Poisson Multi-Bernoulli Filter Driven Information-Controlled Selection of Pose Graph Constraints for SLAM

Area of Science:

• Robotics
• Computer Vision
• Artificial Intelligence

Background:

• Traditional Simultaneous Localization and Mapping (SLAM) methods struggle with high computational demands, data association ambiguity, and limited real-time capabilities in complex environments.
• Existing SLAM approaches often require explicit data association or intensive filtering, hindering scalability and robustness.

Purpose of the Study:

• To propose a novel SLAM framework that overcomes the limitations of traditional methods.
• To enhance computational efficiency, robustness, and real-time performance for SLAM in cluttered environments.

Main Methods:

• A pose-graph-optimization-based Poisson multi-Bernoulli (PMB) SLAM framework is developed.
• The map is unified using a Poisson point process (PPP) for undetected features and multi-Bernoulli (MB) components for detected features.
• An information-controlled pose graph constraint selection strategy (IC-PGCS) couples PMB filtering with pose graph optimization.

Main Results:

• The proposed PMB SLAM framework achieves comparable map feature estimation accuracy.
• Significant improvements in computational efficiency and real-time performance were observed compared to RB-PHD-SLAM and other multi-Bernoulli SLAM methods.
• The framework demonstrates effectiveness in cluttered indoor environments.

Conclusions:

• The proposed PMB SLAM framework offers a robust and efficient solution for complex environments.
• The integration of PMB filtering with pose graph optimization and IC-PGCS enhances SLAM performance.
• This approach provides a viable alternative for real-time SLAM applications in challenging settings.