Snabbfakta
-
- Hauts-de-Seine
Ansök senast: 2024-11-25
CIFRE PhD-Autonomous reinforcement learning quadrotor navigation for complex industrial environment F/ H
Description de la mission
This CIFRE PhD deals with robotics drone navigation with machine learning.
In recent years, the use of drones in various industrial sectors has seen significant growth due to their potential to improve efficiency, safety, and data collection capabilities. However, navigation in complex indoor environment such as factories, in an autonomous fashion raises many complex limitations, including localization, obstacle avoidance, and autonomous planning. Some of these limitations are being addressed in part by the academic sector, such as approaches for using YOLO for inspecting power lines and animals, or for avoiding complex and obstacles rich environments, while also out matching the state-of-the-art in drone control by human and optimal control systems.
The applications, extensions, and fusions of these methods into a comprehensive AI based control and navigation system is next step for autonomous drones for advanced industrial applications. One of these applications that this PhD project seeks to address is the automated inspection of forged ingots for defects and quality control, as the quality assurance for nuclear components follow strict guidelines and standards. Currently this task is done with a human in the loop, however it remains time consuming and challenging. By using a drone system, this inspection could be achieved in a time efficient manner, that can focus as needed on specific locations of the ingot with a higher degree of accuracy, and navigate around it autonomously in order to complete its task and return to base. Future perspectives of this subject include a study of swarm drones, as cooperative navigation and task planning would allow for an increase in the productivity and speed of the overall system, while allowing for complex surveillance or monitoring of sensitive and restricted sites. For this, advanced methods of robotics and deep learning will be employed to achieve the desired result. In particular, Reinforcement learning for task planning and accurate control will be a major focus, thanks to their high generalization capabilities and high performance for a desired target environment.
This PhD is funded by CIFRE, with a joint supervision between Framatome and Central Supelec (SYCOMORE laboratory).
Profil
· Strong background in robotics, computer science, machine learning, embedded systems, or a related field;
· Proficiency in programming languages such as Python or C/C++;
· Good analytical and problem-solving skills, with a strong passion for discovery and cutting-edge research;
· Effective communication skills in English, and the ability to work both independently and as part of a multidisciplinary team;
· Experience with robotics platforms, sensors, deep reinforcement learning, advanced mathematics, or the Robotic Operating System (ROS) would be ideal.
Critères candidat
Niveau d'études min. requis
Bac+5
Niveau d'expérience min. requis
Jeune diplômé
Niveau d'emploi
Cadre
