RoboARCH: Robotics Acceleration with Computing Hardware

Robotics is pushing the limits of conventional computing. Autonomous robots must operate untethered in dynamic and unpredictable environments, requiring many robotics software applications to run online in real-time. Conventional CPU systems are proving unable to deliver the high performance needed by essential latency-critical robotics applications. This is a call to action for researchers across academia and industry: we must leverage nontraditional computing hardware (e.g., custom accelerator ASICs, FPGAs, and GPUs) and navigate enormous design spaces spanning across algorithms, hardware, and physical robot parameters in order to design new high performance systems enabling critical tasks in robotics.

This workshop aims to gather pioneers and innovators working at the intersection of robotics and computer architecture, and to provide an introduction to this exciting emerging field to the computer architecture community. Topics of interest include accelerators and systems for computer vision, mapping, localization, motion planning, control, and end-to-end learning, for all robotics systems, including drones, autonomous vehicles, satellites, submersibles, manipulators, quadrupeds, humanoids, and more!

In particular, the objectives of this workshop are to:

• Present thought-provoking talks spanning across the state-of-the-art of robotics acceleration;

• Provide an introduction to this research topic for the computer architecture community; and

• Encourage open-ended discussion and spark new collaborations using a workshop format with brief small-group breakout sessions between talks (suitable for both in-person and virtual participants).