9.1 Overall Conclusions
This research was done to gain more knowledge about subjective tire handling assessment by professional test drivers with the aim of improving tire handling assessment methods. This supports the development of good handling tires, contributing to the active safety of a vehicle and to the pleasure of driving. This main goal is detailed into the following research objectives:
- To provide information on what handling behavior is considered as good by the driver, by predicting subjective tire handling assessments based on derived objective measures and by analyzing the these measures. This information can support the actual driver's subjective evaluations and provide comparative information for assessments. - To provide information on how the driver's subjective assessment is
formed by analyzing driver model parameters from vehicle handling simulations.
Professional tire test drivers are very capable of doing their work; improvement is therefore searched for in the assessment methods. Specifically, by focusing on the driver, because handling is determined by the closed-loop driver-vehicle system and the driver is the core of subjective assessment.
Three methods, all based on field experiments, are derived which fulfil the research objectives: using a GRNN, based on workload measures and based on driver models. They have in common that they predict the driver's subjective assessment of tire handling, based on vehicle dynamics measurements. The differences lie in the way they derive and utilize these measurements.
The GRNN method (described in chapter 6) and workload measures method (described in chapter 7) provide information on the ‘what’-question:
- The GRNN showed good prediction of the driver’s subjective assessment scores covering driving a handling track, based on vehicle dynamic measurements taken only during the double lane changes.
- The GRNN can even perform well on predicting handling aspects not related to driving the double lane change.
- Analysis of the important measures used by the GRNN can provide information on the vehicle dynamics behavior relevant for the driver. For general handling behavior, regarding all aspects, this showed that metrics based on lateral velocity, steering wheel moment and vehicle slip angle were most relevant.
- Additional advantages of using a GRNN compared to other regression methods are: it works well with a limited dataset, does not need a-prior knowledge about the data or underlying model, has good prediction between data points, behaves well for extrapolation. In addition, constructing the GRNN is easy and adding new examples, with possibly different in- and outputs can be done during usage, incorporating new knowledge when it becomes available.
- Mental workload measures are found to be indicators of driver’s perceived tire handling behavior, even when performance measures do not show differences. More mental workload indicates less tire handling performance, the driver has to put in more effort to keep the performance high.
- HFA showed to be a promising measure in this research for indicating driver’s mental workload during tire handling, but needs more research to make these results broader applicable.
- Steering wheel measure HFA showed to be the best mental workload measure for this research.
- Tire handling assessment is found to be situated in region A3 of the performance - workload model of De Waard (Waard, de, 1996) (Fig. 19), which let driver’s workload measures distinguish between tire handling differences when performance measures cannot.
- The relationship between performance and driver’s workload in region A3 of the performance - workload model of De Waard (Waard, de, 1996) is found to be different for tire handling (Fig. 47). For tire handling, increasing task demand gives linearly increasing mental workload, but at a certain point the mental workload increases rapidly to maximum workload, just when the performance decreases.
The driver model method (described in chapter 8) provides information on the ‘how’-question:
- The parameters of the driver models applied in this research represent driver characteristics. By analyzing the identified driver model parameters, information on these related driver characteristics was derived. Especially the parameter changes due to different task demands and corresponding mental and physical workload provides information on the adapting behavior of the driver.
- From the three derived driver models A, B and C, driver model B showed to be the best model by having a simple, realistic perception part using a discrete preview path, but also having the optimal number of driver model parameters to explain driver behavior.
- The lead term constant of driver model B showed to be the best candidate for predicting subjective tire handling for both professional and nonprofessional drivers.
9.2 Discussion
This research is a first step in predicting and quantifying the driver’s tire feeling. This must provide more knowledge about subjective tire handling assessment by professional test drivers with the aim of improving tire handling assessment methods. With a growing demand for tires, a tire manufacturer is faced with a growing need for tire subjective assessments, whilst limited resources are available. The methods presented here offer new possibilities to gather this information on what handling behavior is considered as good by the driver, and how the driver's subjective assessment is formed.
From a scientific point of view, the “how” question is probably the most interesting. It gives a possible explanation for the driver’s perception and adaptation during tire handling assessment. Although the first driver research started in the late fifties of the previous century, and handling was soon incorporated, driver research related to tire handling is very limited and often focusing on the influence of tire characteristics on the vehicle-part of handling. A possible reason could be, that for tire testing, performance based measures can show equal results, while the driver subjective results are different. This research found that tire handling assessment is situated in the task-related effort region of the performance - workload model of De Waard (Waard, de, 1996) (Fig. 19), where performance measures cannot distinguish between different task demand, like driving with different handling tires. Hence, differences can only be revealed when the focus is on the driver.
In this research, the driver-part of handling is therefore focused on. By using parameters representing relevant, realistic driver behavior, the driver model method is able to reveal how the driver’s perception and adaptation during tire handling manifests itself. Relating these results to the tire handling assessment scores can also provide us with information on how the driver experiences good handling.
From a tire manufacturer point of view, the “what” question is probably the most interesting. It provides information that can support the actual driver's subjective evaluations and that can provide comparative information for assessments. The method using a GRNN can predict subjective tire handling assessments based on derived objective measures. Analyzing these measures can also provide what handling behavior is considered as good by the driver.
The value of the methods derived here for a tire manufacturer depends not only on promising results, but also on the ease of implementation. Nowadays, subjective assessment of tire handling is not only done by high speed driving on handling tracks like the Nürburgring, but more and more on automotive proving grounds like in Papenburg, including repeating standardized maneuvers, like lane changes, on different speeds. This makes the methods derived in this research, which are based on such maneuvers, easier to implement in everyday tire testing.
This research was based on field experiments derived from tire testing practice, to ensure validity of the results, but the number and complexity of the measurements used for vehicle dynamics and drivers was nowhere near representing real testing conditions where no or limited measurements were