Cloud Update of Tiled Evidential Occupancy Grid Maps for the Multi-Vehicle Mapping

Area of Science:

• Intelligent Transportation Systems
• Robotics and Autonomous Systems
• Cloud Computing and Big Data

Background:

• Intelligent vehicles utilize diverse sensors for environmental perception and navigation.
• Increasing connectivity enables vehicles to share sensor data through cloud services, enhancing operational safety and efficiency.
• Existing mapping solutions struggle with sensor disturbances and large-scale environmental coverage.

Purpose of the Study:

• To propose a cloud-based framework for updating occupancy grid maps for multiple intelligent vehicles in large-scale environments.
• To enhance the robustness of occupancy grid maps against sensor noise, occlusion, and dynamic objects using evidential theory.
• To establish a scalable and efficient method for managing and merging map data from multiple vehicles.

Main Methods:

• Utilizing evidential theory to construct evidential occupancy grid maps (EOGM) that account for sensor uncertainties.
• Employing GraphSLAM with LiDAR, motion sensors, and GPS data for map generation along vehicle trajectories.
• Implementing a geodetic quad-tree tile system for organizing and managing EOGM tiles across large areas.
• Leveraging Dempster's combination rule within a cloud platform for merging updated EOGM tiles.

Main Results:

• Demonstrated the creation of EOGMs by multiple vehicles using sensor fusion and GraphSLAM.
• Successfully managed and updated EOGM tiles in a cloud environment using a geodetic quad-tree system.
• Validated the effectiveness of the cloud update framework in a large-scale road environment through experimental evaluation.

Conclusions:

• The proposed cloud update framework effectively enables multiple intelligent vehicles to collaboratively build and maintain occupancy grid maps in large-scale environments.
• Evidential theory provides a robust approach to handling sensor uncertainties, leading to more reliable occupancy grid maps.
• The integration of a geodetic quad-tree system and cloud-based merging ensures scalability and efficiency for real-time mapping applications.