A Two-Stage Feature Point Detection and Marking Approach Based on the Labeled Multi-Bernoulli Filter

Area of Science:

• Robotics
• Artificial Intelligence
• Computer Vision

Background:

• Traditional Simultaneous Localization and Mapping (SLAM) methods face challenges like data association, landmark management, and detection errors.
• Existing robot SLAM algorithms often suffer from low estimation accuracy and suboptimal back-end optimization.

Purpose of the Study:

• To present a novel labeled random finite set (L-RFS) SLAM method to address limitations in current SLAM techniques.
• To develop a robust framework for sensor navigation, obstacle avoidance, and accurate state estimation.

Main Methods:

• The proposed method utilizes the labeled random finite set (L-RFS) framework for SLAM.
• The labeled multi-Bernoulli filter (LMB) is employed for estimating the states of the sensor and feature points.
• B-spline curves are integrated to smooth the sensor's obstacle avoidance path.

Main Results:

• The L-RFS SLAM method demonstrates improved performance in sensor state and feature point estimation.
• The integration of B-spline curves effectively smooths obstacle avoidance trajectories.
• Simulations confirm the algorithm's effectiveness in complex navigation scenarios.

Conclusions:

• The proposed L-RFS SLAM method offers a significant advancement over traditional approaches by mitigating data association issues.
• The algorithm provides accurate state estimation and optimized path planning for mobile robots.
• This research contributes a robust solution for enhanced robot navigation and mapping.