Projects
Embodied Learning for Adaptive Control
The physical properties of embodied agents can have a substantial impact on learning and control outcomes. The physics of sensors, the structure of the environment, and correlations introduced by the dynamics of robot bodies are all crucial to an agent's task-capability. In my work, I design algorithms that exploit and adapt to the physical properties of embodied agents.
Related Publications
Maximum diffusion reinforcement learning
Thomas A. Berrueta, Allison Pinosky, Todd D. Murphey
Nature Machine Intelligence, 2024.
Active learning in robotics: A review of control principles
Annalisa Taylor, Thomas A. Berrueta, Todd D. Murphey
Mechatronics, vol. 77, 102576, 2021.
Experimental applications of the Koopman operator in active learning for control
Thomas A. Berrueta, Ian Abraham, and Todd D. Murphey
The Koopman Operator in Systems and Control, 2020.
Task-Constrained Optimization of Physical Intelligence
The physical intelligence of biological organisms supports and enhances their cognitive abilities by offloading computational burdens to the material make-up of agents. There is a need for task-speficic design principles capable of accounting for and realizing physical intelligence in robotic systems. My work explores novel design principles for exploiting synergies between the material and information-processing capabilities of robotic agents.
Related Publications
Materializing autonomy in soft robots across scales
Thomas A. Berrueta, Todd D. Murphey, Ryan L. Truby
Advanced Intelligent Systems, vol. 6, no. 2, 2300111, 2024.
Emergent microrobotic oscillators via asymmetry-induced order
Thomas A. Berrueta, Jing Fan Yang, Allan M. Brooks, Albert Tianxiang Liu, Ge Zhang, David Gonzalez-Medrano, Sungyun Yang, Volodymyr B. Koman, Pavel Chvykov, Lexy N. LeMar, Marc Z. Miskin, Todd D. Murphey, Michael S. Strano
Nature Communications, vol. 13, 5734, 2022.
Memristor circuits for colloidal robotics: Temporal access to memory, sensing, and actuation
Jing Fan Yang, Albert Tianxiang Liu, Thomas A. Berrueta, Ge Zhang, Allan M. Brooks, Volodymyr B. Koman, Sungyun Yang, Xun Gong, Todd D. Murphey, Michael S. Strano
Advanced Intelligent Systems, vol. 4, no. 4, 2100205, 2022.
Control of Robot Collectives Across Scales
Interactions between agents in a collective can form the basis of physical intelligence and emergent capabilities. By self-organizing their behavior from fully-distributed and simplistic interactions, collectives in nature are capable of robustly surpassing the computational limitations of individuals. By taking advantage of emergence and self-organization, my work develops control principles for robot collectives that work across scales.
Related Publications
Low rattling: A predictive principle for self-organization in active collectives
Pavel Chvykov, Thomas A. Berrueta, Akash Vardhan, William Savoie, Alexander Samland, Todd D. Murphey, Kurt Wiesenfeld, Daniel I. Goldman, and Jeremy L. England
Science, vol. 371, no. 6524, 2021.
Information requirements of collision-based micromanipulation
Thomas A. Berrueta, Alexandra Q. Nilles, Ana Pervan, Todd D. Murphey, and Steven M. LaValle
Proceedings of the Workshop on the Algorithmic Foundations of Robotics (WAFR), 2020.
A robot made of robots: Emergent transport and control of a smarticle ensemble
William Savoie, Thomas A. Berrueta, Zachary Jackson, Ana Pervan, Ross Warkentin, Shengkai Li, Todd D. Murphey, Kurt Wiesenfeld, and Daniel I. Goldman
Science Robotics, vol. 4, no. 34, 2019.