no guidance for optimizing controlocations within the attainable reach zone. As notedabove, the common work-around for this limitation is touse the more-restrictive lap-plus-shoulder-belt surfacesrather than the lap-belt-only surfaces that might bethought to be more applicable to drivers in modernvehicles. However, this approach may be unnecessarilyrestrictive. As the number of controls presented to thedriver increases, the need for a more robust and flexiblemodel of driver reach capability and reach difficulty hasbecome more apparent.

The Figure Model Approach

Over the past decade, digital human figure models(computer manikins) have been used for an increasingnumber of driver workstation assessments (Porter et al1993; Chaffin 2001). Manikins are commonly used tosimulate reach to controls, with the outcomes of themanikin-based studies used to assess control locationsThe primary advantage of manikin-based reachassessments is that the visual depiction of a personperforming a task has considerable value forcommunicating the analysis results. The intendedmeaning of a task assessment depicted with a figuremodel is much clearer to an audience unfamiliar withergonomic assessment than the same problem analyzed

using the SAE J287 reach surfaces.

However, manikin-based approaches to modeling drivereach capability provide considerably less informationthan the statistically based models. Reach informationfrom a single manikin, representing a single set of bodydimensions starting from a particular driving postureprovides little information that is useful for making aquantitative assessment of a vehicle interior designMost users of figure models will repeat a reach analysiswith a family of manikins having different bodydimensions, but this approach provides only theappearance of capturing the range of variability in reach

capability within the driver population.

The primary limitation in reach analyses using figuremodels is that variance in body dimensions is only onefactor influencing reach capability. Equally important isthe drivers posture and position, which are the startingconditions for the reach. People who areanthropometrically similar often sit with widely differenseat positions and torso recline angles. For exampleFlannagan et al. (1998) found that the standard deviationof fore-aft seat position for people of the same stature in


5/12

3

passenger cars and light trucks is about 30 mm. Sincefore-aft seat position is a primary determinant of what adriver can reach in package space, any single seatposition used with a particular figure model will be atbest a rough approximation of the reach capability ofpeople of that size. Monte-Carlo statistical techniquescan be used to perform analyses in an automatedfashion by combining variance in seat position and torsoposture with sampled sets of anthropometric variables.However, the data required to validate such an approachcould be used to generate a statistical model that couldbe much more easily implemented.

Moreover, reach envelopes generated from figuremodels rely on assumptions about joint ranges of motion(ROM). Although joint ROM is configurable in mosthuman models, distributional data for joint ROM acrossindividuals, and the association between joint ROM andanthropometric variables, is not available within mosthuman modeling software packages. Hence, reachanalyses are usually done with a wide range of manikinsizes but only one or a few sets of joint ROM.

Figure models are able to predict comfort or discomfortratings for the terminal posture of a reach, but thisinformation cannot be reliably linked to populationdistributions of ratings. If the discomfort rating from oneof a family of ten manikins used to simulate a reachexceeds some criterion, it is not generally accurate toconclude that 10 percent of drivers would also rate thereach discomfort above the criterion. Even if themanikins are selected to be anthropometricallyrepresentative of the population, the postural variabilitythat is unrelated to anthropometry will not berepresented in the analysis.

Other Population Studies

The design of the SAE driver reach study (Hammondand Roe 1972) was probably influenced by previousreach studies. Figure 3 depicts maximum reachenvelopes for Air Force pilots from Kennedy (1964). Inthat study, pilots sitting on a rigid seat and restrained bya torso harness pushed against measurement rods todemonstrate their maximum reach capability. Themeasurement rods were rotated around the pilot to mapout the maximum reach envelope in a cylindricalcoordinate system. Envelopes for 5th, 50th, and 95th-

percentile maximum reach were presented graphicallyand in tabular form.

Figure 3. Typical data from Kennedy (1964), showingselected percentiles of grasp reach for pilots on one horizontalplane.

Several studies of driver reach have been conductedsince the original SAE controls reach study, but nonehas had an important effect on industry practice in theU.S. Stoudt et al. (1970) published reach data obtainedin a study of over 1000 men and women. However, thedata were collected in a vehicle mockup with a rigid seatand without a steering wheel. The data were alsopresented with respect to the seat, rather than thevehicle package, so the data could not effectively be

used for vehicle design. The increased prevalence oseat belts during the 1960s and 1970s led severainvestigators to conduct studies of the effects ofrestraints on reach (Haslegrave 1970; Bullock 1974Asfour 1978; Garg et al. 1982). None of these studiesincluded populations and test conditions appropriate tothe development of generalized models of driver reachAll noted that adding fixed-length torso belts decreasedfunctional reach, but these data are not applicable tobelts with modern retractor systems.

An important limitation of previous reach studiesincluding the SAE study, is that the results are presentedin tabular form. None of the publications presents ageneral model of reach capability that includesconfigurability for the characteristics of the populationThe J287 model includes configurability for taskconstraints (vehicle geometry), but only to direct thereader to a particular table. As recently as 2000, reachdata are still being presented in tabular form withouconfigurability for population anthropometry (Senguptaand Das 2000).

In early studies, the authors could assume that readersdid not have access to computers that could rapidlycalculate and render surfaces from mathematicaexpressions. Now, however, mathematical expressionsof study outcomes have more utility than tabular databecause they can be more reliably and easily transferredto computer-aided-design (CAD) systems. Theexpression of the results of a reach study inmathematical terms also allows the modeler to includethe effects of anthropometric and task factors withouthaving to add multiple tables of numerical values to apublication.


6/12

4

An Integrated Approach to Reach Capability andDifficulty

This paper outlines a new approach to modeling driverreach that has several unique features intended toovercome some of the limitations of existing models:

1. Data are collected using a computerized targetpositioning system that improves the efficiency andaccuracy of data collection over previous methods.

2. The model addresses reach difficulty and representsthe maximum attainable reach as a maximallydifficult reach. Population distributions for reach-difficulty ratings within a large volume are predicted.

3. The model is explicitly designed to be compatiblewith contemporary vehicle design practices,including new seat position models (Flannagan et al.1998) and the new SAE eyellipse (Manary et al.1998). Like those models, the new reach model isconfigurable for driver anthropometry distributionsand gender mix. The seat position model is an

integral part of the reach model.

4. The model is configurable for package dimensions(seat height, steering wheel position, etc.) throughthe integration with seat position model.

The data collection and modeling approach wasevaluated in a pilot study of eleven men and women.This paper presents the data-collection and modelingmethodologies along with sample results.

METHODS

Computer-Controlled Target Positioning Apparatus

Previous studies have measured participants maximumreach capability by asking them to grasp and push rods(Kennedy 1964; Hammond and Roe 1972). Because acentral goal of the current research is to recordsubjective ratings of submaximal reach difficulty, adifferent approach is required. The test seat is mountedon a motorized, rotating platform, as shown in Figure 4.A push-button target is located on a motorized apparatusthat can move vertically and horizontally. The angle ofthe button-mounting box can also be rotated around a

horizontal axis. By rotating the seat platform andadjusting the horizontal and vertical target position, thetarget can be placed anywhere within the participantsreach envelope. The entire system is under computercontrol, so that a specified target location in a seat-centered coordinate system can be obtainedautomatically.

Figure 4. Model of test facility, showing rotating seat andcomputer-controlled target-positioning apparatus.

Pilot Study

A pilot study was conducted using mockups of drivestations from a passenger car and heavy truckFigure 5 shows a participant completing a reach in the

truck mockup. Each participant was tested using 216target locations distributed throughout the right-handreach envelope. After receiving a visual signal, theparticipant performed a right-handed reach to the targetpressed the button for two seconds with his or her indexfinger, and returned to the home position. The participantrated the difficulty of the reach by moving a labeledslider on the steering wheel. The participants wereinstructed to rate reaches that were at the edge of theireach capability as 10, with submaximal reaches ratedon a continuous scale from 1 to 10. Targets that wereunreachable were scored as 11.

Figure 5. Participant in the pilot study, showing truck driverstation mockup (steering wheel, seat, pedal plane fixture) andtarget positioning apparatus. The participant is wearing targetsused to track motions.

A target location matrix was constructed with targelocations on six radial planes and five vector directionswith respect to horizontal. Figure 6 shows the samplingplanes with respect to the seat H-point and centerlineThe target locations were scaled using initiameasurements of each participants maximum verticallateral, and forward reach. The scaling was designed to


7/12

5

place about 15 percent of the reach target locationsbeyond the participants maximum. Target locationswere concentrated in the outer regions of the reachenvelope where the reach difficulty was expected tochange more rapidly with increasing distance from theH-point. Because the steering wheel interfered withforward reaches, the origin for the sampling vectors onthe -30, 0, and 30-degree planes (see top view in Figure6) was at shoulder height, rather than at H-point height.Figure 7 shows the target locations as a mesh along withthe exclusion zone around the participant.

Figure 6. Target location vectors for pilot study. Angles indegrees.

Figure 7. Target locations as a mesh, showing the exclusionzone surrounding the participant.

RESULTS

Figures 8 and 9 show ratings data for one participant inthe truck mockup by sampling plane and radial distanceAs expected, reach difficulty was a strong function ofradius, with the rate of increase in difficulty with radiusincreasing near the maximum reach. Reaches on the120-degree plane (reaching to the rear of H-point) weremore difficult than those on other planes. Steering-wheel interference caused a steep difficulty gradient onthe forward reaches, with a moderately difficult reachbecoming impossible with only a small increase in radiusdue to torso contact with the steering wheel.

A preliminary examination of the data suggested that asimple linear function with an exponential term couldmodel the shapes of the radially organized data withreasonable fidelity. The data for each radial vector werefit with a function of the form

D(r) = a + b r + Abs[c] Exp[r / d] [1

where D(r) is the difficulty rating at a particular radiaposition r and a, b, c, and d are constants. For thecurrent analysis, d was set to a constant value of 250and a, b, and c were fit by a least-squares method.

Figure 10 shows the fits (dark lines) overlaid on the datafrom Figure 9. Preliminary examination of the residualssuggests that the errors are normally distributed with astandard deviation of about 0.5 to 0.75 ratings pointsacross subjects. The suitability of equation 1 forepresenting the form of the radial difficulty function wasassessed by examining the residuals across thesampling vectors. The mean and standard deviation othe residuals has not been found to be related to eithe

the sampling vector direction or radius, suggesting thathe form of equation 1 is adequate.

A three-dimensional analog of equation 1 in sphericacoordinates can be used to fit the data from aparticipant. The resulting function can be solved for aparticular ratings level to obtain isorating surfaces onwhich the reach difficulty is uniform. Figure 11 showsisorating surfaces for one participant.


8/12

6

0 500 1000 1500

0

500

1000

1500Radial Plane: 60 degrees

8

77

3

6

4

4

6

725

2

9

4

8

6 8

6

2

4

10

9

2

7

5

91

9

5

9

10

9

2

9

2

6

0 500 1000 1500

0

500

1000


53

4

7

8

2

6

2

9

9

7

6

9

2

3

5

53710

10

6

9

8

3

4

3 5

10

8

2

9

2

5

8

0 500 1000 1500

0

500

1000


54

6

9

8

4

8

7

5

7

5

10

9

8

10

4

33

3

10

11

7

6 8

6

115

7

11

8

10

698

2

0 500 1000 1500

0

500

1000

1500Radial Plane: - 30 degrees

1

9

2

9

5

5

3

34

6

10

6

7

8

9

3

1

9

4

5

6

8

910

6

9

11

10

2 9

4

7

10

2

4

4

0 500 1000 1500

0

500

1000


5

89

8

4

6

10

3

1

7

10

5

3

2 9

1

2

8

8

6

7

9

11

5

10

7

10

34

11

1

4

9

2

8

5

0 500 1000 1500

0

500

1000


5

6

1 8

4

5

8

93

3

8

4

10

5

1

75

6

10

119

7

1

6

2

7

10

11

5

5

7

8

4

3

10

11

Figure 8. Reach difficulty ratings for one subject in the truck mockup. Radial planes are labeled as in Figure 6.Origin is at seat H-point on seat centerline; axes in millimeters.

500 1000 1500

1

2

34

5

6

7

8

9

10


500 1000 1500

1

2

34

5

6

7

8

9

10


500 1000 1500

1

2

34

5

6

7

8

9

10


500 1000 1500

1

2

3

4

5

6

7

8

9

10


500 1000 1500

1

2

3

4

5

6

7

8

9

10


500 1000 1500

1

2

3

4

5

6

7

8

9

10


Figure 9. Reach difficulty ratings one participant in the truck mockup. Radial planes are labeled as in Figure 6. Origin is at seat H-point on seat centerline; axes in millimeters. Each line connects data from a single sampling vector. Vectors closer to vertical areshown with longer dashes. Solid line shows data from the most-vertical vector in the plane.


9/12

7

500 1000 1500

1

2

3

4

5

6

7

8

9

10


500 1000 1500

1

2

3

4

5

6

7

8

9

10


500 1000 1500

1

2

3

4

5

6

7

8

9

10


500 1000 1500

1

2

3

4

5

6

7

8

9

10


500 1000 1500

1

2

3

4

5

6

7

8

9

10


500 1000 1500

1

2

3

4

5

6

7

8

9

10


Figure 10. Reach difficulty data from Figure 9 with overlaid with fitted functions.

Figure 11. Isorating surfaces for one participant.

The parameters of the models fitted to individualparticipants data were analyzed with respect toanthropometric variables and task conditions (truck vs.car seat, for example). The resulting relationships canbe used to predict the reach difficulty function for anindividual. The data from the pilot study show that, afteraccounting for overall scale, reach difficulty functions aresimilar across participants. Figure 12 shows isoratingsurfaces for rating = 8 for five participants in the truckseat.

Figure 12. Surface for rating = 8 for five participants in thetruck seat.

Creating useful population models from the findings fromindividual participants requires a more complicatedmodeling procedure. Figure 13 illustrates the approachschematically. Experimental data are analyzed todetermine (1) the structure of individuals reach difficultyfunctions and (2) the relationships between theparameters of these functions and anthropometric andtask variables. Since these models will not account fo


10/12

8

all of the variance across individuals, the residualvariance will be noted for use in subsequent modeling.In predicting reach functions for a particular vehicle, theprocess begins with the distributions of anthropometricvariables for the target driving population (usually staturedistributions and gender mix). The anthropometricinformation is used to generate population distributionsof reach difficulty functions with respect to the seat H-point. Depending on the final form of the models, it maybe possible to do this in closed form, but, more likely,this step will be performed numerically by simulating alarge number of drivers, predicting their reach difficultyfunctions from anthropometric variables, and includingthe random residual variance in the reach difficultyparameters.

The next step is to apply the appropriate distribution ofseat positions. In previous research (Flannagan et al.1998), a model that predicts the distributions of driver-selected seat position from vehicle and anthropometricfactors has been developed. When applied using anumerical simulation approach, the residual variance inthis model can also be incorporated. Each simulated

reach difficulty function from the previous step,representing one driver, is positioned using the predictedseat position from the seating accommodation model.The result is a population of reach difficulty functionspositioned within the vehicle package space. Thesefunctions can then be processed in a number of ways toperform design analyses.

Perhaps the easiest way to apply the reach difficultyfunctions is to calculate isorating surfaces analogous tothe familiar surfaces in J287. For example, the reachdifficulty functions can be solved for the rating = 10.5,which is midway between unreachable (rating = 11) and

the maximum-difficulty attainable reach (rating = 10).This isosurface is equivalent to the maximum reachenvelope for an individual. Computing across asimulated population, the surface in package space towhich, for example, 95 percent of the selectedpopulation can reach, can be computed. A model similarto J287 could be constructed by choosing a fewpopulations of interest and solving for a set of typicalsurfaces (say, 95 percent of the population rating < 7).The positioning of these surfaces in package spacewould be determined by a linear equation of packagevariables, most likely a constant offset from the medianpopulation seat position given by Flannagan et al.

(1998).

For more sophisticated analyses, the simulatedpopulation distributions of reach difficulty would beretained. In one potential application, the difficulty ofreaching to two alternative locations for an in-vehicleinformation device could be evaluated by computing thepopulation distributions of reach difficulty for eachlocation. An overlay of the cumulative ratingsdistributions for the two locations would show clearly theextent to which a change in the control position wouldaffect the difficulty of reach to the control.

Figure 13. Schematic of the development and application ofreach difficulty models.

DISCUSSION

This paper presents a new, integrated approach tomeasuring and modeling reach difficulty and capabilitythat is much more complicated than any previousmethod. Yet, the additional complexity is needed toprovide human-factors practitioners with the predictivepower needed to make quantitative design decisionsThe data collected in previous studies could have beenused to generate more complete and flexible models oreach capability, but the computing power needed toapply such models routinely became available onlyrecently. The current approach is practical only becausevehicle design is now conducted using computesystems easily capable of making the neededcalculations in real time.

One potential criticism of the current approach is that themethod does not determine the maximum reachcapability of drivers with the precision of previousapproaches, in which every data point is a maximumreach. However, the pilot study revealed that, in therelatively unrestricted conditions typical of modernvehicles, maximum reach varies considerably from triato trial. That is, maximum reach is a probabilisticconcept that should be modeled as such. Moreimportantly, true maximum reach for drivers using


11/12

9

modern seat belts is not of substantial interest fordesign. The maximum right-hand reach of a typical U.S.driver extends substantially across vehicle and yet nodesigner would consider putting controls in the outerregions of the reach envelope. Current practice implicitlyacknowledges this fact by using data gathered withfixed-length torso belts (J287) or by exercising figuremodels without moving the torso. As noted above,these approximations do not provide accurate guidancefor locating controls.

The need for a new model of reach could be questionedon the grounds that there are few complaints regardingcontrol reach in modern passenger cars and light trucks.While this may be the case, there are a number ofreasons to believe that a new model of the typeproposed is needed. First, the number and complexityof in-vehicle information systems is expected to increaserapidly, particularly in heavy trucks. Heavy trucksalready have many more controls than typical passengercars, and in-vehicle navigation and information systemsare adding to the control-design challenge. Second,accommodating an aging population will necessitate

increasing the size of controls on already crowdedinstrument panels, pushing controls farther from theeasily reached zones used currently. The currentmodels are not useful for quantifying the implications ofthese changes on the difficulty drivers will have inreaching the controls. The new models are alsoexpected to find application for the design of futuremilitary vehicles in which a large number of computer-based controls are to be accessed by the driver.

The new modeling approach has some limitations thatcould be addressed in future work. Currently, data havebeen collected only for push-button reaches. In the near

term, the models will be extended to other types of reach(three-finger grasp, for example) using the methodsapplied in other studies (e.g., Hammond and Roe 1972)by adding or subtracting constant offsets to convertbetween different target types. The models also do notaccount for the effects of different seat designs, althoughthe test seats included typical bolsters. Seat-restrictioneffects would be most apparent in lateral reaches.

The experimental method requires the participant to viewthe target during the reach and to hold the reach for twoseconds. During actual driving, the visual eccentricity ofa target relative to the road ahead may impose

additional difficulty not accounted for by the currentmodels. For example, two reaches that produced thesame difficulty rating in the laboratory experiment mightbe perceived as having different difficulty if they requiredifferent amounts of glance deviation from the primarydriving task. A related limitation is that these models donot account for the potential effects of reaches on theperformance of the primary task. In evaluating a controllocation, the effects of alternative locations on drivingperformance is clearly of interest.

Any empirical approach to modeling reach capability anddifficulty will always be limited by the extent to which thestudy population represents the target population. Oneparticular concern for reach difficulty modeling isaccurately capturing the effects of age. In studiesintended to be applicable to cars and light trucks used bythe general population, an effort to recruit older driverswith a range of age-associated functional limitations wilbe necessary.

The new methods developed in the pilot study are nowbeing applied in the heavy truck environment. Datacollection using these methods has recently beencompleted with 65 people, including people withexperience driving heavy trucks and buses. These datawill provide the opportunity to construct and test acomplete version of the model.

ACKNOWLEDGEMENTS

This research was sponsored by the University ofMichigan Automotive Research Center and by thepartners of the Human Motion Simulation (HUMOSIM)program at the University of Michigan. HUMOSIMpartners include DaimlerChrysler, Ford, General Motors,International Truck and Engine, United States PostalService, U.S. Army Tank-Automotive and ArmamentsCommand, and Lockheed Martin.

REFERENCES

Asfour, S.S., Ayoub, M.M., Mital, A., Bethea, N.J. (1978)Reach profiles for males and females under restrainedand unrestrained conditions. Proc. Human FactorsSociety Annual Meeting. Human Factors andErgonomics Society, Santa Monica, CA.

Bullock, M.I. (1974). The determination of functional armreach boundaries for operation of manual controlsErgonomics. 17(3), p. 375-388.

Chaffin, D.B. (2001). Digital Human Modeling foVehicle and Workplace Design. SAE InternationalWarrendale, PA.

Flannagan, C.A.C., Manary, M.A., Schneider, L.W., andReed, M.P. (1998). An Improved SeatingAccommodation Model with Application to Different UsePopulations. Technical Paper 980651. SAE

Transactions: Journal of Passenger Cars, Vol. 107.

Garg, A., Bakken, G.M., Saxena, U. (1982). Effects oseat belts and shoulder harnesses on functional armreach. Human Factors, 24(3), p. 367-372.

Hammond and Roe (1972). SAE controls reach studyTechnical Paper 720199. Society of AutomotiveEngineers, Inc., Warrendale, PA.


12/12

10

Haslegrave, C.M. (1970). Study of reach to car controlswhile restrained by a lap and diagonal belt. MIRABulletin, No. 1., p. 11-15.

Kennedy, K. W. (1964). Reach capability of the USAFpopulation. Phase I-the outer boundaries of grasping-reach envelopes for the shirt-sleeved, seated operator.Report No. AMRL-TDR-64-59. Air Force SystemsCommand, Aerospace Medical Research Laboratory,Wright-Patterson AFB, Ohio.

Manary, M.A., Flannagan, C.A.C., Reed, M.P., andSchneider, L.W. (1998). Development of an improveddriver eye position model. Technical Paper 980856.SAE Transactions: Journal of Passenger Cars, Vol. 107.

Porter, J.M., Case, K., Freer, M.T., and Bonney, M.C.(1993). Computer-aided ergonomics design ofautomobiles. In Automotive Ergonomics, ed. B. Peacockand W. Karwowski, 43-77. London: Taylor and Francis.

Roe, R.W. (1993). Occupant packaging. In AutomotiveErgonomics, ed. B. Peacock and W. Karwowski, 11-42.Taylor and Francis, London.

SAE (2002). SAE Handbook. SAE International,Warrendale, PA.

Sengupta, A.K., and Das, B. (2000). Maximum reachenvelope for the seated and standing male and femalefor industrial workstation design. Ergonomics34(9):1390-1404.

Stoudt, H. W., Crowley, T. J., McFarland, R. A., Ryan,A., Gruber, B., and Ray, C. (1970). Static and dynamicmeasurements of motor vehicle drivers. Report Number

DOT/HS 800 261. National Highway Traffic SafetyAdministration, Washington, D.C.

a new approach to modeling driver reach

Documents