Difference between revisions of "RTAB-Map"
Line 5: | Line 5: | ||
Loop closure detection is the process involved when trying to find a match between the current and a previously visited locations in SLAM (Simultaneous Localization And Mapping). | Loop closure detection is the process involved when trying to find a match between the current and a previously visited locations in SLAM (Simultaneous Localization And Mapping). | ||
Over time, the amount of time required to process new observations increases with the size of the internal map, which may affect real-time processing. | Over time, the amount of time required to process new observations increases with the size of the internal map, which may affect real-time processing. | ||
− | RTAB-Map is a novel real-time loop closure detection approach for large-scale and long-term SLAM. Our approach is based on efficient memory management to keep computation time for each new observation under a fixed time limit, thus respecting real-time limit for long-term operation. Results demonstrate the approach's adaptability and scalability using | + | RTAB-Map is a novel real-time loop closure detection approach for large-scale and long-term SLAM. Our approach is based on efficient memory management to keep computation time for each new observation under a fixed time limit, thus respecting real-time limit for long-term operation. Results demonstrate the approach's adaptability and scalability using two custom data set and five standard data sets. |
</english><french> | </english><french> | ||
== Description == | == Description == | ||
− | La détection de fermeture de boucle est le processus impliqué en SLAM (localisation et cartographie simultanées) lorsqu'on tente de trouver une correspondance entre un endroit présent et un autre déjà visité. Plus la carte interne augmente en taille, plus le temps requis pour la détection de fermeture de boucle augmente, ce qui peut affecter le traitement en temps réel. RTAB-Map est une nouvelle approche de détection de fermeture de boucle fonctionnant en temps réel pour du SLAM à grande échelle et à long terme. Notre approche est basée sur une gestion efficace de la mémoire afin de garder le temps de calcul en dessous d'un seuil de temps, respectant ainsi la limite de temps réel à long terme. En utilisant | + | La détection de fermeture de boucle est le processus impliqué en SLAM (localisation et cartographie simultanées) lorsqu'on tente de trouver une correspondance entre un endroit présent et un autre déjà visité. Plus la carte interne augmente en taille, plus le temps requis pour la détection de fermeture de boucle augmente, ce qui peut affecter le traitement en temps réel. RTAB-Map est une nouvelle approche de détection de fermeture de boucle fonctionnant en temps réel pour du SLAM à grande échelle et à long terme. Notre approche est basée sur une gestion efficace de la mémoire afin de garder le temps de calcul en dessous d'un seuil de temps, respectant ainsi la limite de temps réel à long terme. En utilisant cinq ensembles de données standards, notre propre ensemble de données dérivées d'un parcours de plus de 2 km rassemblant des conditions diverses et notre ensemble de données montrant un parcours où le robot visite les mêmes endroits une centaine de fois, les résultats démontrent l'adaptabilité et l'extensibilité de notre approche. |
</french> | </french> | ||
<center> | <center> |
Revision as of 15:40, 26 July 2011
RTAB-Map : Real-Time Appearance-Based Mapping
Description[edit]
Loop closure detection is the process involved when trying to find a match between the current and a previously visited locations in SLAM (Simultaneous Localization And Mapping). Over time, the amount of time required to process new observations increases with the size of the internal map, which may affect real-time processing. RTAB-Map is a novel real-time loop closure detection approach for large-scale and long-term SLAM. Our approach is based on efficient memory management to keep computation time for each new observation under a fixed time limit, thus respecting real-time limit for long-term operation. Results demonstrate the approach's adaptability and scalability using two custom data set and five standard data sets.
Example of sensorimotor learning using directly this loop closure detection approach (new in RTAB-Map 0.3) :
Results[edit]
Note that these results (more recent) may differ from those in the presentation video above...
Figure 1: Summary of the loop closures detected on UdeS data set :
- Green : Loop closures detected
- Yellow : Loop closures rejected
- Red : Unable to detect a loop closure because old places could not be retrieved
Figure 2: Processing time for each image acquired (real-time limit fixed to 700 ms for an image rate of 1 Hz)
Figure 3: Precision-Recall (48% recall at 100% precision)
Videos
Source code[edit]
The code was tested on Windows (Xp, 7), Mac OS X 10.6 and Ubuntu 10.4LTS.
- Standalone application, libraries and ROS packages : rtabmap.googlecode.com
Data sets[edit]
UdeS
- 5395 images at 1 Hz (1.5 hours).
- Images taken while walking through a loop of ~2 km, traversed two times.
- The data set contains indoor and outdoor environments.
UdeS_1Hz.part1.rar UdeS_1Hz.part2.rar UdeS_1Hz.part3.rar UdeS_1Hz GroundTruth
NFSMW
- Images taken from the racing video game Need For Speed: Most Wanted.
- 2 areas visited hundred times each (100 traversals in area 1 then moved to area 2 for another 102 traversals).
- 25098 images at 1 Hz (7 hours).
NFSMW_1Hz.part01.rar NFSMW_1Hz.part02.rar NFSMW_1Hz.part03.rar NFSMW_1Hz.part04.rar NFSMW_1Hz.part05.rar NFSMW_1Hz.part06.rar NFSMW_1Hz.part07.rar NFSMW_1Hz.part08.rar
Community
Community data sets from other loop closure detection approaches :
- Angeli et al. : Lip6Indoor et Lip6Outdoor
- Cummins et al. (FAB-MAP) : NewCollege et CityCentre
- Cummins et al. (FAB-MAP 2.0) : Eynsham (70 km)
Ground truths (readable by RTAB-Map) :
- NewCollege.rar 1073 images at ~0.5 Hz (left and right images merged)
- CityCentre.rar 1237 images at ~0.5 Hz (left and right images merged)
- Lip6Indoor.rar 388 images at 1 Hz
- Lip6Outdoor.rar 531 images at 0.5 Hz
- Eynsham70km.rar 5519 images at ~1 Hz (Note that we removed some images of the original data set to have an approximately image rate of 1 Hz)
Publications
M. Labbé and F. Michaud, “Memory management for real-time appearance-based loop closure detection,” in Proceedings of the IEEE/RSJ International Conference on Intelligent Robots and Systems, 2011. (Accepted)