Abstract:

With the rapidly growing demand for accurate localization in real-world environments, visual simultaneous localization and mapping (SLAM) has received significant attention in recent years. However, those existing methods still suffer from the degradation of localization accuracy in long-term changing environments. To address these problems, we propose a novel long-term SLAM system with map prediction and dynamics removal. First, a visual point-cloud matching algorithm is designed to efficiently fuse 2D pixel information and 3D voxel information. Second, each map point is classified into three types: static, semistatic, and dynamic based on the Bayesian persistence filter (BPF). Then we remove the dynamic map points to eliminate the influence of those map points. We can obtain a global predicted map by modeling the time series of semistatic map points. Finally, we incorporate the predicted global map into a state-of-the-art SLAM method, achieving an efficient visual SLAM system for long-term, dynamic environments. Extensive experiments are carried out on a wheelchair robot in an indoor environment over several months. The results demonstrate that our method has better map prediction accuracy and achieves more robust localization performance.

Figure 1. The proposed LTVS system can remove dynamics in environments and predict the future map, resulting in an excellent performance in localization for intelligent wheelchairs. RDB-D: red, green, blue-depth.

Figure 2. The architecture of our 3D map points descriptor.

Figure 3. The result of map points matching. The map points in red are the mismatched points. (a) and (b) On the right is an amplification of the circled area. (b) The result of map points matching with our matching algorithm.

DOI: 10.1109/MRA.2022.3228492