Annika Thomas

Hey! My name is Annika Thomas and I’m in my 5th year of pursing my Ph.D. at Massachusetts Institute of Technology where I work with Jonathan P. How in the Aerospace Controls Laboratory. Prior to joining MIT, I obtained a B.S. in Mathematics and Physics from The College of Idaho and a B.S. in Mechanical Engineering from Columbia University in the City of New York.

My primary research interests are centered around robotic perception in multi-agent systems. My research is driven by a desire to create environment representations that are both compact and context-rich, with the ultimate goal of creating a platform for autonomous systems to meaningfully understand and interact with the environment. Recently, my research has focused on these key areas:

• Perception and Localization: I am investigating ways to improve environment representations for autonomous agents to efficiently and compactly localize in open-set environments: SOS-Match (IROS 2024), VISTA (RA-L 2026), ROMAN (IROS Long-Term Localization Workshop 2024; RSS 2025), PUMA (ICRA 2024), Object-Based Global Localization in Unstructured Environments (IROS 2023). I also work on dense, scalable scene representations for large environments, including GRAND-SLAM (RA-L 2025), which targets high-fidelity reconstruction and mapping in multi-agent settings. I am one of the lead organizers of the MIT Perception and Localization Seminar.
• Autonomy in Space: I am interested in the applications of robust perception and mapping methods for applications in extreme environments, such as on the moon and Mars. I integrated multi-agent operations into a systems architecture for a long-duration simulation on the moon for our award-winning NASA RASC-AL concept MARTEMIS and led a team to develop navigation and perception algorithms for NASA’s Lunar Autonomy Challenge, where we placed second. Building on this work, we are currently developing LunarLoc (RSS Workshop on Resilient Off-road Autonomous Robotics 2025, IEEE Aerospace 2026), a perception and localization framework for long-horizon, multi-agent lunar navigation. I was invited to speak about enabling collaborative perception on Mars at TEDxMIT and how robotics will enable construction on the moon at TEDxBoston.
• Compute-Aware and Responsible AI for Robotics: An overarching theme in my work is designing perception and mapping systems that scale responsibly with compute. Many modern approaches rely on dense, global optimization at every timestep, which becomes impractical for long-horizon autonomy, multi-agent systems, and deployment on power- and compute-constrained platforms. I am interested in compute-adaptive representations and optimization strategies that allocate resources based on visibility, uncertainty, and task relevance, enabling rich world models that remain efficient, stable, and deployable at scale.

I am also passionate about advocating for underrepresented minorities in STEM. I regularly give talks, mentor students, and engage in outreach initiatives to empower and support the next generation of women pursuing careers in science, technology, engineering, and mathematics.