Research Intern: Scaling Multimodal Robot Policy Learning for Dexterous Manipulation

Honda Research Institute USA; San Jose (CA); Presencial; USD 34.000 - 44.000; Jornada completa

Descripción de la vacante

Honda Research Institute USA in San Jose, CA, seeks a self-motivated research intern to advance robot policy learning for multi-fingered dexterous manipulation using multimodal sensory data. The internship will explore scalable data acquisition and multimodal representation to develop policies that generalize across tasks and environments.

You will work with real robot hardware and publish results when applicable.

Formación

• Ph.D. or highly qualified MS candidate in robotics, computer science, or a related field.
• Experience with representation learning using multimodal robot data (e.g., RGB, RGB-D, and tactile).
• Experience with policy deployment on real-world robots.
• Experience with robotic simulators (e.g., Isaac Sim or MuJoCo).
• Proficient with Python, C++, and PyTorch.
• Experience with the Robot Operating System (ROS 2)

Responsabilidades

• Explore cost-effective and scalable data collection strategies for real robot hardware, complemented by simulation where beneficial.
• Develop and evaluate training approaches that leverage heterogeneous and multimodal data to improve policy generalization across tasks and environments.
• Iteratively investigate the interplay between data collection and model training, evaluating how data design choices affect downstream policy performance.
• Implement baselines and perform benchmark evaluations on hardware and in simulation where applicable.
• Document findings and contribute to internal research reports.
• Contribute to the portfolio of patents and publish research results when applicable at top-tier conferences and journals in robotics and machine learning.

Conocimientos

Python

C++

PyTorch

ROS 2

Multimodal learning

Policy deployment

Educación

Ph.D. or MS candidate in robotics or CS

Herramientas

Isaac Sim

MuJoCo

Descripción del empleo

Research Intern: Scaling Multimodal Robot Policy Learning for Dexterous Manipulation - Honda Research Institute USA

Research Intern: Scaling Multimodal Robot Policy Learning for Dexterous Manipulation

Job Number: P26INT-23

Honda Research Institute USA (HRI-US) is seeking a self-motivated research intern to advance robot policy learning for multi-fingered dexterous manipulation using multimodal sensory data. The internship will explore generalizable policy learning from a data-scaling perspective, including scalable data acquisition and multimodal representation and policy learning. The goal is to develop dexterous manipulation policies that generalize across diverse tasks and environments, including precision-demanding and contact-rich tasks. The successful candidate will work with real robot hardware to implement, evaluate, and iteratively refine the proposed approach.

San Jose, CA

Key Responsibilities

• Explore cost-effective and scalable data collection strategies for real robot hardware, complemented by simulation where beneficial.
• Develop and evaluate training approaches that leverage heterogeneous and multimodal data to improve policy generalization across tasks and environments.
• Iteratively investigate the interplay between data collection and model training, evaluating how data design choices affect downstream policy performance.
• Implement baselines and perform benchmark evaluations on hardware and in simulation where applicable.
• Document findings and contribute to internal research reports.
• Contribute to the portfolio of patents and publish research results when applicable at top-tier conferences and journals in robotics and machine learning.

Minimum Qualifications

• Ph.D. or highly qualified M.S. candidate in robotics, computer science, or a related field.
• Experience with representation learning using multimodal robot data (e.g., RGB, RGB-D, and tactile).
• Experience with policy deployment on real-world robots.
• Experience with robotic simulators (e.g., Isaac Sim or MuJoCo).
• Proficient with Python, C++, and PyTorch.
• Experience with the Robot Operating System (ROS 2)

Bonus Qualifications

• Experience with in-hand dexterous manipulation on multi-fingered robot hands.
• Experience with scalable or low-cost data collection for robot learning (e.g., teleoperation systems, human demonstration capture, or wearable and handheld devices).
• Experience training robot policies on large-scale or heterogeneous datasets (e.g., human videos, cross-embodiment data, or mixed real and simulated data).
• Experience with multimodal sensing for manipulation, particularly tactile and force sensing.
• Experience with large vision-language-action (VLA) models or other robot foundation models.
• Experience with sim-to-real approaches.

Years of Work Experience Required 0

Desired Start Date May 2027

Internship Duration 3 Months

Position Keywords Dexterous manipulation, Embodied AI, Robot Foundation models