OBJECTIVE METRICS FOR VEHICLE HANDLING AND STEERING AND THEIR CORRELATIONS WITH SUBJECTIVE ASSESSMENTS |
G. L. GIL GÓMEZ1, M. NYBACKA2, E. BAKKER1, L. DRUGGE2 |
1Volvo Car Corporation 2KTH Royal Institute of Technology |
|
|
|
ABSTRACT |
This paper focuses on increasing the available knowledge about correlations between objective metrics and subjective assessments in steering feel and vehicle handling. Linear and non-linear correlations have been searched for by means of linear regression and neural network training, complemented by different statistical tools. For example, descriptive statistics, the t-distribution and the normal distribution have been used to define the 95% confidence interval for expected subjective assessments and their mean, which makes it possible to predict the subjective rating related to a given objective metric and its area of confidence. Single- and multi-driver correlations have been investigated, as well as how the use of different databases and different vehicle classes affects the results. A method for automatizing the search for correlations when using the driver-by-driver strategy is also explained and evaluated. Ranges of preferred objective metrics for vehicle dynamics have been defined. Vehicles with characteristics within these ranges of values are expected to receive a higher subjective rating when evaluated. Finally, linear correlations between objective metrics have been studied, linear dependency between objective metrics has been identified and its consequences have been presented. |
Key Words:
Steering feel, Vehicle handling, Driver preference, Objective metrics, Subjective assessments, Regression analysis, Neural network |
|
|
|