Probabilistic Appearance-Invariant Topometric Localization With New Place Awareness

Abstract

Probabilistic state-estimation approaches offer a principled foundation for designing localization systems, because they naturally integrate sequences of imperfect motion and exteroceptive sensor data. Recently, probabilistic localization systems utilizing appearance-invariant visual place recognition (VPR) methods as the primary exteroceptive sensor have demonstrated state-of-the-art performance in the presence of substantial appearance change. However, existing systems 1) do not fully utilize odometry data within the motion models, and 2) are unable to handle route deviations, due to the assumption that query traverses exactly repeat the mapping traverse. To address these shortcomings, we present a new probabilistic topometric localization system which incorporates full 3-dof odometry into the motion model and furthermore, adds an "off-map" state within the state-estimation framework, allowing query traverses which feature significant route detours from the reference map to be successfully localized. We perform extensive evaluation on multiple query traverses from the Oxford RobotCar dataset exhibiting both significant appearance change and deviations from routes previously traversed. In particular, we evaluate performance on two practically relevant localization tasks: loop closure detection and global localization. Our approach achieves major performance improvements over both existing and improved state-of-the-art systems.

Development, integration, and field evaluation of an autonomous citrus‐harvesting robot

“…Over the years, various localization and navigation methods on mobile platforms have been developed for different usage environments, from field weeding robots (McCool et al, 2018) to transportation robots (Ye et al, 2017), from LIDAR detection and navigation approaches (Grimstad et al, 2018), to probabilistic localization techniques (Xu et al, 2021), and to the use of topological maps (Fentanes et al, 2015). The Doppler displacement compensation algorithm based on the carrier frequency detection realized localization and navigation of the small mobile platform in greenhouse environments (Huang et al, 2021).…”

How Many Events do You Need? Event-based Visual Place Recognition Using Sparse But Varying Pixels

“…Thirdly, we are interested spike time similarity metrics [38] to also take the precise timing and order of the events into account when comparing the reference and query traverses. Taken together with sophisticated odometry [5], this could pave the way to place recognition in ultrasonar applications. Finally, we could imagine to apply our methodology to other camera types, including light-field cameras that can trade-off spatial and temporal resolution [39], [40].…”

“…If the score associated to that match is sufficiently high, the match can then be used to e.g. propose a loop closure within a Simultaneous Localization and Mapping (SLAM) system [4], or the VPR system can be used on its own to build a purely topological map [5].…”

Sequential Monte Carlo localization in topometric appearance maps

“…Inspired by (Xu et al 2021), we employ a chi-squared cumulative distribution with three degrees of freedom over such distances:…”