RoadMentor Ground Truthing (Localization)

Introduction:

RoadMentor’s cloud-based aerial to ground fusion technology is a revolutionary approach to improving the accuracy of ego-location in vehicles. By combining data from onboard sensors with airborne imagery and point clouds, RoadMentor’s technology is able to significantly improve the localization of ego-vehicles. This white paper will provide an overview of the RoadMentor technology and its ability to improve ego-location accuracy through the fusion of onboard sensor data and airborne imagery.

Ego-location and onboard sensors:

Ego-location refers to the ability of a vehicle to determine its own position and orientation within a given coordinate system. This is an important capability for autonomous vehicles, as it allows the vehicle to navigate and make decisions based on its surroundings.

Onboard sensors, such as cameras, LiDAR, and radar, are used to gather information about the vehicle’s surroundings. These sensors can provide data on the location, shape, and movement of objects in the environment.

Airborne imagery and point clouds:

Airborne imagery refers to high-resolution images captured from the air, often using aerial photography or satellite imagery. These images can provide detailed information about the features of the environment, including the location and shape of buildings, roads, and other landmarks.

Point clouds are 3D models of the environment created by collecting data from lasers or other sensors. These point clouds can provide detailed information about the shape and location of objects in the environment.

RoadMentor’s technology:

RoadMentor’s technology uses a cloud-based aerial to ground fusion approach to improve the accuracy of ego-location in vehicles. By comparing the onboard sensor data from the cameras, LiDAR, and radar with airborne imagery and point clouds, RoadMentor’s technology is able to estimate the vehicle’s ego-location with high accuracy.

The process of localizing ego-vehicles using RoadMentor’s technology involves the following steps:

1. The vehicle’s onboard sensors gather data about the environment.
2. This data is compared to airborne imagery and point clouds to find features in both sets of data.
3. By matching features in the onboard sensor data and the airborne imagery and point clouds, the vehicle’s ego-location can be estimated.
4. The ego-location data is used by the vehicle to navigate and make decisions based on its surroundings.

Results:

The results of using RoadMentor’s technology have been impressive. In tests, the technology has achieved lateral error rates of less than 20cm and longitudinal error rates of less than 1m. This represents a significant improvement in the accuracy of ego-location, making it possible for autonomous vehicles to navigate with a high degree of precision.

To validate RoadMentor’s technology, Hyperspec physically collected ground control points and compared the data with the registered point cloud derived from the pose estimations of the ego-location of the vehicle.

The process of collecting ground control points involves the following steps:

1. Hyperspec identified a suitable area for collecting ground control points. This area should be representative of the environment in which the vehicle would be operating and should have a sufficient number of identifiable features.
2. Hyperspec used its airborne sensors to collect high-resolution imagery and point clouds of the area.
3. Hyperspec identified a set of ground control points within the imagery and point clouds. These points should be easily identifiable and should have known coordinates within a specific coordinate system.
4. Hyperspec recorded the coordinates of the ground control points and used them as reference points for the vehicle’s ego-location.

To compare the data with the registered point cloud derived from the pose estimations of the ego-location of the vehicle, Hyperspec followed the following steps:

1. RoadMentor’s technology was used to determine the ego-location of the vehicle using the onboard sensors and the airborne imagery and point clouds.
2. The pose estimations from the ego-location of the vehicle were used to create a registered point cloud.
3. The coordinates of the ground control points were compared to the points in the registered point cloud to determine the error in the ego-location of the vehicle.
4. The error data was used to make adjustments to the pose estimations of the ego-location of the vehicle.

This process allowed Hyperspec to validate the accuracy of RoadMentor’s technology by comparing the data collected from the ground control points with the pose estimations of the ego-location of the vehicle. By physically collecting ground control points and comparing the data with the registered point cloud, Hyperspec was able to confirm that RoadMentor’s technology was able to accurately determine the ego-location of the vehicle with lateral error rates of less than 20cm and longitudinal error rates of less than 1m.

Conclusion:

RoadMentor’s cloud-based aerial to ground fusion technology is a powerful tool for improving the accuracy of ego-location in vehicles. By combining data from onboard sensors with airborne imagery and point clouds, RoadMentor’s technology is able to significantly improve the localization of ego-vehicles. The results of using this technology have been impressive, with lateral error rates of less than 20cm and longitudinal error rates of less than 1m. These results demonstrate the potential of RoadMentor’s technology to enable the safe and reliable operation of autonomous vehicles.