RESUMO
The transition to fully autonomous roadways will include a long period of mixed-autonomy traffic. Mixed-autonomy roadways pose a challenge for autonomous vehicles (AVs) which use conservative driving behaviours to safely negotiate complex scenarios. This can lead to congestion and collisions with human drivers who are accustomed to more confident driving styles. In this work, an explainable multi-variate time series classifier, Time Series Forest (TSF), is compared to two state-of-the-art models in a priority-taking classification task. Responses to left-turning hazards at signalized and stop-sign-controlled intersections were collected using a full-vehicle driving simulator. The dataset was comprised of a combination of AV sensor-collected and V2V (vehicle-to-vehicle) transmitted features. Each scenario forced participants to either take ("go") or yield ("no go") priority at the intersection. TSF performed comparably for both the signalized and sign-controlled datasets, although all classifiers performed better on the signalized dataset. The inclusion of V2V data led to a slight increase in accuracy for all models and a substantial increase in the true positive rate of the stop-sign-controlled models. Additionally, incorporating the V2V data resulted in fewer chosen features, thereby decreasing the model complexity while maintaining accuracy. Including the selected features in an AV planning model is hypothesized to reduce the need for conservative AV driving behaviour without increasing the risk of collision.
RESUMO
As driving performance relies heavily on the interpretation of visual information, driving simulators require a visual display that can effectively communicate the virtual environment to the driver. Most high-fidelity visual displays include an expensive system of high-definition projectors and wraparound screens. To reduce the overall cost of a driving simulator while preserving the generalizability of results to naturalistic driving, head mounted displays (HMD) are being considered as a substitute visual cueing system. Recent innovations to virtual reality technologies are encouraging, however, differences between HMDs and more traditional visual displays have not been explored for all types of driving measures. In particular, while existing literature provides insight into the validity of HMDs as a substitute for higher fidelity visual displays in tests of driver behaviour and performance, there is a gap in the literature regarding differences in physiological responses. In the current study, upper body muscle activation and joint angle ranges were compared between an Oculus™ Rift Development Kit 2 HMD and a system of wrap around screens. Twenty-one participants each completed two simulated drives, one per display, in a counterbalanced order. During the simulation, drivers encountered unanticipated pedestrian crossings during which peak surface electromyography, root-mean-square of the surface electromyography signal and joint angles were determined bilaterally on the upper limbs. No significant differences (p ≤ 0.05) were observed between the Oculus™ Rift HMD and the wrap around screens for all dependent variables with the exception of left joint range of motion in female participants, suggesting that the HMD reduced field of view had a minimal effect on driver kinematics and no effect on muscle activation levels. Upper body bracing was observed during the hazard response time segments characterized by significantly increased muscle activity during hazard response time segments and minimal joint movement. Considering the lack of significant kinematic and muscle activation differences between the two visual inputs, HMD technology for hazard response may provide a suitable alternative to wrap around screens for studying kinematic responses during hazardous driving scenarios.
