The "Observe, Orient, Decide, Act" (OODA) loop can be applied to encapsulate the agility of cyber-physical or cyber-human systems that depend on continuous iterations of these steps. Systems with faster OODA loops react more quickly to changes in their environment. This work analyzes the OODA loop of the SteelEagle edge-enabled autonomous drone system, which transforms consumer aerial photography drones into fully autonomous UAVs by offloading computation to the edge using a drone payload with cellular connectivity.

We identify bottlenecks and opportunities for optimization, leading to a faster SteelEagle OODA loop and thus improved performance in active vision tasks such as obstacle avoidance and object tracking. This enables the drone to fly safely at higher speeds in crowded spaces, increasing the practicality of SteelEagle drones in applications such as search and rescue and infrastructure inspection. Our findings show that the hardware encoding of the H.264 video stream on the drone makes up about two-thirds of the drone-to-cloudlet latency.

Because of its dependence on offloading, SteelEagle is currently limited in its ability to operate in degraded network conditions. To mitigate these limitations, we analyze the use of onboard computation with SteelEagle by considering a new payload that can run float16-quantized DNNs. We discuss how onboard computational abilities can be combined with offloading to achieve an optimal system based on computation accuracy, energy efficiency, and latency.

64 pages

Thesis Committee:
Mahadev Satyanarayanan (Chair)
Padmanabhan Pillai

Srinivasan Seshan, Head, Computer Science Department
Martial Hebert, Dean, School of Computer Science

Return to: SCS Technical Report Collection
School of Computer Science

This page maintained by reports@cs.cmu.edu