Localization

Localization Packages

Sylphase INS

Discussion

Currently we are investigating 3 types of sensor to be used used for localization:

Inertial Navigator Sensors and GNSS (GPS, GLONASS, etc.)

INS and GNSS are usually coupled into an single device.

Pros

• Directly yields position and orientation data, minimal computation required.

• RTK and other corrections can greatly boost the accuracy of the position measurement.

Cons

• GNSS data rate is slow < 10Hz

• Position can drift if too reliant on inertial sensors.

• GNSS is known to produce large errors if environmental conditions negatively affect the signals.

LIDAR Localization

Pros

• High definition geometric data that can be compared to a know point cloud.

• Plenty of existing algorithms to do this: e.g NDT Matching

• Operates in almost any lighting conditions

Cons

• Slow data acquisition rate, at high speeds a mechanical spinning LIDAR would not capture the world correctly

• Somewhat costly, not as bad anymore

• Mechanical spinning LIDARS will be affected my the fast movements of the vehicle.

Camera Localization

Camera based localization would rely on algorithm suchs as [ORB-SLAM](https://webdiis.unizar.es/~raulmur/orbslam/) to determine the position and orientation of the sensor with respect to a pre-constructed map.

Pros

• Cheaper than LIDARs most of the time

• High data acquisition rate, 100+ FPS

Cons

• Camera based computations can be costly which would reduce the effectiveness of their high data acquisition

• Lighting can extremely affect the usability of the sensor.