Field tests and machine learning approaches for refining algorithms and correlations of driver's model parameters

Area of Science:

• Human-Computer Interaction
• Artificial Intelligence
• Transportation Engineering

Background:

• The Adaptive Integrated Driver-vehicle InterfacE (AIDE) European Integrated Project developed a driver's model to predict driver intentions.
• Validating dynamic parameters like driving task demand and driver distraction is crucial for advanced driver-assistance systems.

Purpose of the Study:

• To validate algorithms and correlations for predicting driver intentions through dynamic parameters.
• To model and validate driving task demand and driver distraction using machine learning techniques.

Main Methods:

• Field tests were conducted using a driving simulator to collect driver behavioral data.
• Machine learning techniques, including adaptive neuro fuzzy inference systems (ANFIS) and artificial neural networks (ANN), were employed.
• Two distinct models for task demand and distraction were developed, one for each machine learning technique.

Main Results:

• Models for driving task demand and driver distraction were successfully developed and validated.
• Artificial neural networks demonstrated promising results for predicting driver distraction, exhibiting low prediction errors.
• Comparison between predicted and expected outcomes for both machine learning techniques was performed.

Conclusions:

• Machine learning techniques, particularly ANNs, are effective in modeling and predicting driver distraction.
• The validated parameters contribute to the advancement of the AIDE driver's model.
• The findings support the development of more sophisticated driver-assistance systems.