- Grenoble, France
- +33 7 82 69 48 20
- Date of birth: 14 / 09 / 1990
- Place of birth: El-hofof, Kingdom of Saudi Arabia
- Nationality: Egyptian
In the context of transfer learning techniques, the main objective of the thesis is to learn Human-Robot Interaction task models usable in planning systems embedded in a robot architecture.
The learning techniques device will be evaluated in carefully devised situations of Human-Robot Interaction on two available robotics platforms: I-Cub (GIPSA-lab) and Baxter (LIG-lab).
Taking advantage of the availability of these two robots with different sensorimotor capabilities, we will also explore transfer learning techniques between robots.
In this internship, I am exploring the topic of "representational re-description". What we want to make the robot use the experience that it learned on one task, in order to accelerate the learning of a new task. We believe that it is essential for the robot to adapt the representation of its previous knowledge in order to suit the new task. In this internship, I explore different techniques in order to achieve this.
Following an elite research work about robot damage recovery - was published in Nature in 27th of may 2015 - , my task was to investigate how to extend this work from the single objective optimization domain to multiple objective optimization domain. During this work, I gained decent experience in the area of "expensive multiple objective optimization", and "surrogate model based optimization", and the latest challenges in this field. We also tried many machine learning techniques in order to overcome the complex problems rising from the shift room single objective to multiple objective.
Professor of Computer Science - Hubert Curien Lab
Director of the Computer Science and Image Processing Department
Dr. Mohamed Abdel Salam
Team Manager, Mentor Graphics