Answers are grounded in this paper and cite their sources.
This thesis develops a framework for understanding driver interaction, enhancing autonomous driving integration and safety.
The automotive industry is witnessing significant advancements in autonomous and assisted driving functions. However, a major challenge remains in ensuring these systems are accepted by drivers and effectively integrated into their driving routines. Hence, the necessity of a comprehensive framework becomes crucial to analyze and enhance the interaction between drivers and assisted driving functions. In this thesis, we design and implement a driver interaction based framework for shared lateral driving. The research is divided into three main objectives: developing a methodology to capture the dynamics of shared control, constructing a driver interaction classifier, and modeling concepts for individualized driver support. We introduce a novel driver-steering interaction model to analyze the interplay between driver and assistance system torques. The model allows a comprehensive analysis of the driver torque, the assistance torque, and their combined effect on the vehicle's steering behavior. To accurately represent these interactions, the driver-steering interaction model requires the formulation of a quadratic program. We apply the Varying Coefficient (VC) method to effectively formulate this QP. We develop a driver interaction classifier based on a designed real-driving experimental framework. We identify and suggest suitable classification features: conflict and passivity, path consistency, adaptation and individual path pattern features. We categorize the identified driver interaction strategies into five distinct classes: adaptation, persistence, selective persistence, nonintervention and uncertainty. The classification is validated through a comparison with subjective expert assessments. Additionally, we apply Dynamic Mode Decomposition (DMD) to analyze the underlying dynamics for each class. We present concepts for adapting the system behavior in real-time to account for the driver interaction strategies. We design a Model Predictive Control (MPC) system and conduct a performance analysis to ensure it aligns with desired system behavior for each class. We apply online active-set methods and interior point methods to solve the optimization problems in real-time. The framework is implemented with a focus on real-world applicability. Additionally, we suggest further concepts for adapting the system behavior based on independent components of the framework.
No takes yet. Share an insight, caveat, or question.
A. Amouri (2025) studied this question.
Synapse has enriched 8 closely related papers on similar clinical questions. Consider them for comparative context: