Regent State Estimation Intern at REGENT will research and design multi-agent information fusion algorithms for Seaglider fleets. The role involves building simulations in Python or MATLAB and validating algorithms with real and simulated data.
Responsibilities
REGENT's perception team is taking on a new kind of problem: instead of relying on a single Seaglider's sensors to understand its surroundings, what if a fleet of Seagliders could share what they see and build a more complete picture together? That's the question at the center of this internship. You'll join the Perception team to study existing approaches to multi-agent information fusion, then design or adapt an algorithm suited to how our platform actually operates, and demonstrate it against simulated and, where available, real data.
This is genuinely greenfield work. There's no existing playbook to extend here, which means you'll have real room to shape the direction of the project, backed by a defined deliverable, a named mentor, and weekly check-ins so you're never working in a vacuum. Your simulation work will be grounded in real platform and sensor characteristics so your results carry meaning beyond the page, and at the end of your term you'll present your findings directly to the perception engineers who would build on them next.
If you're the kind of person who gets energized by open problems rather than intimidated by them, and you want to see your research connect to an actual vehicle system rather than stay theoretical, this is a rare chance to help define a new area of perception work at a company reinventing how we move across water.
Survey the Landscape: You'll research existing approaches to multi-agent and distributed information fusion to understand what's been tried and what might translate well to REGENT's Seagliders.
Design a Fusion Approach: You'll propose or adapt an algorithm for combining sensor data across multiple Seagliders into a shared, more complete picture of the environment.
Build Grounded Simulations: You'll build simulation environments in Python or MATLAB that reflect real platform and sensor characteristics, so your results are meaningful beyond a theoretical exercise.
Test and Validate: You'll test your algorithm in simulation and, where available, against real data, refining your approach based on what you find.
Work Closely With a Mentor: You'll partner with a dedicated mentor through weekly check-ins, getting guidance and feedback as you navigate genuinely new territory for the team.
Communicate What You Learn: You'll translate your technical work into clear findings that others on the team can understand and build on.
Present to the People Who'll Use It: At the end of your term, you'll present your algorithm and results directly to the perception team that could put this work into practice next.
What You'll Bring
Strong Math Fundamentals: You have a solid foundation in probability theory, statistics, and linear algebra, the toolkit you'll need to engage with multi-agent estimation problems.
