MAPPING AND LOCALIZATION SYSTEM FOR AUTONOMOUS VEHICLES

VALUE PROPOSITION

A fundamental task for autonomous vehicles is to accurately determine its position at all times. Multiple key sub-systems rely either fully or partially on the performance of the localization algorithm. It has been estimated that decimeter level localization accuracy is required for autonomous vehicles to drive safely and smoothly. GNSS-based (Global Navigation Satellite System) techniques struggle to achieve this level of accuracy except for open sky areas. Map-based localization frameworks, especially those that utilize Light Detection and Ranging (LiDAR) based localization methods, are popular because they can achieve centimeter level accuracy regardless of light conditions. However, a key drawback of any localization method that relies on 3D point-cloud maps is the enormous size of the map itself. Consequently, there is a need for efficient representations of such maps while maintaining high-accuracy localization capabilities. The representation format should contain sufficient information for vehicles to localize and be lightweight (i.e., low storage) enough to be stored and downloaded into vehicles in real-time when needed. Furthermore, it is important to note that environments do change rather frequently, and it is therefore important to have the ability to update the map to reflect these changes.
 

DESCRIPTION OF TECHNOLOGY

The proposed mapping framework requires less than 0.1% of the storage space of the original 3D point cloud map. In essence, mapping framework emulates an original map through feature likelihood functions. In particular, the mapping framework models planar, pole and curb features. These three feature classes are long-term stable, distinct and common among vehicular roadways. Multiclass feature points are extracted from LiDAR scans through feature detection. A new multiclass-based point-to-distribution alignment method is also used to find the association and alignment between the multiclass feature points and the map.

BENEFITS

• Enhanced Localization Accuracy: The proposed technology enables centimeter-level localization accuracy, which is crucial for safe and smooth autonomous vehicle navigation.
• Reduced Storage Requirements: The technology requires less than 0.1% of the storage space of the original 3D point cloud map, making it highly efficient for storage and transmission in city-scale environments.
• Real-time Alignment: The system features an efficient and robust real-time alignment algorithm for on-vehicle LiDAR scans, allowing for quick and accurate mapping of the environment.
• Adaptability to Changing Environments: The lightweight nature of the proposed map representation allows for easy updates to reflect changes in the environment, ensuring the system remains accurate and reliable.
• Versatility in Lighting Conditions: Unlike GNSS-based techniques, the proposed mapping framework can achieve high-accuracy localization regardless of light conditions, making it suitable for various environments.
• Improved Safety and Efficiency: By providing accurate vehicle localization, the technology contributes to the overall safety and efficiency of autonomous vehicles, as multiple key subsystems rely on the performance of the localization algorithm.

APPLICATIONS

• Autonomous Vehicles
• Indoor Navigation Systems
• Augmented Reality and Virtual Reality
• Robotics
• Smart Cities
• Disaster Response and Recovery

IP Status

US Patent 11,790,542

LICENSING RIGHTS AVAILABLE

All Licensing rights available

Inventors: Hayder Radha, Daniel Morris and Su Pang

Tech ID: TEC2019-0119

For more information about this technology,

Contact Jon Debling, Ph.D. at deblingj@msu.edu or +1-517-884-1653