Yaw Rate Error – a Dynamic Measure of Lane Keeping Control Performance for the Retrospective Analysis of Naturalistic Driving Data

The aim of this work is to define and evaluate a “yaw rate error” (YRE) derived from naturalistic driving data to quantify driver steering performance during lane keeping. This measure of lane keeping performance is based on the predicted kinematic control error at any instance. Scope is limited to the demonstration that such a quantity exists, that can be computed from naturalistic driving data, and that it correlates with instantaneous control performance in real-world driving. The YRE is defined as a measure of conflict: the difference between current vehicle yaw rate and kinematic values required to be consistent with forward lane boundary crossing. A second, well-known measure is computed for comparison: the predicted time to lane crossing (TTLC). All data is obtained from naturalistic driving databases containing detailed information (over 200 signals at 10 Hz.) on driver input and vehicle response as well as aspects of the highway and traffic environment. As a continuously updated measure of the control correction required by an alert driver, it is expected that the YRE will be more informative of driving situations than the simpler kinematic measure TTLC. This latter measure is only loosely related to the closed loop control of vehicle motion. For example a very small TTLC can represent either a critical case where the vehicle is about to depart the lane and requires a large correction, or it could be a case where the vehicle is close to the lane boundary but with small lateral velocity requiring only a small correction. The YRE represents the severity of the possible lane departure in a natural way, accounting for current position, path direction, and path curvature. While no in-depth statistical analysis is conducted for YRE, it is proposed as a new tool for post-hoc analysis of driver steering performance during lane keeping. INTRODUCTION Driving is a control task based on visual input; it includes filtering of input for relevance, extracting signals or patterns from that visual information, and hence provides a reference to guide steering and speed control. Control action then involves manual effort by the driver to modulate vehicle motion using further force and acceleration cues [1-3]. Here we focus on the visual reference for lane keeping in terms of a conflict measure or error criterion. In broad terms we seek a simple measure of the control reference for when the driver is concerned with staying in the lane but less concerned with some optimal path within that lane. To this end we introduce and evaluate a suitable measure of “yaw rate correction required” or yaw rate error. Since no preferred path is computed, the YRE is computed for multiple lane boundary points and the most critical of these will represent the overall correction required. This metric has been used previously in driver modeling [4] and applied to collision avoidance [5] The approach is analogous to longitudinal speed control in traffic, where control action required can be found in terms of the vehicle deceleration required to avoid a collision with the vehicle in front. Again, this contrasts with the predicted time to collision (TTC), based on instantaneous positions and velocities of the vehicles [6]. While in the speed control problem there is essentially a single target point, the more complex lane keeping activity involves multiple conflict points and more complex vehicle kinematics. We focus on yaw velocity rather than the related variables of path curvature and lateral acceleration because of the focus on visual reference. Yaw velocity is directly available to the driver as the perceived angular rate of distant or peripheral objects across the field of view. Path curvature by contrast