FIT-SLAM -- Fisher Information and Traversability estimation-based Active SLAM for exploration in 3D environments

TL;DR

This paper tackles active exploration and robust localization for unmanned ground vehicles in GNSS-denied 3D environments by integrating perception-aware planning into SLAM. It introduces FIT-SLAM, which projects 3D space into a global 2D traversability map to drive frontier-based exploration, while a two-level utility function combines exploration progress with SLAM-informed information gain from observed landmarks along planned paths. The approach merges a graph-based SLAM backend (with 6D poses and landmarks and sliding-window optimization), a geometry-based traversability estimator, frontier detection/clustering, and Fisher-information-based path evaluation using TOED to manage memory. Experimental validation in simulation and real-world planar scenarios demonstrates higher exploration rates and lower localization covariance compared to baselines, with numerous loop closures arising from landmark-informed planning. This framework offers a scalable, information-driven strategy for efficient 3D exploration with robust SLAM in challenging environments.

Abstract

Active visual SLAM finds a wide array of applications in GNSS-Denied sub-terrain environments and outdoor environments for ground robots. To achieve robust localization and mapping accuracy, it is imperative to incorporate the perception considerations in the goal selection and path planning towards the goal during an exploration mission. Through this work, we propose FIT-SLAM (Fisher Information and Traversability estimation-based Active SLAM), a new exploration method tailored for unmanned ground vehicles (UGVs) to explore 3D environments. This approach is devised with the dual objectives of sustaining an efficient exploration rate while optimizing SLAM accuracy. Initially, an estimation of a global traversability map is conducted, which accounts for the environmental constraints pertaining to traversability. Subsequently, we propose a goal candidate selection approach along with a path planning method towards this goal that takes into account the information provided by the landmarks used by the SLAM backend to achieve robust localization and successful path execution . The entire algorithm is tested and evaluated first in a simulated 3D world, followed by a real-world environment and is compared to pre-existing exploration methods. The results obtained during this evaluation demonstrate a significant increase in the exploration rate while effectively minimizing the localization covariance.

Figures (10)

• Figure 1: Overview of the proposed framework: A 2D thresholded traversability map is analyzed to extract frontiers goals (stars). Each goal candidate $C_i$ is ranked based on the information computed for the planned path. Such information is linked to the observed landmarks (red dots) in the camera field-of-view (FOV) $\Phi$ (blue area) but also the maximal reduction in entropy after reaching the goal with robot orientation $\Theta_s^*$. This allows to ensure good localizability while exploring new areas, resulting in a more accurate mapping.
• Figure 2: Our ASLAM solution framework: First a global traversability map is built based on graph-based SLAM and 3D perception. Then, based on the rover capabilities, traversability scores are thresholded and the frontiers are detected. Goals are defined for each frontier and ranked on the basis of information gain. Finally, path safety for each goal are evaluated using predicted perception entropies. Depending on the other constraints of the mission, the final path is selected and executed.
• Figure 3: Estimated local traversability map. (a) The processed geometric traversability was obtained with a 3D LiDAR detection. (b) The reprojection of the traversability on the synchronized image showing (in red) the navigability risks.
• Figure 4: Our robotic platform equipped with the sensors (Depth Camera, 3D LiDAR, IMU and wheel odometry) required for the algorithm.
• Figure : (a)