Atlas Fusion 2.0 - A ROS2 Based Real-Time Sensor Fusion Framework

Kód výsledku v IS VaVaI

<a href="https://www.isvavai.cz/riv?ss=detail&h=RIV%2F00216305%3A26220%2F24%3APU151292" target="_blank" >RIV/00216305:26220/24:PU151292 - isvavai.cz</a>

Výsledek na webu

<a href="https://doi.org/10.1007/978-3-031-71397-2_3" target="_blank" >https://doi.org/10.1007/978-3-031-71397-2_3</a>

DOI - Digital Object Identifier

<a href="http://dx.doi.org/10.1007/978-3-031-71397-2_3" target="_blank" >10.1007/978-3-031-71397-2_3</a>

Jazyk výsledku

angličtina

Název v původním jazyce

Atlas Fusion 2.0 - A ROS2 Based Real-Time Sensor Fusion Framework

Popis výsledku v původním jazyce

In this paper, we present a novel, easy-to-use ROS2-based realtime sensor fusion framework capable of making high-level detections from raw sensor data provided by their respective drivers. This framework is a direct successor of Atlas Fusion developed by Brno University of Technology robotics lab. As opposed to its predecessor, it is based on ROS2 and more in line with its philosophy - each functionality is encapsulated in its own process (node). This allows for the composition of a unique sensor-fusion pipeline, code testing in isolation, better profiling, and easier usage of the state-of-the-art ROS2 packages developed by other research teams. Algorithms used are real-time, so the framework can be used in development, simulations (with previously collected dataset), deployed to a physical autonomous agent and the high-level detections can be shared between multiple agents. The Atlas-Fusion 2.0 framework has been developed in a way that allows for code distribution between several physical devices which helps with dividing responsibility and building redundancy into the system. With RVIZ and PlotJuggler, one can visualize every part of the processing chain from raw data up to high-level detections to assess current performance. It also has inbuilt basic profiling capabilities to publish the current delay each algorithm introduces into the system. This framework has been evaluated and tested on a sensory framework used to collect the Brno Urban Dataset and its winter extension. As the boundary of the state-of-the-art algorithms in sensor data processing is pushed rapidly, this package, in our opinion, provides a very streamlied way of experimenting with them and testing their performance.

Název v anglickém jazyce

Atlas Fusion 2.0 - A ROS2 Based Real-Time Sensor Fusion Framework

Popis výsledku anglicky

In this paper, we present a novel, easy-to-use ROS2-based realtime sensor fusion framework capable of making high-level detections from raw sensor data provided by their respective drivers. This framework is a direct successor of Atlas Fusion developed by Brno University of Technology robotics lab. As opposed to its predecessor, it is based on ROS2 and more in line with its philosophy - each functionality is encapsulated in its own process (node). This allows for the composition of a unique sensor-fusion pipeline, code testing in isolation, better profiling, and easier usage of the state-of-the-art ROS2 packages developed by other research teams. Algorithms used are real-time, so the framework can be used in development, simulations (with previously collected dataset), deployed to a physical autonomous agent and the high-level detections can be shared between multiple agents. The Atlas-Fusion 2.0 framework has been developed in a way that allows for code distribution between several physical devices which helps with dividing responsibility and building redundancy into the system. With RVIZ and PlotJuggler, one can visualize every part of the processing chain from raw data up to high-level detections to assess current performance. It also has inbuilt basic profiling capabilities to publish the current delay each algorithm introduces into the system. This framework has been evaluated and tested on a sensory framework used to collect the Brno Urban Dataset and its winter extension. As the boundary of the state-of-the-art algorithms in sensor data processing is pushed rapidly, this package, in our opinion, provides a very streamlied way of experimenting with them and testing their performance.

Druh

D - Stať ve sborníku

CEP obor

—

OECD FORD obor

20204 - Robotics and automatic control

Projekt

<a href="/cs/project/8A21013" target="_blank" >8A21013: Automotive Intelligence for Connected Shared Mobility</a><br>

Návaznosti

P - Projekt vyzkumu a vyvoje financovany z verejnych zdroju (s odkazem do CEP)

Rok uplatnění

2024

Kód důvěrnosti údajů

S - Úplné a pravdivé údaje o projektu nepodléhají ochraně podle zvláštních právních předpisů

Název statě ve sborníku

Lecture Notes in Computer Science

ISBN

9783031713965

ISSN

—

e-ISSN

—

Počet stran výsledku

13

Strana od-do

1-13

Název nakladatele

Springer Nature

Místo vydání

neuveden

Místo konání akce

Palermo, Itálie

Datum konání akce

17. 10. 2023

Typ akce podle státní příslušnosti

EUR - Evropská akce

Kód UT WoS článku

—