Research Intern, Continual Adaptation for Robotics
At Toyota Research Institute (TRI), we envision a future where technology helps humanity thrive. Our world-class teams work across disciplines and collaborate with leading universities to advance research in AI, human happiness, robotics, driving, and materials science, helping shape the future of mobility, amplify human ability, and make our lives safer and more sustainable.
This is a paid 12-week internship opportunity and is a hybrid, in-office role.
Here’s a glimpse into the Internship experience from some of our TRI interns!
The Opportunity
We are seeking a research intern to advance continual adaptation for robot policies. Our goal is to enable robots to improve from experience, personalize their behavior, preserve learned capabilities, and build on prior learning to acquire new behaviors more efficiently.
You will work closely with our robotics and machine learning researchers on a project shaped around your interests, expertise, and the team’s priorities. The work will combine algorithm development with experiments in simulation and on real robots, emphasizing reliable adaptation with limited robot experience and human effort.
The Research
We are interested in how robots can improve through ongoing experience while preserving and building on what they have learned. Your project could explore one of the following directions:
Continual learning: Preserve existing capabilities through successive adaptations and use prior experience to learn new behaviors more efficiently. Investigate how useful adaptations can be retained, combined, and reused, including settings where the original training data are unavailable.
Efficient learning and adaptation: Enable robots to improve with limited experience by identifying where adaptation is needed and making effective use of prior knowledge, new interactions, and guidance.
Personalization from human guidance: Use demonstrations, corrections, or preference feedback to change how a robot performs a task while maintaining task success. Investigate when guidance is sufficient and when further learning is needed.
Possible approaches include reinforcement learning, imitation learning, agentic tool use, modular policies, and distillation, depending on the research question.
Required Languages
🇬🇧 English