CAT-SLAM: probabilistic localisation and mapping using a continuous appearance-based trajectory

Abstract

This paper describes a new system, dubbed Continuous Appearance-based Trajectory Simultaneous Localisation and Mapping (CAT-SLAM), which augments sequential appearance-based place recognition with local metric pose filtering to improve the frequency and reliability of appearance-based loop closure. As in other approaches to appearance-based mapping, loop closure is performed without calculating global feature geometry or performing 3D map construction. Loop-closure filtering uses a probabilistic distribution of possible loop closures along the robot’s previous trajectory, which is represented by a linked list of previously visited locations linked by odometric information. Sequential appearance-based place recognition and local metric pose filtering are evaluated simultaneously using a Rao–Blackwellised particle filter, which weights particles based on appearance matching over sequential frames and the similarity of robot motion along the trajectory. The particle filter explicitly models both the likelihood of revisiting previous locations and exploring new locations. A modified resampling scheme counters particle deprivation and allows loop-closure updates to be performed in constant time for a given environment. We compare the performance of CAT-SLAM with FAB-MAP (a state-of-the-art appearance-only SLAM algorithm) using multiple real-world datasets, demonstrating an increase in the number of correct loop closures detected by CAT-SLAM.

Absolute geo-localization thanks to Hidden Markov Model and exemplar-based metric learning

“…Their solution remains sensitive to strong perceptual changes, and the same camera is used for database images and query images. Odometry information has been incorporated into FAB-MAP [18]. Another type of approach is to learn specificities of each place, as our solution that learns a visual similarity measure for each place.…”

Towards Persistent Localization and Mapping with a Continuous Appearance-based Topology

“…The proposed algorithm outlined in this section extends the linear 'trajectory-based' representation of [7] to a generalised graph-based representation. The steps of the algorithm for each new update of control input u k and visual bag-of-words observation z k are as follows:…”

“…Particle resampling is performed using the Select with Replacement method as in [7]. Any particles selected to replace the new location weight are sampled to a random edge on the graph (with a random direction).…”

“…The incorporation of a motion model into the image retrieval process has the potential to greatly enhance the performance of an appearance based system. In CAT-SLAM [6,7], the local movement information of the robot is fused with appearance information using a particle filter to dramatically improve the recall of location over the use of appearance information alone.…”

Understanding how camera configuration and environmental conditions affect appearance-based localization

“…We will ground our investigation using extensive field testing of a visual topometric localization algorithm [5]. The conclusions, however, could be generalized to other algorithms that employ visual localization [8], [14]. The key contributions of this paper are: 1) a detailed study of the relationship between system configuration and visual localization performance; 2) an analysis of the underlying causes for certain regions of the viewing sphere around a vehicle to be more effective for localization; and 3) a large-scale, publicly available data set of panoramic imagery with ground truth localization for the research community.…”