Time Predictable and Fault-tolerant Software Framework for Multicore UAVs
- This project is part of a NASA grant, entitled "NASA Learn II:
Multichannel Sense-and-Avoid Radar for Small UAVs. (PI: Dr. Chistopher
Allen)" Our role is on builing a software framework for a
multicore computing platform.
- Sponsor: NASA
Future UAVs will perform complex and computationally demanding
real-time tasks -- such as vision based SLAM, optical flow, and radar
signal processing -- to achieve a high degree of autonomy. To this end,
commodity off-the-shelf (COTS) heterogeneous multicore processors,
which can be found in smartphones and tablets, will play a key role as
they can provide enormous computing capacity without sacrificing size,
weight, and power (SWaP) constraints of UAVs. For example, Nvidia’s
recent mobile processor includes 4 ARM CPU cores, each up to 2.3GHz,
and 192GPU cores that can deliver 365GFLOPs in just 5W power
envelope. We are building a powerful multicore based UAV testbed and a
software framework that provides predictable timing guarantees
among demanding UAV applications.