Register now After registration you will be able to apply for this opportunity online.
Reinforcement Learning for Human-Robot Collaboration with NVIDIA Isaac Sim
To train reinforcement learning robot policies for human-robot collaborations, this thesis aims at leveraging the Isaac Gym to simulate collaborative scenarios and train robot policies for seamless robot assistance.
Keywords: Robotics, Reinforcement Learning, Deep Learning, Isaac Sim, Simulation, Human Robot Collaboration
To effectively assist humans in shared tasks, robots must understand their environment, the specific task, and respond appropriately to their human collaborators. Recent advancements in reinforcement learning (RL) have demonstrated the potential to develop flexible policies that can efficiently support humans in such scenarios. However, this requires scalable training environments. This thesis seeks to address this need by developing a scalable and dynamic simulation platform for Human-Robot Collaboration using NVIDIA's Isaac Gym. The platform will simulate a range of collaborative scenarios, allowing for the training and evaluation of RL policies tailored to human-robot interactions. To accurately replicate real-world collaborations, the project will involve capturing human task performance with an existing motion capture system and digitalizing the task environment using a computer vision system. These components — human actions, environmental context, the robot, and the task — will be integrated into a unified simulation platform. Once the platform is established, the next step will be to train and benchmark RL policies. The ultimate objective is to create robotic systems capable of seamlessly assisting humans in a variety of collaborative tasks, thereby enhancing both efficiency and safety.
**This thesis is conducted in collaboration with Accenture Digital Experience Labs in Sophia Antipolis.**
To effectively assist humans in shared tasks, robots must understand their environment, the specific task, and respond appropriately to their human collaborators. Recent advancements in reinforcement learning (RL) have demonstrated the potential to develop flexible policies that can efficiently support humans in such scenarios. However, this requires scalable training environments. This thesis seeks to address this need by developing a scalable and dynamic simulation platform for Human-Robot Collaboration using NVIDIA's Isaac Gym. The platform will simulate a range of collaborative scenarios, allowing for the training and evaluation of RL policies tailored to human-robot interactions. To accurately replicate real-world collaborations, the project will involve capturing human task performance with an existing motion capture system and digitalizing the task environment using a computer vision system. These components — human actions, environmental context, the robot, and the task — will be integrated into a unified simulation platform. Once the platform is established, the next step will be to train and benchmark RL policies. The ultimate objective is to create robotic systems capable of seamlessly assisting humans in a variety of collaborative tasks, thereby enhancing both efficiency and safety.
**This thesis is conducted in collaboration with Accenture Digital Experience Labs in Sophia Antipolis.**
After familiarization with the Nvidia Omniverse / Isaac Sim platforms your tasks will include:
- building the core simulation platform
- capturing human task execution (environment and human pose) with a MoCap and CV system
- creating gyms for different tasks
- Implementing and benchmarking RL Algorithms on this platform
- exploring task-related reward functions and their effect on the resulting robot assistance
After familiarization with the Nvidia Omniverse / Isaac Sim platforms your tasks will include:
- building the core simulation platform
- capturing human task execution (environment and human pose) with a MoCap and CV system
- creating gyms for different tasks
- Implementing and benchmarking RL Algorithms on this platform
- exploring task-related reward functions and their effect on the resulting robot assistance
- Strong programming skills (Python, C#, C++, …) - Experience with machine learning, data science or computer vision (Pytorch, OpenCV, …) - The ability to take initiative and shape the direction of the project - Enthusiasm for tackling practical challenges
Not specified
- Collaboration with Accenture Digital Experiences Labs - Master Thesis / Semester Project - Human Robot Collaboration - Reinforcement Learning
Please send your CV and master thesis grades to Sophokles Ktistakis (ktistaks@ethz.ch)
Please send your CV and master thesis grades to Sophokles Ktistakis (ktistaks@ethz.ch)
