Conference Presentation, Keynote
MIT-AVT: Data Collection Device (for Large-Scale Semi-Autonomous Driving)
- Transitioning the Ryder unit to a Jetson TX2 single-board computer to increase compute power and support added sensors for potential real-time systems.
- Developing onboard algorithms in future iterations to process video in real-time, automatically retaining interesting segments and discarding uninteresting data to improve edge case training efficiency.
- Shifting data collection strategies from recording all data to selectively recording specific useful epochs as insights into human behavior in semi-autonomous vehicles improve.
- Maintaining daily data collection volumes of 500 to 1,000 miles while expanding the fleet with new vehicles such as a Tesla Model 3 and a Cadillac CT6 Super Cruise system.
- Prioritizing rich internal environment sensor information regarding driver face, glance, cognitive load, and body pose alongside external environment data to ensure safe semi-autonomous and autonomous design.
- Treating the understanding of human supervisor attentiveness and effective road monitoring as a fundamental requirement for autonomous vehicle technology.