H-SLAM: Rao-Blackwellized Particle Filter SLAM Using Hilbert Maps

Area of Science:

• Robotics
• Artificial Intelligence
• Computer Vision

Background:

• Occupancy grid maps are crucial for robotic applications like path planning and autonomous manipulation, offering probabilistic spatial representations.
• Existing methods may face challenges with memory footprint and online processing capabilities.

Purpose of the Study:

• To present a novel Simultaneous Localization and Mapping (SLAM) framework, H-SLAM, designed for online generation of occupancy grid maps.
• To leverage Hilbert Maps for a continuous, probabilistic spatial representation with reduced memory requirements.

Main Methods:

• The H-SLAM framework utilizes Hilbert Maps for environment representation.
• A Particle Filter is employed to manage and update the robot's state estimation.
• The system is designed for online operation, processing data in real-time.

Main Results:

• Experimental results from both simulated environments and real Autonomous Underwater Vehicles (AUVs) were analyzed.
• The H-SLAM approach demonstrated consistent environment representation.
• The system proved capable of online execution, meeting real-time processing demands.

Conclusions:

• H-SLAM provides an effective and efficient method for generating occupancy grid maps online.
• The use of Hilbert Maps contributes to a compact and continuous probabilistic representation.
• The framework is suitable for robotic applications requiring real-time mapping and localization, including AUVs.