Memory Management for Real-Time Appearance-Based Loop Closure Detection

TL;DR

RTAB-Map presents a memory-managed appearance-based mapping framework for real-time loop closure in large-scale, long-term SLAM. By maintaining a bounded Working Memory of recent/frequent locations and transferring others to Long-Term Memory, it preserves real-time performance while enabling retrieval of neighboring locations when needed. A discrete Bayesian filter tracks loop-closure hypotheses, with a threshold-driven selection that merges locations and updates the dictionary through a retrieval mechanism. Results on four public datasets show 100% precision and favorable recall under real-time constraints, demonstrating scalability and adaptability for long-term operation.

Abstract

Loop closure detection is the process involved when trying to find a match between the current and a previously visited locations in SLAM. Over time, the amount of time required to process new observations increases with the size of the internal map, which may influence real-time processing. In this paper, we present a novel real-time loop closure detection approach for large-scale and long-term SLAM. Our approach is based on a memory management method that keeps computation time for each new observation under a fixed limit. Results demonstrate the approach's adaptability and scalability using four standard data sets.

Figures (3)

• Figure 1: Flow chart of our memory management loop closure detection processing cycle.
• Figure 2: Precision-recall curves for each data set.
• Figure 3: Processing time in terms of locations processed over time for each data set. The time limit $T_{\mathrm{time}}$ used is shown as a red line.