Spring 2026 GRASP SFI: Jaimie Carlson, Amazon Robotics, “Autonomous Robotics in Industry at Scale”

This was a hybrid event with in-person attendance in Levine 307 and virtual attendance…

ABSTRACT

While robotics engineering courses often focus on getting a single-robot demonstration to work once, software engineers in industry have to maintain a system with many robots at a high standard for a long period of time. This talk presents some general robotics design principles the author has learned during six years of working at Amazon Robotics, to help achieve the goal of reliable large-scale autonomy. First, large-scale autonomous design often involves co-designing a robot and its environment, and figuring out the tradeoffs of adding complexity to one or the other. Second, it requires constant attention to metrics – where engineers might use the mean error of a system to measure a smaller project, those working on large-scale systems have to evaluate error metrics at the 99th-percentile or higher, and specifically compare those errors to system requirements. Some case studies are presented for debugging “99th percentile” edge cases of emergent behavior in a complex system. Third, best practices for software engineering, like realistic time estimation, code simplicity, and testing in reality and simulation, become especially important in the interdisciplinary field of robotics. Fourth, communication becomes critical: roboticists don’t only write code – they also spend a good deal of time writing design documents, requirement definitions, and correction-of-error documents to navigate multi-layered problems. Whether students choose to pursue robotics in academia or industry, these principles can help them design reliable and understandable systems at scale.