The behavior of self-driving cars may difer from people's expectations (e.g. an autopilot may unexpectedly relinquish control). This expectation mismatch can cause potential and existing users to distrust self-driving technology and can increase the likelihood of accidents. We propose a simple but efective framework, AutoPreview, to enable consumers to preview a target autopilot's potential actions in the real-world driving context before deployment. For a given target autopilot, we design a delegate policy that replicates the target autopilot behavior with explainable action representations, which can then be queried online for comparison and to build an accurate mental model. To demonstrate its practicality, we present a prototype of AutoPreview integrated with the CARLA simulator along with two potential use cases of the framework. We conduct a pilot study to investigate whether or not AutoPreview provides deeper understanding about autopilot behavior when experiencing a new autopilot policy for the frst time. Our results suggest that the AutoPreview method helps users understand autopilot behavior in terms of driving style comprehension, deployment preference, and exact action timing prediction.
CCS CONCEPTS• Human-centered computing → Human computer interaction (HCI); • Computing methodologies → Artifcial intelligence.
The task of driver attention prediction has drawn considerable interest among researchers in robotics and the autonomous vehicle industry. Driver attention prediction can play an instrumental role in mitigating and preventing highrisk events, like collisions and casualties. However, existing driver attention prediction models neglect the distraction state and intention of the driver, which can significantly influence how they observe their surroundings. To address these issues, we present a new driver attention dataset, Co-CAtt (Cognitive-Conditioned Attention). Unlike previous driver attention datasets, CoCAtt includes per-frame annotations that describe the distraction state and intention of the driver. In addition, the attention data in our dataset is captured in both manual and autopilot modes using eyetracking devices of different resolutions. Our results demonstrate that incorporating the above two driver states into attention modeling can improve the performance of driver attention prediction. To the best of our knowledge, this work is the first to provide autopilot attention data. Furthermore, CoCAtt is currently the largest and the most diverse driver attention dataset in terms of autonomy levels, eye tracker resolutions, and driving scenarios.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.