Computer Science Thesis Proposal
— 11:30am
Location:
In Person
-
Traffic21 Classroom, Gates Hillman 6501
Speaker:
MIHIR BALA
,
Ph.D. Student, Computer Science Department, Carnegie Mellon University
https://www.mihirbala.com/
Democratizing Drone Autonomy via Edge Computing
Fully autonomous flight by low-cost, lightweight commercial off-the-shelf (COTS) drones could transform many use cases involving real-time computer vision. In this proposal, we propose SteelEagle, a platform that emulates the performance of heavier, more powerful drones on a lighter, cheaper, and more portable frame using edge computing. A key consideration of this system is the use of consumer off-the-shelf (COTS) drones and computing / communication payloads. This approach avoids customization of hardware and modification of privileged software. It thus avoids the need for re-certification (e.g., by the FAA or the FCC). However, a COTS approach also introduces new obstacles.
Thermal limitations of lightweight COTS communications devices pose latency, frame rate, and quality challenges that render traditional sensor offloading techniques ineffective, and force such systems to intelligently manage communication, computation, and prediction. Despite these limitations, we believe that it is feasible to build a capable lightweight COTS drone platform on modern hardware which can equal and in some cases exceed the performance of similar weight devices. Such a platform could expand the use of autonomous drones and drive investment towards lighter, cheaper, and more intelligent aerial vehicles.
Thesis Committee:
Mahadev Satyanarayanan (Chair)
David O'Hallaron
Jeff Schneider
Padmanabhan Pillai (Intel Labs)