Application of the CMCDOT framework to embedded perception and navigation for indoor robotics - Inria - Institut national de recherche en sciences et technologies du numérique Accéder directement au contenu
Vidéo Année : 2021

Application of the CMCDOT framework to embedded perception and navigation for indoor robotics

Afficher 

Résumé

The CMCDOT framework is a broad perception system, based on Bayesian filtering of dynamic occupancy grids (CMCDOT), allowing parallel estimation of occupancy probabilities for each cell of a grid, inference of velocities, collision risk prediction and dynamic object segmentation. From various heterogeneous sensor data, ground form is estimated, instantaneous occupancy grids are generated and filtered using hybrid sampling methods (classical occupancy grids for static parts, particle sets for parts dynamics) into a Bayesian unified programming formalism. Based on this perception framework, navigation systems have been developed and integrated, allowing path finding-and-following, dynamic obstacle avoidance, localization, thus automation of various mobile robots. Also included are communication tools, allowing data fusion from infrastructure systems. The software is composed of ROS packages, which encapsulate the optimized core system on GPU Nvidia (Cuda), allowing real-time application on embedded boards (Tegra X2). First developed in an automotive setting, it is now exploited in other areas of mobile robotics, and are particularly suited to highly dynamic and uncertain environment management. Thanks to an important engineering support over the years (notably thanks to IRT Nanoelec), this software has grown to be a core research and development tool of the team, an important technology demonstration and transfer vector, through maintained experimental platforms (most notably automated Zoe) and associated research contracts and software licensing with industrial partners.

Dates et versions

hal-04100016 , version 1 (15-06-2023)

Licence

Copyright (Tous droits réservés)

Identifiants

  • HAL Id : hal-04100016 , version 1

Citer

Thomas Genevois, Lukas Rummelhard, Christian Laugier. Application of the CMCDOT framework to embedded perception and navigation for indoor robotics. 2021. ⟨hal-04100016⟩
73 Consultations
10 Téléchargements

Partager

Gmail Facebook X LinkedIn More