PhD position in Machine Learning for Mesh-based Representations of Robots and Deformable Objects

At the Soft Robotics Laboratory, we offer a doctoral position on the topic of real-time mesh-based reconstruction of deformable objects and robots from unstructured point clouds. This position at the Soft Robotics Laboratory (SRL) collaborates with the Swiss Data Science Center (SDSC) and the Computational Robotics Lab (CRL). In this collaboration, we develop an efficient, robust, and real-time mesh-based representation of articulated or deformable objects and robots, with material properties augmented in such a representation. Such a tool would greatly simplify the solution to complex downstream tasks such as real-time scene generation in moving cluttered environments or manipulation of articulated and/or deformable objects. We aim to release generalizable real-time points to mesh reconstruction framework that leads to a leap in performance improvements on existing grasping benchmark tasks as well as our own proposed manipulation benchmark challenges.

This doctoral position will focus mainly on the physics-based complex object simulation and data-driven point cloud to mesh reconstruction aspects, while the manipulation tasks will be aided by colleagues in SRL and CRL. Simulation to reality validation needs to be performed through experimental setups that the candidate should be able to design and use. A strong interest in working hands-on with robotic systems and validating simulations in the real world on robots and complex objects is desirable

As a PhD student, you will develop and publish new software frameworks and their real-world validation. You will regularly present your work at international robotics and machine learning conferences. Your responsibilities will also include supervising bachelor and master students in their thesis works, supporting the Soft Robotics Laboratory in teaching its graduate classes and preparing grant proposals.

You are:

• interested in the computer vision of robots for manipulation and are motivated to independently explore various research fields to combine their knowledge to achieve this goal.
• a diligent worker driven to publish new insights and lead the research community forward by communicating your findings to a smaller community of researchers and a broader public audience.
• curious about novel technologies, learning about different complex objects, and understanding their physical properties.

You persevere through:

• challenges faced throughout the project and are able to quickly adapt when experiments do not deliver the desired results.

You have background knowledge in:

• in object representations, computer graphics/vision, and physics-based simulation. Ideally, you have previously worked with robotics learning and (differentiable) simulations, with hands-on experience manipulating deformable objects, reconstructing mesh representations online, and matching with simulated and/or ground truth shapes.
• Machine learning knowledge, especially in the field of deep learning for computer graphics, can be beneficial. Proficient communication skills in English are required.
• computer science or engineering background, with BSc and MSc degrees or equivalent, in computer science, mechanical or electrical engineering, computational engineering, applied physics, or a related field. Your academic record is outstanding.

ETH Zurich is a family-friendly employer with excellent working conditions. You can look forward to an exciting working environment, cultural diversity, and attractive offers and benefits.