Parallelized SLAM: Enhancing Mapping and Localization Through Concurrent Processing

Area of Science:

• Robotics
• Computer Vision
• Artificial Intelligence

Background:

• Simultaneous Localization and Mapping (SLAM) systems require substantial computational resources, limiting real-time applications on low-end hardware.
• Current offline map creation for SLAM is sequential and time-consuming, even on powerful computers.

Purpose of the Study:

• To develop a parallel mapping method (pSLAM) to significantly reduce the time required for offline SLAM map generation.
• To enable faster deployment of SLAM for real-time navigation on resource-constrained devices.

Main Methods:

• The proposed pSLAM method partitions video sequences into multiple subsequences for independent processing.
• Individual submaps generated from subsequences are merged to create a unified global map.
• Experimental validation across diverse scenarios and comparison with state-of-the-art SLAM methods (UcoSLAM, OpenVSLAM, ORB-SLAM3).

Main Results:

• pSLAM achieved processing speeds up to 6 times faster than sequential approaches.
• The method maintained robustness comparable to traditional sequential SLAM.
• pSLAM outperformed UcoSLAM, OpenVSLAM, and ORB-SLAM3 in all evaluated scenarios.

Conclusions:

• Parallel processing significantly accelerates offline map creation for SLAM.
• pSLAM offers a viable solution for efficient real-time navigation using pre-generated maps on low-end systems.
• The proposed method demonstrates superior performance and speed compared to existing state-of-the-art SLAM techniques.