Today, and possibly for a long time to come, the full driving task is too complex an activity to be fully formalized as a sensing-acting robotics system that can be explicitly solved through model-based and learning-based approaches in order to achieve full unconstrained vehicle autonomy. Localization, mapping, scene perception, vehicle control, trajectory optimization, and higher-level planning decisions associated with autonomous vehicle development remain full of open challenges. This is especially true for unconstrained, real-world operation where the margin of allowable error is extremely small and the number of edge-cases is extremely large. Until these problems are solved, human beings will remain an integral part of the driving task, monitoring the AI system as it performs anywhere from just over 0% to just under 100% of the driving. The governing objectives of the MIT Advanced Vehicle Technology (MIT-AVT) study are to 1) undertake large-scale real-world driving data collection that includes high-definition video to fuel the development of deep learning-based internal and external perception systems; 2) gain a holistic understanding of how human beings interact with vehicle automation technology by integrating video data with vehicle state data, driver characteristics, mental models, and self-reported experiences with technology; and 3) identify how technology and other factors related to automation adoption and use can be improved in ways that save lives. In pursuing these objectives, we have instrumented 23 Tesla Model S and Model X vehicles, 2 Volvo S90 vehicles, 2 Range Rover Evoque, and 2 Cadillac CT6 vehicles for both long-term (over a year per driver) and medium-term (one month per driver) naturalistic driving data collection. Furthermore, we are continually developing new methods for the analysis of the massive-scale dataset collected from the instrumented vehicle fleet. The recorded data streams include IMU, GPS, and CAN messages, and high-definition video streams of the driver's face, the driver cabin, the forward roadway, and the instrument cluster (on select vehicles). The study is on-going and growing. To date, we have 122 participants, 15610 days of participation, 511638 mi, and 7.1 billion video frames. This paper presents the design of the study, the data collection hardware, the processing of the data, and the computer vision algorithms currently being used to extract actionable knowledge from the data.

Accurate, robust, inexpensive gaze tracking in the car can help keep a driver safe by facilitating the more effective study of how to improve (i) vehicle interfaces and (ii) the design of future advanced driver assistance systems. In this study, the authors estimate head pose and eye pose from monocular video using methods developed extensively in prior work and ask two new interesting questions. First, how much better can they classify driver gaze using head and eye pose versus just using head pose? Second, are there individual‐specific gaze strategies that strongly correlate with how much gaze classification improves with the addition of eye pose information? The authors answer these questions by evaluating data drawn from an on‐road study of 40 drivers. The main insight of the study is conveyed through the analogy of an ‘owl’ and ‘lizard’ which describes the degree to which the eyes and the head move when shifting gaze. When the head moves a lot (‘owl’), not much classification improvement is attained by estimating eye pose on top of head pose. On the other hand, when the head stays still and only the eyes move (‘lizard’), classification accuracy increases significantly from adding in eye pose. The authors characterise how that accuracy varies between people, gaze strategies, and gaze regions.

When asked a great number of people believe that, as pedestrians, they make eye contact with the driver of an approaching vehicle when making their crossing decisions. This work presents evidence that this widely held belief is false. We do so by showing that, in majority of cases where conflict is possible, pedestrians begin crossing long before they are able to see the driver through the windshield. In other words, we are able to circumvent the very difficult question of whether pedestrians choose to make eye contact with drivers, by showing that whether they think they do or not, they can’t. Specifically, we show that over 90% of people in representative lighting conditions cannot determine the gaze of the driver at 15m and see the driver at all at 30m. This means that, for example, that given the common city speed limit of 25mph, more than 99% of pedestrians would have begun crossing before being able to see either the driver or the driver’s gaze. In other words, from the perspective of the pedestrian, in most situations involving an approaching vehicle, the crossing decision is made by the pedestrian solely based on the kinematics of the vehicle without needing to determine that eye contact was made by explicitly detecting the eyes of the driver.

Semantic scene segmentation has primarily been addressed by forming high-level visual representations of single images. The problem of semantic segmentation in dynamic scenes has begun to receive attention with the video object segmentation and tracking problem. While there has been some recent work attempt to use deep learning models on the video level, what is not known is how the temporal dynamics information is contributing to the full scene segmentation. Moreover, most existing datasets only provide full scene annotation on non-consecutive images to ensure the variability of scenes, making it even harder to explore novel methods on video-level modeling. To address the above issues, our work takes steps to explore the behavior of modern spatiotemporal modeling approaches by: 1) constructing the MIT DriveSeg dataset, a large-scale video driving scene segmentation dataset, densely annotated for pixel-level semantic classes with 5000 consecutive video frames, and 2) proposing a joint-learning framework that reveals the contribution of temporal dynamics information in regard to different semantic classes in the driving scene. This work is intended to help assess current methods and support further exploration of the value of temporal dynamics information in video-level scene segmentation.

We use an immersive virtual reality environment to explore the intricate social cues that underlie non-verbal communication involved in a pedestrian’s crossing decision. We “hack” non-verbal communication between pedestrian and vehicle by engineering a set of 15 vehicle trajectories, some of which follow social conventions and some that break them. By subverting social expectations of vehicle behavior we show that pedestrians may use vehicle kinematics to infer social intentions and not merely as the state of a moving object. We investigate human behavior in this virtual world by conducting a study of 22 subjects, with each subject experiencing and responding to each of the trajectories by moving their body, legs, arms, and head in both the physical and the virtual world. Both quantitative and qualitative responses are collected and analyzed, showing that, in fact, social cues can be engineered through vehicle trajectory manipulation. In addition, we demonstrate that immersive virtual worlds which allow the pedestrian to move around freely, provide a powerful way to understand both the mechanisms of human perception and the social signaling involved in pedestrian-vehicle interaction.