Cartographie en temps réel d'un environnement industriel avec robot

Toulouse CDD (36 mois)

À propos de CESI LINEACT

L’objectif de la recherche à CESI est de produire des connaissances et des méthodes au service de la communauté scientifique.

Son organisation actuelle est liée aux domaines de formation et répond à un besoin d’identification au sein de l’écosystème recherche. Elle met ses résultats au service des entreprises partenaires de CESI, et contribue dans ses domaines, à relever les défis technologiques posés aux territoires qui l’accueillent et la nourrissent.

Elle est également au service des apprenants CESI pour participer à leur formation par la recherche.

LINEACT CESI a été labellisé Équipe d’Accueil – EA 7527 – pour la qualité des travaux de recherche de ses équipes par le Ministère de l’Enseignement Supérieur, de la Recherche et de l’Innovation.

Le poste

In the context of industry 5.0 with human and robots sharing the same working area, the need of an updated map with the exact position of static and dynamic elements is required [1]. Indeed, to avoid collision and make the human-robot collaboration possible, the information of the position of each element is needed. It is necessary that a robot have a map of its closest dynamic environment [2] to manage its trajectory. Today, in the literature [1, 2], there are lots of works for static scenes but only few works have addressed the problem of the map representation in dynamic scenes [3]. In recent years, there has been a growing interest in adding high-level information to many robotic applications to achieve more efficient robots with a diversity of planned tasks, even capable of reacting to unexpected events. To do so, the field of mobile robotics is starting to include semantic information in navigation tasks, leading to a new concept: semantic navigation [1]. This type of navigation brings the human way of understanding the environment closer to the way robots understand it [2,4]. Semantic navigation approaches follow some common principles, including a framework for topological mapping that includes geometric information, adding a topological abstraction.

 In the case of displacement of humans and robots in the same area, the map of the environment should be updated at a high frequency level. One of the major challenges of creating dynamic map is the real-time constraint [5, 6, 7]. Indeed, in addition of the time required for the data processing and data fusion, latency periods should be considered. In one part, the data processing time is considerable since there are different kinds of data: video stream (from static camera and robot’s camera), position, velocity, distance measures (from ultrasound Lidar, odometers, cameras, INS, laser, ...) [8, 9, 10]. Digital twins will be used to perform the data processing and map computation. In another part, transmission times between sensors and digital twin should be taken in consideration in particular if augmented reality is used as one of the data providers. A tradeoff should be studied between the map execution time and the precision of the map (number of data, sensors precision and algorithm performance) [11, 12] with a considerable high reliability index to ensure the security of people.

This thesis aims at tackling the challenges specific to the industries 5.0 and Human-Robot collaboration. The work goal is to provide a real-time map of human-robot industrial environment. The number of use cases is very large. Obviously, these maps will be used for the planification of robot tasks providing safety to human and robots, sharing the same working space. We can imagine that this kind of map will be used to provide an inventory in case of a fire or accident.

Contributions on the following topics are expected:

·       Develop a data fusion algorithm to be used in our context

·       Build a semantic map using different kinds of information (provided by different static and dynamic sources).

·       Respect real-time constraints in considering data processing time and latency periods. 

·       Provide a reliability index associated to the map

The PhD thesis major steps can be described as follows:

·       State of the Art of a real-time data processing, data fusion, 2D/3D mapping methods and static/dynamic environment.

·       Development of a data fusion algorithm

·       Development/Usage of a static element mapping model.

·       Contribution in a real time dynamic elements mapping model.

·       Experimentation, simulation and validation of the model (reliability study, ...).

The expected deliverables are:

·       At least, two communications in major conferences and one JCR journal paper.

·       The deployment of the algorithms on LINEACT platforms.

·       An evaluation of the algorithms performance on a real system.

Profil recherché

Hard skills:

·       Industrial robotics

·       Good programming skills

·       Knowledge in ROS and Gazebo

·       Data processing

·       Notion in real-time issues is a must-have

·       Scientific writing

·       Good English speaking and writing proficiency

Soft skills:

·       Autonomy

·       Adaptability

·       Communication

·       Creativity

·       Problem solving

·       Teamwork

CDD (36 mois)
Propulsé parTaleez