A Mapping Method Tolerant to Calibration and Localization Errors Based on Tilting 2D Laser Scanner

Autonomous mobile machines use onboard sensors for navigation and obstacle avoidance. The accuracy of the sensor data in global frame is however dependent on the localization accuracy of the machine. Simultaneous localization and mapping algorithms (SLAM) are widely used with 3D laser scanners for mapping the world. They use scan matching algorithms to solve the accuracy problem by matching prior sensor data of the environment with the newly acquired data. However matching scans is not always possible. Insufficient amount of prior data or too few features in the scan can prevent the scan matching algorithm from finding a match. Thus it is important that also the mapping algorithm is tolerant to some degree of error in localization and calibration. We present a method for generating obstacle maps from smaller data segments at a time, thus making the mapping system more tolerant to navigation and calibration errors. The obstacle mapping method is tested with modified Avant multipurpose loader.