Offre en lien avec l’Action/le Réseau : BigData4Astro/Doctorants

Laboratoire/Entreprise : Romea team, INRAE Clermont-Ferrand
Durée : 6 mois
Contact : zhongkai.zhang@inrae.fr
Date limite de publication : 2021-11-30

Contexte :
Mobile robot control can be achieved by either model-based or data-driven methods. Model-based methods have stability guarantee, but they need analytical models with a higher accuracy. It is usually difficult to obtain an accurate model for high-speed and off-road mobile robots because of the presence of sliding. Data-driven methods need a huge amount of data instead of an accurate model, but they lack of stability guarantee. It is natural to combine both methods for control design in order to get the advantages of each method. Existing hybrid methods
assume that the data is enough to predict the model, which is usually not guaranteed in real application. Therefore, the main objective of this master project is to investigate a methodology to combine data-driven model(Bayesian neural network) with model-based control(model predictive control) to achieve stable path following tasks, even if the amount of data is not enough to recover the robot model. Results aims at adapting an off-road mobile robot behaviour to the diversity of encountered situations in an agricultural context. The proposed trainee will take part of experiments conducted on different robot available at INRAE.

Sujet :
Hybrid Data-driven/Model-based Methods for Mobile Robot Control

Profil du candidat :
see the attached document

Formation et compétences requises :
Technical Skills: machine learning, control theory, robotics
Software: Python, C++, Pytorch, ROS
Language: English

Adresse d’emploi :
CLermont-Ferrand

Document attaché : 202111290803_Hybrid Data-driven and Model-based Methods for Mobile Robot Control.pdf