This study covered two areas of work. Firstly, the issues of applying non-linear genetic algorithms to model subjective-objective links where limited numbers of data points are available were addressed. It has been seen from this that, with care, suitable neural network solutions can be used to find links within data where they are not immediately visible, and allowances must be made for noise that is inherent in subjective ratings. The second area of work was the collection of the neural network results into practical design recommendations for a vehicle to achieve good subjective handling ratings. Generally speaking, the ideal ranges of metric values appear to be in line with those that experienced vehicle test engineer tend to favour. 1. Phase delays of yaw rate and lateral acceleration and dynamic response times are preferred (within the range of the tests to be as short as possible. 2. Results from metrics concerned with steering torques suggest that light steering is preferred, again so far as the range of test vehicle is concerned. The development of modern shaped electronic power steering systerns goes further towards improving the subtleties of this metric. 3. The results suggested that a preferred range of values existed for steady state yaw rate gain (0.1-0.2). Preferred values of yaw rate gain(< 0.25) at a frequency of 0.7 Hz confirm that large increases in yaw rate gain from the steady state condition are disliked. This is consistent with common understanding that the frequency response curve should be flat within the working bandwidth to maintain the feeling of consistency. 4. The results confirmed that drivers are consistent in subjectively rating understeering vehicle configurations more positively than oversteering ones. Overall, the methodology described here has been shown to be a powerful tool in uncovering links between measured vehicle metrics and subjective ratings, where large amounts of noise are evident and links are not clearly defined by linear functions.
Identification of subjective-objective vehicle handling links using neural networks for the foresight vehicle
Identifikation der subjektiv-objektiven Fahrzeugführung mit Hilfe neuraler Netze für das 'Voraussehende Fahrzeug'
2002
9 Seiten, 1 Bild, 2 Tabellen, 7 Quellen
Aufsatz (Konferenz)
Englisch
SAE Technical Papers | 2002
|British Library Conference Proceedings | 2002
|Vehicle handling assessment using a combined subjective-objective approach
Kraftfahrwesen | 1998
|Vehicle Handling Assessment Using a Combined Subjective-Objective Approach
British Library Conference Proceedings | 1998
|