This internship will focus on leveraging a human demonstration for teaching a robot how to grasp an object in a particular manner. More precisely, the main idea is to develop a proof of concept where a human demonstrator shows to a robot how to grasp an object in a certain way (for example, with the intention to give a knife to someone else), and the robot should grasp the object in a similar fashion. One major difficulty is the fact that the hand of the human and the gripper of the robot are different, in particular because their numbers of fingers and their degree of freedoms may not be identical. Additionally, it is not reasonable to consider that pairs of corresponding grasps can be annotated at large scale.
One possible solution for this problem is to design a cost function that would score different robot grasps according to their similarity to the human grasp. This cost function could be used to filter robot grasp hypotheses produced by an off-the-shelf technique (e.g. GraspIt, GraspNet, ...), or to produce new ones, by training a machine leaning algorithm to map human grasps to robotic grasps in a weakly-supervised way. The research will focus on the design of such cost function and machine learning algorithm.
Experiments will be conducted both in simulation and on a real robot arm.
NAVER LABS Europe has full-time positions, PhD and PostDoc opportunities throughout the year which are advertised here and on international conference sites that we sponsor such as CVPR, ICCV, ICML, NeurIPS, EMNLP etc.
NAVER LABS Europe is an equal opportunity employer.
NAVER LABS are in Grenoble in the French Alps. We have a multi and interdisciplinary approach to research with scientists in machine learning, computer vision, artificial intelligence, natural language processing, ethnography and UX working together to create next generation technology and services that deeply understand users and their contexts.