News

Speaker: Stefano Carpin (University of California, Merced)

Monday, 6 July 2026, 11:30 — Sala Consiglio, Floor 8, Dept. of Computer Science, Via Celoria 18.

Abstract

Autonomous robots are increasingly used for data collection and monitoring in precision agriculture, but enabling non-specialist growers to assign them complex tasks remains a challenge. This talk presents our recent work using large language models (LLMs) to bridge that gap. First, we describe a system that translates natural-language mission requests into robot-executable plans, following the IEEE Standard for Robot Task Representation and executed on a field robot via ROS2. Testing on an orchard testbed shows that LLMs can handle many missions well but may struggle with spatial reasoning and optimization problems, a gap we close by embedding a hand-designed solver into the pipeline. Owing to the inherent ambiguity of natural language, we introduce a second architecture that formally verifies whether a generated mission plan actually matches the user’s intent, using an independent LLM to produce a linear temporal logic specification checked against the plan via the SPIN model checker. Evaluated on both lab missions and real prompts from farm workers, the system detects and self-corrects misalignments without human intervention, moving toward robots that growers can instruct in plain language with confidence that the robot will do what was asked.

Speaker bio

Stefano Carpin is a Professor of Computer Science and Engineering and Associate Dean for Research and Graduate Programs at UC Merced. He earned his Laurea (MSc) and Ph.D. in electrical engineering and computer science from the University of Padova, Italy (1999, 2003), and held faculty positions at Jacobs University Bremen before joining UC Merced in 2007, where he founded and leads the robotics laboratory. He chaired the Department of Computer Science and Engineering from 2018 to 2021 and currently serves as site director for the NSF Engineering Research Center for the Internet of Things for Precision Agriculture. His research focuses on mobile and cooperative robotics and robot algorithms, with over 150 publications. He is a Senior Member of the IEEE and has served as associate editor for several top robotics journals and conferences. His teams won first place in the 2009 RoboCupRescue Virtual Robots competition, and he received the Best Conference Paper Award at CASE in 2018. His research has been funded by NSF, DARPA, USDA, the Office of Naval Research, the Army Research Lab, NIST, CITRIS, Microsoft Research, and General Motors.