Analysing vehicle dynamics objective and subjective testing in winter conditions

(1)

http://www.diva-portal.org

Postprint

This is the accepted version of a paper presented at 24th International Symposium on Dynamics of Vehicles on Roads and Tracks.

Citation for the original published paper:

Gil Gómez, G., Nybacka, M., Bakker, E., Drugge, L. (2015)

Analysing vehicle dynamics objective and subjective testing in winter conditions.

In: Taylor & Francis

http://dx.doi.org/10.1201/b21185-81

N.B. When citing this work, cite the original published paper.

This is an Accepted Manuscript of an article published by Taylor & Francis in The Dynamics of Vehicles on Roads and Tracks: Proceedings of the 24th Symposium of the International

Association for Vehicle System Dynamics (IAVSD 2015) on May 3, 2016, available online: http://

www.tandfonline.com/10.1201/b21185-81.

Permanent link to this version:

http://urn.kb.se/resolve?urn=urn:nbn:se:kth:diva-181060

(2)

1 INTRODUCTION

Currently the development and evaluation of winter vehicle handling characteristics are almost solely based on subjective assessments (SA) done by expert drivers. This is both expensive and time consuming, and therefore in conflict with the general goal of shortening project development and testing time meanwhile fulfilling more restrictive and demanding specifications.

A more effective vehicle dynamics evaluation would be to use Computer Aided Engineering (CAE) for objective testing. Although objective evaluations are well defined for high friction conditions, or summer testing, the correlations between objective metrics (OM) and SA are still under development (Chen & Crolla 1998, Harrer et al. 2006, King et al. 2002, Nybacka et al.

2014a, b). Furthermore, these objective methods are not directly applicable for low friction testing, or winter testing, because the added challenge of constantly changing surface conditions.

This causes both low signal-to-noise ratio measurements and low robustness, or repeatability, of the results. This difficult situation is increased by the fact that there is only one short winter test season per year and hemisphere, which limits the possibilities to shorten project times, as several winter test periods are normally required. All the above are motivations to understand:

− How expert drivers perform their winter SA.

− How objective tests, measurements and OM shall be defined for winter tests.

− How OM and SA correlate in winter conditions and compares with summer conditions.

During this research, a winter test expedition was performed in order to develop the methods to evaluate vehicle dynamics handling on winter conditions, see Figure 1. The goals of the performed tests were to study how winter objective testing should be performed when using steering robots. By identifying the suitability of different objective test manoeuvres, methods and/or procedures; the effects that the surface conditions have on the results and their repeatability, and if key OM are robust enough to describe the vehicle performance during changing surface

Analysing vehicle dynamics objective and subjective testing in winter conditions

G. L. Gil Gómez & E. Bakker

Vehicle Dynamics, Volvo Car Corporation, Gothenburg, Sweden M. Nybacka & L. Drugge

Dept. of Aeronautical and Vehicle Engineering, KTH Royal Institute of Technology, Stockholm, Sweden

ABSTRACT: This paper presents a test procedure developed to gather good quality data from objective and subjective testing on winter conditions. As the final goal of this test is to analyse the correlation between objective metrics and subjective assessments on winter for steering and handling, this procedure has to ensure a minimum change of the surface properties, which has a major influence on vehicle performance, during the whole test campaign. Therefore, the method presented keeps the total test time very low and allows similar vehicle configurations to be tested, objectively and subjectively, very close in time. Moreover, continuous maintenance work on the ice is performed. Reference vehicles are also used to monitor the changes on vehicle performance caused by weather conditions, which are inevitable. The method showed to be very effective. Initial results on objective metrics and subjective assessments are also presented.

(3)

conditions. Furthermore, it was also a goal to understand the human factors lying behind expert drivers’ winter SA. All of them are preliminary steps to start the analysis of correlations between OM and SA on winter conditions, i.e. important steps towards the vision of reducing development costs and time by using CAE in the future.

Gathering quality data in an environment that changes within hours, as it is during winter road conditions, imposes high requirements on both material and human resources, especially for objective testing, since these kinds of tests is not normally performed during such conditions. Thus, well-planned test procedures, logistics and methods were compulsory. The goal is to obtain as many subjective and objective data for a given vehicle configuration during the shortest possible time, assuring the lowest change on non-controlled test parameters. This paper focuses on presenting the methodology followed in order to ensure gathering of quality data in a short period of time to reduce effects of surface changes and at the same time making it possible to register these surface changes.

Figure 1. Subjective and objective tests during winter conditions driven on ice with a thin compact layer of snow. Left: Test tracks and vehicles. Centre: Steering robots for OM. Right: Test drivers’ SA.

2 MATERIAL AND METHODS 2.1 Driver selection

Previous studies, searching for correlations between OM and SA indicated that for this kind of research, highly skilled and experience test drivers should be used for gathering the subjective data. Farrer (1993) recommended that this type of test should be run by expert drivers, given the highly technical nature of the experiments involved. Dang et al. (2014) indicated that ex- perts, compared to normal drivers, are able to understand each question thoroughly and to con- nect them with the different sensations felt during the drive. Harrer et al. (2006) explained that many of the previous investigations lacked consistency in their evaluations due to the use of randomly chosen customers, from different age groups and with general difficulties in interpret- ing the descriptions from SA. Harrer et al. (2006) therefore recommended the use of only expe- rience vehicle dynamics engineers with advanced driving training for SA. Chen & Crolla (1998) suggested using expert drivers, as they are able to perform manoeuvres that are more advanced and still have enough mental capacity left for evaluations. In similar studies on driveability, Schoeggl & Ramschak (2000) pointed out that the use of expert drivers leads to a lower SA variation, a key factor for finding good correlations between OM and SA.

All these studies were done for summer conditions. Winter conditions are even more demanding for the drivers. Expert drivers were therefore selected for this new study. Five expert drivers participated in evaluating the complete test vehicles fleet. These five drivers were all male, their age span and years of experience as test drivers, in form of mean and standard deviation, are presented in Table 1.

Table 1. Detailed information about the test drivers involved in collecting SA.

____________________________________________________________________________

No. of Age Experience

______________________ __________________

drivers Mean Std.Dev. Mean Std.Dev. Gender

____________________________________________________________________________

5 44 13 13 12 Male

____________________________________________________________________________

(4)

Figure 2. Four similar vehicles (first row) make it possible to run quasi-simultaneous subjective and objective steering and handling testing of configuration vehicles (first and third column), at the same time that unmodified reference vehicles (second and fourth column) are used, for both subjective and objective testing, as indicators of the changes of the surface conditions. Each row represents a test series, i.e. objective and subjective testing for configurations 1 to 4 in this example (10 configurations in the complete test).

2.2 Vehicle set-up selection

A way of lowering the negative effect of the variation of friction is using a reference vehicle;

this applies for both SA and OM. The reference vehicle works as an indicator of the changing road conditions. This implies that the reference and the test vehicle need to be evaluated in the shortest possible time, both objectively for all test manoeuvres and subjectively by all drivers.

In order to accomplish this, the following test procedure was developed: Four identical vehicles were used. Two would be kept unchanged as reference vehicles, for objective and subjective testing respectively. The remaining two vehicles would be changed into 10 different chassis configurations, to serve each of them as objective and subjective test vehicles, as shown in the example in Figure 2, for four configurations.

The test configurations were selected from an orthogonal factorial test design matrix, elimi- nating uninteresting test configurations; the allowed modified parameters were connection- disconnection of front and/or back antiroll bars, standard or sport dampers front or back; loaded or unloaded rear with 120 kg; and their combinations. As tests were blind, a configuration similar to the reference vehicle was included in order to test drivers SA repeatability, see Table 2.

The order of the vehicle configurations was decided in order to optimized workshop time, i.e.

the most time demanding changes (dampers and load) were done between configurations 3 to 4 and 8 to 9, coinciding with test drivers lunch; and between configurations 5 to 6, between the end of first test day and beginning of the second test day.

Table 2. All vehicle configurations tested.

_______________________________________________________________________________

Vehicle Front axle Rear axle Rear

______________________ __________________

configuration ARB* Damper ARB* Damper loaded

_______________________________________________________________________________

Vehicle 1 Yes Softer No Softer No Vehicle 2 No Softer Yes Softer No Vehicle 3** Yes Softer Yes Softer No Vehicle 4 No Softer Yes Stiffer Yes Vehicle 5 Yes Softer Yes Stiffer Yes Vehicle 6 Yes Stiffer No Stiffer Yes Vehicle 7 No Stiffer Yes Softer No Vehicle 8 Yes Stiffer Yes Softer No Vehicle 9 Yes Softer No Stiffer No Vehicle 10 No Softer Yes Stiffer No

_______________________________________________________________________________

*ARB = Antiroll bar. **Vehicle 3 similar to the reference vehicle

(5)

2.3 Test procedure and manoeuvres The chosen objective manoeuvres were:

− Constant Radius circle (CR),

− Frequency Response (FR),

− Sine With Dwell (SWD), modified with the dwell in the first half of the sinus,

− Throttle Release In a Turn (TRIT),

in order to gather data from the steady-state response, the frequency response, and the transitory response to steering inputs and to powertrain inputs, respectively. All manoeuvres were slightly modified from the standard tests, basically by reducing speed and steering wheel angle (SWA) amplitudes, in order to adapt the manoeuvres to the space available on the test tracks, with safety margins, and to the lower lateral grip available on ice, with respect to dry tarmac conditions for which the standard tests are designed. Moreover, the sine with dwell manoeuvre was modified, switching the dwell to the first half period of the sinus, i.e. a kind of dwell with sine. The motivation for doing so was that it had been observed that expert test drivers repeated that manoeuvre constantly when evaluating the vehicles. The idea was therefore to try to reproduce that manoeuvre and to try to objectively measure what test drivers were subjectively evaluating.

In order to study repeatability, and to increase the statistical significance of the data, objective tests were repeated several times: CR three times, SWD four times for nine increasing- SWA-amplitudes steps, TRIT six times, FR four times. In this initial study, the response of the vehicles was considered symmetric and tests were therefore executed in only one SWA direction.

Pfeffer et al. (2008) indicates that “performing open-loop manoeuvres in conjunction with a steering robot greatly increases the reproducibility and eliminates influences caused by the driver”, because the frequency and amplitude of the test driver's manoeuvres do not remain constant. Their results show that the scatter of measurement data, i.e. the spread of objective parameters in test repetitions, is reduced when using a steering robot in comparison to a test driver in average from 13.5% to 4.0% in the weave test and from 13.3% to 2.1% in the step in- put test, i.e. the scatter is reduced in the order of 70% and 84% respectively. Therefore, in these tests the steering robots were chosen for objective testing, see Figure 1-centre.

Observations of SA on low friction by expert drivers, with very high driving skills, indicate that subjective testing are done at very high levels of tyres lateral slip and body side slip. Driv- ers’ corrections to the vehicle, via SWA and throttle and braking pedals are constants. There- fore, close-loop manoeuvres might be even less suitable for winter testing and the effect of the driver would probably be even greater than in Pfeffer’s summer conditions analysis. Conse- quently, all the selected manoeuvres, except CR, are open loop manoeuvres. Thus, the effect of the robot controller will have to be analysed for the CR test, which is done by path following.

The above also indicates a possible drawback in using steering robots with open-loop manoeuvres on winter testing: high slip and loss of grip are expected and the steering robot, in open loop will not counteract it. Furthermore, after a loss of vehicle control because of the low friction available, a larger braking distance is needed. Altogether, this means that higher safety margins are required: large frozen and prepared surfaces are required in order to perform tests at the same level of e.g. side slip that test drivers reach. When large areas are not possible lower levels are needed, i.e. lower SWA and speed would be required. That is, the tests would not be as close to the limits as for subjective testing. This would unfortunately reduce the signal-to- noise ratio in OM between different vehicles. As noted by Krüger et al. (2000) the handling dif- ferences due to small design changes are noticeable or measurable only in driving conditions with high lateral accelerations, where high lateral acceleration can be interpreted as close to the handling limit conditions. In winter conditions, this happens at lower lateral acceleration.

The tests were run on frozen lakes, in one of the winter proving grounds in the north of Swe- den. The test surface was therefore ice with a thin layer of compact snow, see Figure 1. Two different tracks had to be used in these tests. The ideal condition would have been to use only one track in order to ensure the ice to be always the “same” in all measurements. Although most of the test were programmed for the larger of the handling tracks, which also included a handling area, the CR test was the exception, as it had to be executed on the circular handling area.

Each test driver received a copy of the test plan and a checklist with the test procedure. Alt- hough they drove freely, as it is normally done, the protocol included some limitations: three

(6)

laps per vehicle, with option to a fourth lap if required, in this way test time was controlled.

They had also to drive always first 3-4 laps with the subjective reference vehicle (2^nd column in Fig. 2) and immediately after the subjective configuration test vehicle (1^st column in Fig. 2). In order to reduce testing time, two test drivers would drive these two vehicles simultaneously.

Straightaway, by using two steering robots and two robot operators, objective tests were run, simultaneously for the objective reference (4^th column in Fig. 2) and for the objective configuration vehicles (3^rd column in Fig. 2). In this way, the five subjective and all objective tests (49) for each reference and configuration vehicle were run altogether in less than 90 minutes.

In order to optimize the available time, driving time between the different test tracks and the workshop, as well as the required workshop time, were measured and included in the plan. Be- cause of safety reasons, objective testing was not allowed simultaneously in the same track than subjective testing. In order to make the best use of time, meanwhile subjective testing was performed, the objective test vehicles were used for the CR test, in another test area, and for workshop time, i.e. changing the objective configuration vehicle. In a similar way objective testing time was measured and included in the test-plan and objective testing was done in parallel with changing the subjective configuration vehicle. Drivers’ and test operators’ breaks and lunches were scheduled. A van allowed comfort to the test drivers while waiting by the test track as well as optimized the transport of drivers from the test tracks to the office-buildings. A half-day plan can be seen in Table 3, the sequence to follow is similar for the rest of the test period.

The changes on the surface properties is not only caused by weather conditions but also by the test itself. Tyre lateral slip tends to accumulate small amounts of snow and create small snow dunes. Surfaces that are constantly driven through, e.g. corners apexes, end up with no snow and with the ice being polished. The consequence is therefore a heterogeneous road surface with low friction areas. In order to ensure the quality of the ice, a support vehicle worked constantly distributing the snow and scratching the ice.

Table 3. Test plan sequence for half a day, the rest of the test day follows the same concept but it is not shown in order to improve readability.

10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 160 170 180 190 200 210 220 230 240

Driver D1 D2 D3 D4 D5 D1 D2 D3 D4 D5 D1 D2 D3 D4 D5

Vehicle Ref Ref Ref Ref Ref Ref Ref Ref Ref Ref Ref Ref Ref Ref Ref

Driver D1 D2 D3 D4 D5 D1 D2 D3 D4 D5 D1 D2 D3 D4 D5

Vehicle V1 V1 V1 V1 V1 V2 V2 V2 V2 V2 V3 V3 V3 V3 V3

Obj.Test FR SWD TRIT FR SWD TRIT

Vehicle Ref Ref Ref Ref Ref Ref

Obj.Test FR SWD TRIT FR SWD TRIT

Vehicle V1 V1 V1 V2 V2 V2

Track 2 CR V1 Ref V2 Ref

V1 V2 V1 V2 V2 V3 V2 V3

V1 V2 V1 V2 V2 V3 V2 V2

4

Break BREAK COFFEE BREAK COFFEE BREAK LUNCH

Transport time Workshop time Time in minutes

Test track 1

Time in hours 1 2 3

D1, D2, … = Driver number. V1, V2, … = Vehicle configuration number. Ref = Reference vehicle.

2.4 Subjective questionnaire

Another important factor, as explained in Gil Gómez et al. (2015), is the selection of the SA questions and of the rating scale to answer them. Käppler et al. (1992) based on a review of test procedures and scales, explained that test standards and scale deficiencies may lead to inade- quate results. The SA questionnaire was therefore modified with respect to the one used for summer SA in previous research done by the authors (Nybacka et al. 2014a, b, Gil Gómez et al.

2015). The lesson learned in those studies was applied, as well as some modifications in order to study their effect on the SA, resulting in the SA questionnaire presented on Appendix A. The main characteristics of the questionnaire are summarized below:

− It IS a one-page questionnaire, which facilitate the drivers’ work as it allows them having constantly an overview of all the questions.

− Winter test questions: the questions are adapted to the attributes assessed during winter conditions, which differ from those assessed during normal summer testing.

(7)

− Weather & temperature: are to be logged in each test, as changes in surface conditions are especially important on winter and therefore keeping track of conditions during testing is even more important in winter.

− Shorter questionnaire: with almost one third of the number of questions compared to the previous questionnaire (13 vs 36 questions), this to reduce the mental workload of the drivers, improving their focus on the characteristics to be felt and facilitating remembering the characteristics of the reference vehicle.

− Different SA levels with their different rating scales are used, see Gil Gómez et al. (2015):

Questions 1-6 are evaluation SA (drivers give a rating to an attribute); instead in questions 7-13, the drivers, working as sensors, also try to measure the attribute. However, the tree structure of Nybacka et al. (2014a, b) and Gil Gómez et al. (2015) is avoided, with all the questions presented without any hierarchy at the same level. In this way, no previous relation between SA questions is introduced to the driver via the questionnaire. Consequently, the relation, e.g. multicollinearity, of different questions can be studied avoiding the bias that could have been introduced in the old questionnaire.

The SA question has also been upgraded with a series of new features, see Table 4:

− A comment field is available to complement the numerical information.

− A “Not sure” field allows the drivers to indicate when they are not sure about the assessment. Still they can guess giving a rate, as some analysis methods need always an answer;

however, in that way low confidence is given to that rating. E.g. Roll control was crossed several times, and the comment field indicated that on winter testing low lateral accelerations are reached and that those low levels do not allow a proper evaluation of this characteristic. Note that this question was introduced because of the different vehicle configurations were obtained by removing anti roll bars, see Table 2.

− An evaluation (rating) field was included even for level 5 questions (8.5 of 10 in the example in Table 4). In that way it is not only possible to indicate how the attribute is felt, i.e. es- timation or ‘X’ in the example, but also how appropriate it is.

− A desired level 5 field was included in order to understand the direction of improvement.

That is obtained by the desired field ‘O’ which indicates where the test driver would like the

‘value’ of the attribute to be. The difference between the ‘O’ and the ‘X’ indicates the required level of change, meanwhile the rating (8.5) indicates the expected effect of imple- menting the change. In the example in Table 4, the vehicle is felt little-to-medium understeer, the driver would prefer it to be more neutral (still in the understeer side), however the change might not be critical as the rating in that attribute, 8.5, is already high.

− Extra complementary questions: “What is the character of the car?” In addition, general feedback comments: “What is missing, other assessment criteria, what should we change in the questionnaire?” were included in order to complement the information from numerical ratings and to identify potential improvements in the new procedure, respectively.

3 RESULTS & DISCUSSION 3.1 The test procedure in numbers

Section 2.3 and Table 3 presented the schedule followed in order to optimize the test, i.e. in order to execute as many tests as possible in the shortest amount of time. Remember that the goal is not only to reduce time/cost but also to ensure the minimal change in the properties of the test track surface, i.e. the ice, which means allowing the shortest possible time between tests.

Table 5 presents a summary of what the schedule presented in Table 3 allowed to do in a test campaign of two working days (approximately 16 hours), with four vehicles, two steering robots, two robot operators, five expert drivers and one mechanic, all of them with a copy of the schedule and the support of two of the researchers as test coordinators.

Table 4. Detail of the subjective assessment questionnaire.

Not

Sure Comments

Desired Understeered O Neutral Oversteered

Current -4 -3 -2 X -1 0 1 2 3 4

7 Balance @ Constant Throttle 8.5

Grade (X) first what you feel in "Current field" then grade (O) what you think will give the optimal grade in the "Desired field"

(8)

Table 5. Overview of the results of the two days test campaign in numbers.

________________________________________________________________________________________________________

∼1000 objective tests (3 CR + 36 SWD + 6 TRIT + 4 FR) x (10 configurations + 10 references) 100 subjective tests 5 drivers x 20 vehicles

∼1000 minutes test driving 5 drivers x 20 vehicles x 10 minutes/(driver x vehicle)

∼1800 minutes transport 10 minutes/way x 2 ways x 10 times x 9 people*

~ 550 minutes workshop 25 minutes/modification x 2 vehicles** x 11 times

>4000 km driving >1000 km logged per vehicle x 4 vehicles

________________________________________________________________________________________________________

* 9 people = 5 drivers + 2 robot operators + 2 test coordinators

** 2 vehicles = subjective testing configuration vehicle + objective testing configuration vehicle

3.2 Preliminary results from winter testing

The aim of this paper was to present a methodology designed to gather good quality data in order to affront the task of study the correlation between OM and SA in winter conditions. How- ever, some preliminary results of the winter test work are also presented in this section.

Despite the low signal-to-noise ratio characteristic of winter conditions, the developed procedure allows the objective evaluation of vehicle handling on winter condition. Figure 3 shows an example where differences in repeated measures of the same vehicle configuration can be clearly distinguished from the repeated measures of other vehicle configurations, each configuration plotted in different colours versus each repetition plotted in different line shapes. It also shows how vehicle 3 (dark grey lines, 2^nd lowest) are difficult to differentiate from the reference vehicle (lowest black lines) as they have very similar OM, which was desired, as both vehicles shared the same configuration.

Figure 3. OM example – slip angle vs lateral acceleration. Vehicle 1: top black lines. Vehicle 2: light grey lines under vehicle 1. Reference vehicle: bottom black lines. Vehicle 3: dark grey lines between vehicle 2 and reference vehicle. Only vehicle 1 to 3 and reference plotted.

Figure 4 presents the SA rating distributions for the vehicles shown in Figure 3. These SA make it possible to perform a similar study as in Gil Gómez et al. (2015), but for winter condi- tions: i) classify the vehicles from best to worst (e.g. from best to worst, vehicle 3 or reference – vehicle 2 – vehicle 1), ii) investigate the correlations OM to SA (relation between Fig. 3 and 4), iii) study the repeatability of SA via a blind test (e.g. vehicle 3 versus reference in Fig.4), iv) understand the use of the rating scale, v) understand how drivers set ratings.

Drivers rating tendency, rating repeatability and the use of the scale present similar results to those obtained during summer testing and studied at Gil Gómez et al. (2015).

(9)

2 4 6 8 10 0

10 20 30 40 50

Reference vehicle

Distribution of ratings in %

Mean = 7.6077 Standard Dev. = 0.41685

2 4 6 8 10

0 10 20 30

Vehicle 1

2 4 6 8 10

0 5 10 15 20 25

Vehicle 2

2 4 6 8 10

0 5 10 15 20 25

Vehicle 3

Figure 4. SA rating distribution (histogram) per vehicle. Only vehicle 1 to 3 and reference plotted.

Figure 5 is another rating distribution, but this time per driver. Here it is also possible to do a similar analysis to the one presented in Gil Gómez et al. (2015). However, the most interesting part here is that comparing this plot with its analogous (in Gil Gómez et al. 2015) it is possible to see that drivers 2, 4 and 5 (respectively drivers 2, 3 and 4 in Gil Gómez et al. 2015) present a quite similar tendency in their ratings distributions, with almost same mean, standard deviation, and rating tendency vs. the rest of the drivers. Meanwhile driver 1 (driver 1 in Gil Gómez et al.

2015) keeps his usage of the rating scale, with approximately the same standard deviation, but his mean rating has been reduced by 0.8 points, passing to be the most pessimist driver, i.e.

lowest ratings, from having been the most optimistic in the previous tests.

This is very important, because this driver had been informed about his too optimistic rating tendency, with respect to the rest of drivers, and he changed his behaviour. Which indicates that presenting to the drivers their ratings, with respect to the group, could be used as a method to calibrate test drivers in order to reduce the identified high spread on SA by this kind of studies (Chen & Crolla 1998, Gil Gómez et al. 2015).

Note that the analysis of level 5 questions would also allow to identify SA questions that are not properly felt by the drivers, e.g. roll control during winter conditions, which was already identified by some drivers’ comments and by preliminary observation of the gathered data.

2 4 6 8 10

0 5 10 15 20 25

Driver: 1 Distribution of rates in %

Mean = 6.4738; Median= 6.5 Standard Dev. = 0.83535

2 4 6 8 10

0 10 20 30

2 4 6 8 10

0 10 20 30 40

Mean = 6.9716; Median= 7 Standard Dev. = 0.66251

2 4 6 8 10

0 10 20 30

Mean = 7.0322; Median= 7 Standard Dev. = 0.83384

2 4 6 8 10

0 5 10 15 20 25

Figure 5. SA rating distribution (histogram) per driver.

4 CONCLUSIONS

In conclusion, it can be said that the test procedure developed, and presented in this paper, shows promising results for the evaluation of OM and SA and their correlations during winter conditions, which is an enabler for CAE winter testing.

When measuring OM on winter, and when gathering SA to be correlated with OM, a reference vehicle is required. That is because the drivers need to have a reference against which to set their ratings, since the low signal-to-noise ratio of vehicle performance on winter makes it difficult to distinguish if changes are produced by the new vehicle configuration or by the rapid change of the ice surface. Including a reference vehicle, measured at the same time as the configuration vehicle, allows therefore to measure and document these changes.

The method and procedure proposed here, by using four similar vehicles, two for objective testing with two steering robots and two for SA, with a reference vehicle in each group, does

(10)

not only make it possible to increase the quality of the data, it also allows to gather a huge quantity of information in a very short time (1000 objective tests and 100 subjective tests in only two working days). This efficiency is not only desired for time and cost reduction, but also to keep the changes of the ice (i.e. road conditions) as low as possible through all the vehicle configurations and all the tests.

Analysis of SA demonstrate that driver-rating tendency is similar on winter compared to summer test conditions. It has also been demonstrated that the driver tendency might be influ- enced and thus driver-rating calibration is possible.

ACKNOLEDGMENT

The authors would like to acknowledge gratefully the financial support from TRENoP (Transport Research Environment with Novel Perspectives) at KTH (the Royal Institute of Technology in Stockholm) and the funding programme FFI - Strategic Vehicle Research and Innovation [2012-04609]. The authors would also like to extend their thanks to Volvo Car Cor- poration and to all the teams: test drivers team, objective testing team, workshop team and proving ground team without whom this work would not have been possible.

REFERENCES

Chen, D. C. & Crolla, D. A. 1998. Subjective and objective measures of vehicle handling: drivers & ex- periments. Vehicle System Dynamics 29, 576-597.

Dang, J., Chen, H., Gao, B., Li, Q. & Li, M. 2014. Optimal design of on-center steering force characteris- tic based on correlations between subjective and objective evaluations. SAE Int. J. Passenger Cars – Mech. Syst. 7, 3, doi: 10.4271/2014-01-0137.

Farrer, D. 1993. An objective measurement technique for the quantification of on-centre handling quality.

In: International Congress and Exposition. Detroit, Michigan, 1-5 March 1993. Society of Automo- tive Engineers. Warrendale, Pa. doi:10.4271/930827.

Gil Gómez, G.L., Nybacka, M., Bakker, E. & Drugge, L. 2015: Findings from subjective evaluations and driver ratings of vehicle dynamics: steering and handling, Vehicle System Dynamics: International Journal of Vehicle Mechanics and Mobility, DOI:10.1080/00423114.2015.1050402

Harrer, M., Pfeffer, P. E. & Johnston, D. N. 2006. Steering feel: objective assessment of passenger cars analysis of steering feel and vehicle handling. In: FISITA World Automotive Conference 2006. Yoko- hama, Japan, 22-27 October 2006.

King, R., Crolla, D., Ash, H. & Whitehead, J. 2002. Identification of subjective-objective vehicle handling links using neural networks for the foresight vehicle. In: SAE World Congress. Detroit, Michigan, 4-7 May 2002. Society of Automotive Engineers. Warrendale, Pa. doi: 10.4271/2002-01-1126.

Käppler, W. D., Ruijs, P. A., van Randwijk, M. J. & Godthelp, J. 1992. Methodology for predicting car and truck handling assessments. In: XXIV FISITA Congress. London, 7-11 June 1992. Institution of Mechanical Engineers. London

Krüger, H., Neukum, A. & Schuller, J. 2000. A Workload Approach to the Evaluation of Vehicle Han- dling Characteristics. In: SAE World Congress. Detroit, Michigan, 6-9 March 2000. Society of Auto- motive Engineers. Warrendale, Pa. doi: 10.4271/2000-01-0170.

Pfeffer, P. E., Harrer, M. & Lin, J. 2008 Vehicle dynamics measurements. Proceedings of the Institution of Mechanical Engineers, Part D: Journal of Automobile Engineering, May 1, 2008 Vol. 222, pp.

801-813.

Nybacka, M., He, X., Su, Z., Drugge, L. & Bakker, E. 2014a. Links between subjective assessments and objective metrics for steering, and evaluation of driver ratings. Vehicle System Dynamics 52, Suppl. 1, 31-50.

Nybacka, M., He, X., Gil Gómez, G. L., Bakker, E. & Drugge, L. 2014b. Links between subjective as- sessments and objective metrics for steering. Int. J. Automotive Technology 15, 6, 893-907.

Schoeggl, P. & Ramschak, E. 2000. Vehicle Driveability Assessment using Neural Networks for Devel- opment, Calibration and Quality Tests. Society of Automotive Engineers. Warrendale, Pa. doi:

10.4271/2000-01-0702.

(11)

APPENDIX A

Figure A1. Subjective assessment questionnaire.

Date:Driver: Place: Car:R12345678910Tyre: Weather:Temp: RatingNot SureComments DesiredUndersteeredNeutralOversteered Current-4-3-2-101234 DesiredUndersteeredNeutralOOversteered Current-4-3-2-101234 DesiredUnclearClear Current12345 DesiredBrakingAccelerating Current12345 DesiredLow dampingHigh damping Current12345 DesiredLowHigh Current12345 DesiredLow responseHigh response Current1234513

11 12 Response

Attribute 1 2 3 54 6 7 8 9 10

Overall handling Predictability and concistency

iC O M S A S U B J E C T IV E E V A L U A T IO N O N I C E

Balance @ Power On/Off

Cornering Stability / Controllability on ICE Pleasure to Drive Controllability Capacity feel Yaw Rate Overshoot Yaw Damping Roll control

Connection to the Road Balance @ Constant Throttle Neutral Yaw (Centre) Position Grade (X) first what you feel in "Currentfield" then grade (O) what you think will give theoptimalgradein the "Desired field"