Safety Guarantee Skill Discovery for Visual

Programming manipulation behaviors has become increasingly difficult with the growing number and complexity of manipulation tasks, particularly in a dynamic, unstructured environment. Recent progress in unsupervised skill discovery algorithms has shown great promise in learning an extensive collection of behaviors without extrinsic supervision while assuming oracle state estimation of the surrounding environment.
As skill discovery methods typically encourage exploratory and dynamic behaviors, it can often be the case that a large part of learned skills remain too dangerous and unsafe.
In this context, we have introduced the novel problem of Safe Guarantee Skill Discovery (SGSD) [1], which aims at learning, in a task-agnostic fashion, a repertoire of reusable skills that is inherently safe to compose for solving downstream tasks.
In this internship, we will extend and evaluate SGSD on manipulation tasks of daily life environments such as kitchen scenarios, and explore methods for learning such safe skills beyond safety-critic [2,3] leveraging visual input [4] to enhance the usability of our approach.
The goal of the internship is to extend the Safe Guarantee Skill Discovery approach to visual-based tasks to enhance its usability in daily life manipulation scenarios.

Reading and understanding the NAVER LABS global Safety Guarantee Skill Discovery approach.
Participating in the development of the NVIDIA Isaac Gym environment for visual manipulation on our Franka Emika Panda platform.
Adapting the Safety Guarantee Skill Discovery approach to visual manipulation tasks. This will include original methodological research contributions.
Experiments, evaluation.
Participating in the writing of a research paper.

**Supervisor**: Julien Perez and David Emukpere (NAVER LABS Europe) and Taeyoon Lee (NAVER LABS Korea)
Required skills
- Strong competencies in differential programming languages such as PyTorch and Tensorflow.
- Relevant experience in machine learning, and reinforcement learning is a strong plus
- Ability to propose and implement research ideas
- Team player

Optional skills
- Experience in robotics is a plus

References
[1] Safety-Guaranteed Skill Discovery for Robot Manipulation Tasks - Sunin Kim, Jaewoon Kwon, Taeyoon Lee, Younghyo Park and Julien Perez.

**[2] Learning to be Safe**: Deep RL with a Safety Critic - Krishnan Srinivasan, Benjamin Eysenbach, Sehoon Ha, Jie Tan and Chelsea Finn. 2020.
[3] Constrained Markov Decision Processes. Eitan Altman, 1999.

**[4] RMA**: Rapid Motor Adaptation for Legged Robots - Ashish Kumar, Zipeng Fu, Deepak Pathak and Jitendra Malik, RSS, 2021
Application instructions
Please note that applicants must be registered students at a university or other academic institution and that this establishment will need to sign an 'Internship Convention' with NAVER LABS Europe before the student is accepted.

About NAVER LABS

NAVER is the #1 Internet portal in Korea with activities that span a wide range of businesses including search, commerce, content, financial and cloud platforms.

NAVER LABS, co-located in Korea and France, is the organization dedicated to preparing NAVER’s future. NAVER LABS Europe is located in a spectacular setting in Grenoble, in the heart of the French Alps. Scientists at NAVER LABS Europe are empowered to pursue long-term research problems that, if successful, can have significant impact and transform NAVER. We take our ideas as far as research can to create the best technology of its kind. Active participation in the academic community and collaborations with world-class public research groups are, among others, important tools to achieve these goals. Teamwork, focus and persistence are important values for us.

NAVER LABS Europe is an equal opportunity employer.