Study Assessed the Effects of Adaptation to Driver Assistance Systems Using Driver Performance and Eye Glance Data over Four Simulator Sessions.
McLean, Virginia, United States
Driver Adaptation to Vehicle Automation: The Effect of Driver Assistance Systems on Driving Performance and System Monitoring
Summary Information
Under Society of Automotive Engineers (SAE) Level 1 driving automation, either lateral or longitudinal movement can be temporarily controlled by an automation system under driver supervision, but the driver remains responsible for the driving functions of the vehicle overall. Using the Federal Highway Administration Highway Driving Simulator, researchers assessed the effects of driver adaptation to two Level 1 driver assistance systems, cooperative adaptive cruise control (CACC) and lane-keeping-assist (LKA), on drivers’ performance over time. Adaptation was measured by comparing driver responses to normal roadway conditions and unexpected critical events which took place during earlier versus later experiences with the driver assistance system.
Methodology
Over the course of two weeks, 48 licensed drivers completed four experimental driving sessions in the simulator, during which they gained experience with different driver assistance systems assigned to them. One group of participants gained experience with CACC, another with LKA, and another with both CACC and LKA (CACC + LKA). Additionally, a control group was not exposed to any driver assistance system. During the driving simulator tests, data were collected for driver performance metrics, such as speed and lane position variability, eye glance behavior, and physiological analysis. Unexpected events such as a deer crossing the road were also utilized to assess participant reactions early after introduction to the simulator and later during the last session of the study. A linear mixed-effects model and general linear mixed model were used to analyze how driving behavior changed over time, and were supplemented by before and after questionnaires.
Findings
- The statistical tests showed a significant increase in drivers’ gazes towards the windshield when using CACC, or CACC + LKA, after the introductory session, suggesting drivers shifted from monitoring speed to monitoring the external environment. Participants in the LKA and control groups spent much more time looking toward the interior of the simulated vehicle.
- Based on analysis of the questionnaires, participants had a neutral level of trust in the LKA and CACC technologies both before and after the simulated driving, suggesting that the repeated experience with the technologies did not affect opinions.
- Rates of crashing in response to the critical event for each of the conditions are shown in Table 1. In three of the four cases, the crash rate was reduced over the sessions, but were not statistically significant. Additionally, test results indicated that collision speeds were higher in the initial session (39.21 miles per hour) than the last session (38.91 miles per hour) where drivers had more experience with the driver assistance systems.
Table 1: Critical Event Crash Rates Early and Later in Study
Session |
Control |
LKA |
CACC |
CACC + LKA |
Session 1 (first) |
91.7 percent |
75.0 percent |
91.7 percent |
100 percent |
Session 4 (last) |
75.0 percent |
91.7 percent |
75.0 percent |
66.7 percent |