Skip to Main content Skip to Navigation

Perception et génération de trajectoires en cobotique dans des environnements dynamiques

Abstract : The appearance in the early 2010s of collaborative robots in parallel with the development of industry 4.0 prompted manufacturers to rethink robotics. Theintegration of robots in contact with humans makes it possible to combine their strengths in terms of flexibility and adaptability to complex tasks. However, many challenges are posed to the robotics community by these new practices and in particular: howto secure the interaction between Human and Robot. We propose to address this problem by proposing a complete pipeline allowing the robot to move autonomously in the space shared with humans while guaranteeing its safety. For this we build a representation of the robot's workspace in the form of a 3D semantic occupation grid from the robot's perception of its environment. We then project in real time the obstacles and humans present in the robot's work area to the configurationsspace, a representation adapted to robotic arms. Using the medialaxis in the configurationsspace, the robot's trajectory towards the objective is calculated as far as possible from obstacles. We adapt this trajectory to the dynamic change of the environment and modulate its speed according to the distance separating the human and the robot. The interest of this method has been demonstrated in simulation and with real experiments on collaborative robots in an industrial context.
Document type :
Complete list of metadatas

Cited literature [108 references]  Display  Hide  Download
Contributor : Abes Star :  Contact
Submitted on : Thursday, February 20, 2020 - 9:07:07 AM
Last modification on : Friday, February 21, 2020 - 1:37:04 AM
Long-term archiving on: : Thursday, May 21, 2020 - 1:12:06 PM


Version validated by the jury (STAR)


  • HAL Id : tel-02485229, version 1



Guillaume Fuseiller. Perception et génération de trajectoires en cobotique dans des environnements dynamiques. Automatique / Robotique. Université de Limoges, 2019. Français. ⟨NNT : 2019LIMO0103⟩. ⟨tel-02485229⟩



Record views


Files downloads