Driver Attention in Automatic Transmission Cars

INTRODUCTION In automatic transmission vehicles, the gear-shifting event allows the vehicle operate over a wide range of speeds and is a key component for speed control in the vehicle. When changing gears, there is often a jolt in the vehicle's movement due to a transmission setting by the manufacturer to reduce response time excessively for the gear change. Conversely, drivers might perceive an unexpected delay in acceleration when the transmission control is programed to provide drivers with a more smooth feeling. Due to this, it is required to find optimal configurations of gear shifting patterns to satisfy drivers across various conditions. However, the effort to satisfy drivers for shift quality also implies the importance of valid measurement technique to evaluate drivers' subjective perception of quality of the gear-shifting event, in forms of quantitative metric.

Historically, automobile companies have used trained drivers to assess the shift quality by rating it on a scale of 1-10 after driving the vehicles [1]. Another approach may include the

examination of variation of forward and backward acceleration data (G data) captured using acceleration sensors attached in a vehicle. Experts often use this approach on vehicle evaluation as an in-direct assessment method [2], [3]. There have also been studies to develop a useful measurement method [4]. Zhang et al. conducted a study to evaluate shift quality based on six different factors, these being shift time, acceleration, fuel economy, exhaust emission, jerk, and sound. The evaluation was applied to three different types of automobiles (average, premium luxury, and racing). They developed an artificial network to predict shift quality and validated it using the ratings of a professional driver [5].

The existing measures, however, have general limitations to assess drivers' subjective perception of shift quality. First, the results from a measurement using trained or professional drivers might not be extrapolated to the general driving population. Second, representative drivers often use a single scale assessment rating (e.g., a scale from 1-10 in which one is “extremely bad” and ten is “extremely good”) despite the fact

Measurement and Modeling of Perceived Gear Shift Quality for Automatic Transmission Vehicles

Byeong wook Jeon Hyundai Motor Group

Sang-Hwan Kim University of Michigan-Dearborn

ABSTRACT This study was conducted to develop and validate a multidimensional measure of shift quality as perceived by drivers during kick-down shift events for automatic transmission vehicles. As part of the first study, a survey was conducted among common drivers to identify primary factors used to describe subjective gear-shifting qualities. A factor analysis on the survey data revealed four semantic subdimensions. These subdimensions include responsiveness, smoothness, unperceivable, and strength. Based on the four descriptive terms, a measure with semantic scales on each subdimension was developed and used in an experiment as the second study. Twelve participants drove and evaluated five vehicles with different gear shifting patterns. Participants were asked to make kick-down events with two different driving intentions (mild vs. sporty) across three different speeds on actual roadway (local streets and highway). After each event, participants were asked to complete the rating of the four descriptive terms as well as a comprehensive rating on the gear-shifting event. Along with this subjective evaluation, changes of mechanical properties during the event also were recorded. This included shift time and delay, acceleration profile, and jerk. Consequently, a series of statistical analysis revealed the relationship between each subdimension on subjective quality and their associated mechanical properties. Models of comprehensive shift quality, based on the ratings of the subdimensions for driving intentions and speeds, were also developed. The study provided a unique framework to assess subjective driving quality related to mechanical control variables in a multidimensional way, which is applicable to improve the quality of a vehicle.

CITATION: Jeon, B. and Kim, S., "Measurement and Modeling of Perceived Gear Shift Quality for Automatic Transmission Vehicles," SAE Int. J. Passeng. Cars - Mech. Syst. 7(1):2014, doi:10.4271/2014-01-9125.

2014-01-9125 Published 05/09/2014

Copyright © 2014 SAE International doi:10.4271/2014-01-9125

saepcmech.saejournals.org

423

that it may limit the evaluation of diverse psychological aspects influencing their perceptions of shift quality. Third, despite the fact that various dimensions of perceived shift quality were successfully captured using general drivers, the results do not provide the vehicle designers with solutions to improve the gear shift quality but rather present comparisons of the quality between competing vehicles. Therefore, there exists a need for further development of a shift quality metric to take into account drivers' common psychological factors. This metric will compose the perception of shift quality with an understanding of the relationship between mechanical variables of the gear shifting and the cognitive demands of drivers' subjective perception of shift quality so that vehicle designers can manipulate the mechanical properties to improve drivers' satisfaction. It was expected to be achievable through the application of Kansei Engineering [6].

Kansei Engineering was originally used to design new products based on customers' feelings and demands. Major activities for Kansei Engineering involve: 1) capturing the customers' common feelings about the product in terms of psychological estimation; 2) identification of design characteristics of the product; and 3) development of relationships between design characteristics and customers' feelings, which allow adjustments of the product to maximize customers' satisfaction [6]. In order to capture customers' feelings, the semantic differential (SD) method [7] was primarily used to measure and decompose the psychological meaning of the product. Even though Kansei Engineering has been developed and used in Japan, typically for designing products, a similar approach using SD has been used in many human factors studies. For instance, the NASA-TLX (Task Load Index) was developed by NASA to assess the level of human operators' mental workload [8]. Based on the SD method, the human workload was decomposed into six factors, including mental demand, physical demand, temporal demand, performance, effort, and frustration. These six subdimensions of perceived workload are then used to calculate the overall workload level. This measurement technique has been considered one of most validated measures to capture an operator's workload on a given task. Another example is the research on development and validation of multidimensional measures of display clutter [9], [10]. In that study, the investigators developed a metric for assessing subjective clutter in a flight cockpit display. They also used the SD method to identify six latent factors in the perception of clutter to develop a measure. They then demonstrated the validity of the measure through experiment and found relationships between the subdimensions of perception with pilots' performance and physical characteristics of the displays.

The objective of this study was to: 1) develop a multidimensional measure of perceived shift quality. This included discovering the underlying common psychological factors among drivers with respect to the perceived quality of gear-shift as well as generating a quantitative metric to represent drivers' subjective perception in a multidimensional manner; 2) validate the measure through an experimental study. It was expected that the multidimensional measure is

viable to express general shift quality; and 3) find relationships between various aspects of perceived shift quality and mechanical properties which were determined through physical characteristics of the vehicle based on the experiment data. This involved modeling approaches and is expected to support determining the best adjustment of transmission control variables in order to improve driver satisfaction. Consequently, the research encompassed two general studies, one for the measure development and another for experiment. However, it should be noted that even though the measure was built for evaluating general shift quality, its primary focus was on kick-down gear shifting events. A kick-down shift event refers to the automatic change of gear to a lower gear during abrupt acceleration in automatic transmission vehicles. It is considered to be one of the critical vehicle quality issues. Also, the present study was targeting the driving population in the U.S.

STUDY I- DEVELOPMENT OF MEASURE In order to assess drivers' perceptions of shift quality, a measure consisting of semantic pairs was developed based on the SD theory [7].

Collection of Semantic Pairs In order to identify the perceptual qualities of gear shifting that common drivers might internally define, we collected descriptor terms relative to the perceived shift quality from literature, magazine, and brainstorming. The terms were used as a base for identifying semantic pairs of terms that might represent a driver's concepts of gear shifting in making subjective assessment. An initial list of semantic pairs was generated, which consisted two antonymous adjectives. Several experts with substantial experience in vehicle testing using human subjects were then asked to validate the list of semantic pairs to determine whether it covered most possibilities in describing subjective shift quality. They were also asked to suggest other semantic pairs. Consequently, we identified the following 28 pairs of terms (See Table 1).

Table 1. List of semantic pairs

Jeon et al / SAE Int. J. Passeng. Cars - Mech. Syst. / Volume 7, Issue 1 (May 2014)424

Data Collection Based on the 28 semantic pairs of descriptor terms for shift quality, a survey form was developed to ask drivers to rate the shift quality of their personal vehicle. The survey form included 28 semantic pairs, using a 10-point scale, for rating of the shift quality. Figure 1 shows the sample of the survey form. The reason in which the survey asks the driver to rate their own vehicle rather than a controlled vehicle is not to evaluate each vehicle's shift quality but to assess the relationship between the semantic pairs in order to find primary common factors and groups of sematic pairs for describing the quality. Along with the semantic pairs, another 28 questions were included to rate the utility of each semantic pair, on a 10-point scale, for describing the quality. The analysis of this data was used to select representative semantic pairs among associated pairs in a group.

The survey data was collected on-line. Forty-three English speaking U.S. citizens with automatic transmission vehicles participated the survey. The survey group ranged in age from 20 to 60 years (M=35.68) and consisted of 19 males and 14 females.

Figure 1. Sample of survey form

Factor Analysis Factor analysis using principal components was conducted to investigate the relationship among the 28 pairs of gear-shifting descriptors (principal components analysis is a specific form of factor analysis that does not make assumptions about the distribution of the response measure's variance). This analysis was intended to 1) examine the underlying factors or latent variables in drivers' perceptions of shift quality; and 2) determine how many factors are necessary to explain the pattern of variance in the perception of shift quality.

The ratings for the pairs of descriptor terms were submitted to a principal components analysis with an orthogonal equamax rotation. Through the analysis on scree plot and the cumulative eigen value, four factors were identified as latent variables explaining 76.4% of the total variance of the perception of shift quality. Table 2 shows the 4 factors and the factor loading values of each pair to the factor. Along with the factor loading values for each factor, the ratings on utility of pairs for describing shift quality, collected from the survey, were used to select the representative sematic pairs for each factor.

Consequently, 4 descriptor terms consisting of semantic pair were identified, including “Responsiveness (responsive/ hesitant)”, “Strength (strong/weak), “Smoothness (smooth/ rough),” and “Unperceivable (unperceivable/apparent)”. This means that drivers' subjective perceptions on shift quality could be comprised of multidimensional meanings and the four latent factors are primary subdimensions of internal perception. Figure 2 and 3 show MDS (multidimensional scaling) map by projecting the multidimensional meaning space of the 28 terms in the driver's mental model into a 3-dimensional space and a 2-dimensional space, respectively. The figure also supports that the 28 terms can be grouped into 4 latent factors.

Table 2. Rotated component matrix

Development of Measure Based on the four latent factors and the semantic pairs found using factor analysis, a multidimensional measure of perceived shift quality was developed. The set of shift quality descriptor terms was used as an anchor in the collection of bipolar subjective rating scales covering the underlying dimensions. The scales were integrated into an overall quality index, which required drivers to rank the importance of each dimension. Therefore, the measure was consisted of two forms, a ranking

Jeon et al / SAE Int. J. Passeng. Cars - Mech. Syst. / Volume 7, Issue 1 (May 2014) 425

form and a rating form. The ranking form was to collect the importance of each dimension for each individual (see Appendix). This form was completed once all experimental trials were finished. The evaluators (drivers in the study) were asked to select which subdimension term of two subdimensions terms was most important to them. Thus, the weight of each subdimension was calculated based on number of selected terms among 8 pair-wise comparisons (8=4C2). The rating form was used to capture the subjective ratings on each subdimension consisting of the semantic pairs while using a bipolar semantic scale. The rating form was completed after finishing each specific task event during the experimental trials (see Appendix). The ranking and ratings for specific gear- shifting events were combined in an overall shift quality score (rank-weighted sum of ratings across dimensions). This measurement approach was very similar to the design of examining other psychological phenomenon, such as the NASA-Task Load Index (TLX) for measuring subjective mental workload [8] and the display clutter matric for assessing subjective perception of visual display clutter [9][10]. However, the measure developed in the present study was used to collect data using actual vehicles.

Figure 2. Multidimensional scaling map of descriptor terms in 3-dimensional space

Figure 3. Multidimensional scaling map of descriptor terms in 2-dimensional space

STUDY 2- EXPERIMENT An experiment was conducted to validate the multidimensional measures for assessing perceived shift quality and to identify the relationship between the drivers' perception of quality and the mechanical properties in the vehicle during the gear- shifting event.

Method

Equipment Three types of current mid-size sedan vehicles were selected for the experiment (Vehicle A, B and C). The vehicles were equipped with similar engines (2400cc) and body sizes. Since vehicle A was a product of the sponsoring research company, three (3) different Transmission Control Units (TCUs) were prepared and used in vehicle A by replacing them throughout the experiment. The three TCU's were programed to generate different gear shifting patterns, including slow-smooth (A1), fast-rough (A3), and the regular pattern which is applied for the vehicle in market (A2). Consequently, five (5) vehicle samples (A1, A2, A3, B, and C) were used in the experiment. Each vehicle had a different calibration for gear-shifting response patterns. In order to prevent a subjective bias from vehicle brands on the rating of shift quality, substantial instructions were provided to the participants.

During the kick-down gear shifting events, the mechanical data, such as the vehicle's speed and gear shifting profile, were captured through an OBD (On-Board Diagnostics) interface. In addition to this, an accelerator sensor was attached in the vehicle to capture longitudinal acceleration. The data from the OBD interface and the acceleration sensor were recorded in real-time in a portable workstation during the experiment. The sampling rate of these mechanical data was 100 Hz.

Participants The study was focusing on U.S. driver population therefore twelve English-speaking drivers participated in the experiment. The drivers were balanced across age and gender. There were 6 males and 6 females. Of the drivers, 6 were young drivers (ranging in age from 21 to 37 years) and 6 were older drivers (ranging in age from 52 to 59 years old). The mean age of the 12 participants was 40.7 years of age. The average years of driving experience was 23.3 years with a standard deviation of 13.4 years. The driving experience years for young drivers and old drivers were 11.2 years (with standard deviation of 5.2 years) and 35.5 years (standard deviation: 3.1 years), respectively.

Jeon et al / SAE Int. J. Passeng. Cars - Mech. Syst. / Volume 7, Issue 1 (May 2014)426

Experiment Design and Driving Scenario As experimental conditions, three types of driving speeds (20, 40, and 60 MPH) and two types of acceleration intensions (mild and sporty) were manipulated across subjects and sample vehicles, in terms of within subject design. Thus, each participant was asked to generate kick-down gear shifting events for 5 types of sample vehicles, three types of speed conditions, and two types of acceleration intentions. A test route near Ann Arbor, Michigan, was selected for the experiment and all participants were asked to drive the 5 sample vehicles on the same route. The test route consisted of a local road (which the speed limit is 45MPH and under) and a highway (which the speed limit is 70 MPH and under). Each test trial for a sample vehicle consisted of three phases according to the driving speed. The first and second phases were to generate the kick-down events when driving at 40MPH and 20MPH on the local road. The third phase was for the events during 60MPH driving on the highway. In each phase, the participants were asked to perform two types of kick-down events (mild and sporty). The mild kick-down was defined as pressing down on the accelerator approximately 30 percent of the way down in a mild manner. Example situations also were provided to drivers, such as “adjusting the vehicle speed due to an increased of speed limit of 35MPH to 45MPH.” The sporty kick down was defined as pressing accelerator pedal very abruptly and as far down as possible. Example situation included entering on the freeway or getting ready to pass another vehicle.

Consequently, the experiment included a total of 360 observations (12 drivers × 5 vehicles × 3 speed conditions × 2 intention conditions) on the kick-down gear shifting events. In order to minimize the statistical biases, including learning effects, the order of the five vehicles for each trial and the two driving intentions in each phase were randomized and balanced across participants and trials.

Procedure The experiment was conducted after obtaining Institutional Review Board (IRB) approval for the research. The experiment protocol involved an initial briefing to the participants on the objectives of the study. Participants then gave their informed consent to the research protocol and a completed a demographic survey for recording participant experience. Participants were then presented with a briefing on the test route, two types of kick-down events with driving scenarios, and a subjective data collection form. Following this briefing, participants were required to complete practice trials in order to familiarize them with the vehicles, test route, and kick-down events. Participants then completed five experimental trials involving three phases of driving (speeds) by two driving intentions. After performing each kick-down gear shifting event, participants were asked maintain their feelings regarding the

gear-shifting shock until they were able to pull the vehicle over to a safe place. Once the vehicle was stopped in a safe place, participants were asked to provide ratings on the 4 sematic pairs of the shift quality using 20-point bipolar scales in the rating form (anchors are shown in Table 3). In order to maintain semantic consistency between participants, definitions, synonym, and acronym of each anchor of semantic pairs were provided to participants. Along with the ratings of each subdimension, a rating on the unidimensional measure of overall shift quality (SQR) was collected using a 20-point scale on the rating form. After completing all five test trials, participants were then asked to rank the relative importance of each of the pairs of the shift quality descriptor terms according to their preferences. Upon completion of the ranking form, participants were given debriefing instructions and payment for their participation. The entire experiment took approximately 2.5 hours per a participant.

Measures Two general data sets were collected in the experiment. The first data set was the subjective perception of the shift quality, which is the multidimensional measure involving a ranking of the dimension of the shift quality as well as the ratings of each. Each driver completed pairwise comparisons of the dimensions including responsiveness, smoothness, unperceivable, and strength, in terms of importance to their perceived shift quality.

The comparisons yielded a weighting factor for each dimension ranging from 0 (contributing least to their perceived quality) to 3 (contributing most to their perceived quality). Ratings on the bipolar semantic scales of each dimension were collected after completing specific kick-down gear shifting events. The vehicle and experimental conditions, such as speed and intention, were noted for each data file. The ratings were multiplied by the rankings for each dimension and summed to create a rank-weighted shift quality score (SQS), as in equation (1). The resulting score also has a value ranging from 0 to 20 (minimum to maximum quality)

(1)

Overall unidimensional perceived shift quality ratings (SQRs) were also collected as another measure of validation for the SQS.

The second data set collected was the mechanical/physical data captured from each sample vehicle during the gear shifting events. The data captured from the OBD system and the acceleration sensor was encoded for various types of information. Figure 4 shows a sample of typical mechanical property changes during a kick-down gear-shifting event. The data included gear change information (Gear Shift), turbine speed changes (NT), acceleration change information (G), and jerk (differential of acceleration) change information (Jerk). A

Jeon et al / SAE Int. J. Passeng. Cars - Mech. Syst. / Volume 7, Issue 1 (May 2014) 427

total of 32 types of data were recorded and the data was categorized by units, including time (e.g., time to start shift, total shift time, time for acceleration to start, time for reaching initial max, etc.), acceleration (e.g., values at initial and final max, amount of acceleration drop from initial max, etc.), gradient of acceleration changes (e.g., gradient drop and increase), jerk (initial, middle, and final), and turbine rotation (e.g., gradient of turbine rotation increase).

Figure 4. Sample of mechanical data profile for kick-down gear-shifting Event

Results and Discussion

Multidimensional Measure of Shift Quality Correlation analyses were conducted to identify whether driver ratings on the underlying dimensions of shift quality were consistent with overall perceived shift quality ratings (SQR). It was expected that the new multidimensional measure of the gear-shift quality targeted variations of gear-shifting feelings influential in the perception of the quality; that is, ratings across dimensions would trend in a similar manner and with the overall rating (SQR). Results revealed significant positive linear relationships of all four subdimensions with overall shift quality.

Table 3 shows the Pearson coefficients and p-values for the subdimension and overall shift quality ratings and the anchors of the various scales.

The gear-shifting scores (SQSs) were calculated by multiplying the driver's rankings of dimensions with their ratings on the dimensions for each gear-shifting event. The SQSs were highly correlated with the overall shift quality ratings (SQRs) (r= 0.811, p<0.0001). These results indicated that the ratings among the dimensions of gear shifting were consistent and that the multidimensional measure was valid for describing drivers' subjective perceptions of shift quality.

Effects of Subdimensions on Overall Quality In order to identify the contribution of the multiple dimensions of subjective perception towards the overall perceived shift quality, multiple regression analyses were conducted. Since the weight of each dimension captured using the ranking form was obtained by each individual participant regardless of intention, it was required to examine common weights of subdimensions for driving intentions. Therefore multiple regression models of SQSs in terms of ratings on the 4 subdimensions as independent variables were developed. Three models were formulated; one for mild kick-down, one for sporty kick-down, and one for an aggregated model across the two driving intention conditions. The models for overall, mild kick-down, and sporty kick-down were statistically significant (R2=0.910, p<0.0001 for overall, R2=0.900, p<0.0001 for mild kick-down, and R2=0.913, p<0.0001 for sporty kick-down) in predicting the overall ratings of perceived shift quality (SQS). Table 4 shows the results of the regression analysis. The proportion was calculated as the percentage of estimates (coefficient) among the 4 parameters in the model in order to illustrate the relative level of contribution of each subdimension to the overall quality rating. As shown in the table, the “Strength (weak/strong)” factor appeared to account to greatest portion to overall quality, followed by “Responsiveness (hesitant/ responsive)” and “Smoothness (rough/smooth)”. The “Unperceivable (apparent/unperceivable)” factor was the smallest contributor. It is also found that the contribution of “Smoothness” is decreased in sporty kick-down compared to mild kick-down, while the contribution of “Responsiveness” and “Unperceivable” are slightly increased.

Table 3. Correlations of subdimension with overall quality ratings (SQR) and descriptor terms uses as scale anchor

Jeon et al / SAE Int. J. Passeng. Cars - Mech. Syst. / Volume 7, Issue 1 (May 2014)428

Table 4. Regression analysis results for model of calculated quality score (SQS) in terms of subdimensions

Effects of Mechanical Properties on Subjective Perception Based on the subjective data, including drivers' ratings on the 4 subdimensions, calculated shift quality scores (SQS), and the 32 types of mechanical data captured during the gear-shifting events, a series of ANOVAs (analysis of variances) were conducted to identify which mechanical data affects the subjective perception of shift quality. Since many variables among the 32 types of mechanical data have internal correlations, diagnoses of multicollinearity were conducted to reduce the number of variables for the mechanical data. For example, ‘hesitant time G’ for time from accelerator pedal pressing to begin of final acceleration, which was captured using the acceleration sensor, was highly correlated with ‘total time shift’ collected by engine turbine revolution information. The ANOVAs revealed the mechanical variables affecting subjective perceptions on each subdimension and overall shift score. From the results, it was found that 10 variables were commonly affective across the 4 subdimensions and overall quality. These variables are representative variables among those that have internal correlations. The definitions and units of the 10 physical variables are described in Table 5.

Table 6 shows the summary of affective mechanical variables to subjective perceptions on overall and each of subdimensions for two acceleration intentions (mild and sporty), with an indication of significantly strong effects (marked as ‘XX’) as well as marginally significant or moderate effects (marked as ‘X’).

From the results, it was inferred that, in general: 1) the perception of responsiveness is affective to time variables such as “total_G_time”, “Time_SFT_A”, and “Time_SFT_B”; 2) smoothness is affective to the degree of second acceleration after completing the mechanical gear change, such as “G_inc_ fin_grad”; 3) jerks during the gear-shift affect the feeling of unperceivable and apparent; and 4) the subjective feeling of strength may be affected by most variables however acceleration (G) and jerks are primary variables to the strong/ week feeling. Figure 5 illustrates the general relationship of the mechanical properties with subjective perceptions.

Table 5. Units and definitions of 10 physical variables

Figure 5. Mechanical variables affecting subjective perception

Jeon et al / SAE Int. J. Passeng. Cars - Mech. Syst. / Volume 7, Issue 1 (May 2014) 429

Table 6. Summary of affective mechanical variables to subjective perception

Predicting Subjective Shift Quality Based on Mechanical Property Variables Regression models were developed to predict drivers' perception of shift quality on the 4 subdimensions and comprehensive quality (SQS) for the two driving intentions (mild and sporty) in terms of the mechanical variables. On the basis of the previous findings, the 10 mechanical variables were used as independent variables. The objective of this model development was to allow engineers to predict the expected scores of shift quality without costly experimentation using human subjects. It is also expected that the manufacturer might be able to use the model to identify which ones and how the specific mechanical variables should be controlled in order to improve the quality of the vehicle.

Consequently, ten (10) models were produced, including models of the overall shift quality score (SQS) and four subdimensions all across two driving intentions. For each regression model graphical analyses and diagnostic tests were conducted on the residuals to assess the normality assumption and the results revealed all models were statistically significant. Since the regression coefficients, which represent the effect that the mechanical variable values have on the dependent variables, such as score and ratings, were significantly different for the two driving intentions, the separated models for the driving intentions were developed. This means that the shift quality for same gear changes should be calibrated differently by the driver's intentions on acceleration. The equation (2) shows the form of multiple linear regression models and table 7 shows the values of coefficient (ai) and associated variable names (Xi) of the model of SQS for sporty kick-down in terms of the 10 mechanical property variables. In the equation ‘k’ means constant value of the regression model, as an intercept.

(2)

Table 7. Variables and coefficients of model of SQS for sporty kick-down

Unfortunately, it should be noted here that the detailed coefficient values for all model are not provided due to confidentiality constraints given by Hyundai Motor Company.

A similar approach has been existed to predict and compare general drivability based on mathematical or statistical models (AVL-Drive) [11]. The drivability evaluation tool was designed to generate predicted scores of drivability for various driving conditions throughout analysis of physical data captured from vehicles during actual driving or simulation. However, while it might be useful for benchmarking vehicles, it may not provide detailed information of impact of changes of physical properties on drivers' various psychological subdimensions consisting perceived gear-shifting quality. The examination of impact on subdimensions may allow manufacturers to differentiate vehicle characteristics such as sporty and mild vehicles as well as economy and premium vehicles.

Utility of Measure and Models Based on the four latent subdimensions in the measure and model developed through this experimental study, an evaluation tool was developed. The tool is programed to

Jeon et al / SAE Int. J. Passeng. Cars - Mech. Syst. / Volume 7, Issue 1 (May 2014)430

generate the predicted score of the overall shift quality and ratings on each of the four subdimensions, using a series of values for the 10 mechanical variables captured during kick-down gear shifting events done by engineers. This approach implies that expert engineers can evaluate shift quality without costly experimentation using numerous normal drivers as participations. However, the approach still reflects the normal drivers' perception of shift quality in a multidimensional manner. The tool also provides two graphical outputs (see Figures 6 and 7). As shown in Figure 6, the expected score and ratings on the four subdimensions were depicted using a radar chart with five points along with the overall scores located at the center top point. Two points on the right hand side represent subdimensions related to sporty feelings (responsiveness and strength) and another two points on left hand side include mild feelings (smoothness and unperceivable). This chart can also be used for the comparison of different gear-shifting characteristics by visualization of meaning space differentiation. For example, in Figure 6, it can be confirmed that vehicle A presents more sporty transmission characteristics than vehicle B, with a higher overall shift quality rating. Figure 7 shows a sample of another output from the evaluation tool. As shown, the graph identifies the differences of values of the mechanical control variables between two vehicles' gear-shifting characteristics. This output can be used to identify which ones and how mechanical properties are superior or inferior compared to other vehicles as well as to provide a guideline to adjust the mechanical properties to improve the overall shift quality.

Figure 6. Expected score and ratings on a radar chart

Figure 7. Comparison of mechanical properties

CONCLUSIONS The present study achieved the objectives of developing a new multidimensional measure of shift quality and validation of the measure through an experimental study along with model development. Specifically: 1) four latent factors composing the drivers' perception of shift quality have been identified using SD method. These factors include responsiveness, smoothness, unperceivable, and strength; 2) a measure to assess the overall gear-shifting score was developed based on the ratings and ranking of the 4 subdimensions; 3) the multidimensional measure was validated by having internal correlations with unidimensional measures; 4) the 10 mechanical variables in the transmission control affecting each subdimension of perceived gear shift quality were identified; 5) models were developed to predict overall quality score and ratings on the subdimensions and an evaluation tool based on the models.

There are several limitations on this study. First, even though the measures were developed to assess general shift quality, only the shift quality for kick-down events was focused in this study. Since there are other types of gear shifting, such as up shift events, further studies on other shifting events are required. Second, the vehicles used to collect data in the experiment were comparable mid-size sedans. That means, the model developed based on the experimental data might not be generalizable for other types of vehicles that have different ranges of mechanical property values, such as luxury sedans with greater engine volumes than the vehicles used in the study. Thus, in order to develop a more general model, various types of vehicles are needed for testing and data collection. Third, the drivers that participated in the study were average U.S. citizens who speak English as a primary language. Since the latent factors used to describe shift quality was based on semantic pairs in English, the underlying factors for other countries using other languages could be different from those found in this study. In addition to this, the drivers' common mental model (mental stereotype) to perceive gear shifting might be different due to international or cultural differences. Therefore, it is required to develop the appropriate measures and models for each language and culture while applying similar methods used in this study.

However, the approach and results of this research are expected to be applicable not only for the development and evaluation of shift quality but also for assessing various qualities within vehicles in drivers' perspectives, such as driving comfort, sound, material, etc.

REFERENCES 1. Dorey, R. and Martin, E., “Vehicle Driveability - The Development

of an Objective Methodology,” SAE Technical Paper 2000-01-1326, 2000, doi:10.4271/2000-01-1326.

2. Schwab, L., “Development of a Shift Quality Metric for an Automatic Transmission,” SAE Technical Paper 941009, 1994, doi:10.4271/941009.

3. Jeon, B. W., “A Study on the Automatic Calibration of AT Shift Quality in Test bed Environment,” KSAE 09-B0113, pp.695-701, 2009.

Jeon et al / SAE Int. J. Passeng. Cars - Mech. Syst. / Volume 7, Issue 1 (May 2014) 431

4. Moir, G. B., “An Investigation into Objective Measures of gear-shift quality,” in Proceedings of the Institution of Mechanical Engineers, 209, No. D4/1995: 273-279, 1995.

5. Zhang, J., Lei, Y., Hua, X., Wang, J., Ge, A., “Proposed Shift Quality Metrics and Experimentation on AMT Shift Quality Evaluation,” in Proceeding of Third International Conferences on Natural Computation, 494-498, 2007.

6. Nagamach, M., “Kansei Engineering: A New Ergonomic Consumer- oriented Technology for Product Development,” International Journal of Industrial Ergonomics, Vol. 15, 3-11, 1995.

7. Osgood, C. E., Suci, G. J., Tannenbaum, P. H., “The Measure of Meaning,” University of Illinois Press, 1957.

8. Hart, S. G., Staveland, L. E. “Development of NASA-TLX (Task Load Index): Results of empirical and theoretical research,” In Handcok P. A. & Meshkati N. (Eds.), Human Mental Workload, Vol. 1, 139-183, 1988.

9. Kim, S-H., Prinzel, L. J., Kaber, D. B., Alexander, A. L., Stelzer, E. M., Kaufmann, K., Veil, T., “Multidimensional Measure of Display Clutter and Pilot Performance for Advanced Head-up Display,” Aviation, Space, and Environmental Medicine, 82 (10), 1013-1022, 2011.

10. Kaber, D. B., Alexander, A. L., Stelzer, E. M., Kim, S-H., Kaufmann, K., Hsiang, S. “Perceived Clutter in Advanced Cockpit Displays: Measurement and Modeling with Experienced Pilots,” Aviation, Space, and Environmental Medicine, 79 (11), 1007-1018, 2008.

11. List, H. and Schoeggl, P., “Objective Evaluation of Vehicle Driveability,” SAE Technical Paper 980204, 1998, doi:10.4271/980204.

CONTACT INFORMATION Byeong Wook Jeon [email protected]

Sang-Hwan Kim [email protected]

ACKNOWLEDGMENTS Portions of this research were reported at the SAE 2013 World Congress, April 16-18, 2013, Detroit, MI. We would like to thank Heather Harrelson for helping data collection and analysis.

DEFINITIONS/ABBREVIATIONS SD - Semantic Differential TCU - Transmission Control Unit IRB - Institutional Review Board SQS - (multidimensional) Shift Quality Score SQR - (unidimensional) Shift Quality Rating OBD - On-Board Diagnostics NT - Turbine Speed G - Acceleration

Jeon et al / SAE Int. J. Passeng. Cars - Mech. Syst. / Volume 7, Issue 1 (May 2014)432

APPENDIX

RANKING AND RATING FORM

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

Positions and opinions advanced in this paper are those of the author(s) and not necessarily those of SAE International. The author is solely responsible for the content of the paper.

Jeon et al / SAE Int. J. Passeng. Cars - Mech. Syst. / Volume 7, Issue 1 (May 2014) 433

Copyright of SAE International Journal of Passenger Cars: Mechanical Systems is the property of SAE International and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use.