ERI: Autonomous Monitoring and Planning with Lightweight Robotic Teams Operating in Dynamic Aquatic Environments

This Engineering Research Initiation (ERI) award will support fundamental research to advance autonomous environmental monitoring with teams of lightweight robotic systems operating in dynamic aquatic environments. Rivers, lakes, bayous, and coastal waters are increasingly affected by fuel and oil spills, chemical releases, and harmful algal blooms. These events evolve over hours to days, requiring repeated measurements across large areas to determine how contamination is evolving. However, sparse manual sampling can miss important changes, while computation-intensive exhaustive robotic sampling can exceed onboard computing power, battery energy, and communication range available to small autonomous systems. To address these challenges, this project seeks to develop methods and tools that enable teams of autonomous surface vehicles to make informative sampling decisions directly on the vehicles and coordinate with limited communication. The resulting advances will support faster, safer, and more repeatable monitoring of water resources and serve the national interest by promoting scientific discovery, advancing environmental stewardship, and supporting public health. The project will also broaden impact through open-source software and datasets, new educational materials in computer science and reinforcement learning, research mentoring for undergraduate and graduate students, robotics workshops for regional teachers, summer outreach for K-12 students, and dissemination of results through publications and a workshop at a major robotics conference. The objective of this project is to create lightweight methods and cost-effective algorithms that enable teams of autonomous surface vehicles to gather informative measurements in nonstationary spatiotemporal environmental fields. Specifically, the research seeks to enable robust, resource-efficient operation by allowing robots to estimate changing environmental fields, select informative sampling locations, and coordinate despite constraints on computation, energy, and communication. In pursuit of this objective, the project will make three main contributions: (i) developing low-rank spatiotemporal field models and information-theoretic candidate selection methods for efficient estimation of evolving environmental fields; (ii) designing compact information-state-driven decision policies that select informative sampling locations in real time and can be compressed for resource-constrained robotic platforms; and (iii) creating decentralized coordination methods that use limited-range communication to reduce redundant sampling without centralized control. The work will include theoretical analysis, field-informed simulations, and multi-day field deployments using autonomous surface vehicles carrying multimodal water quality sensors. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.