Particle filter — Scan matching SLAM recovery under kinematic model failures

Simultaneous localization and mapping comprises two highly correlated procedures, which renders it a greatly difficult problem. Its difficulty is further increased due to the high levels of uncertainty introduced by both the environment and the robot's sensors and actuators. In the current paper we present methods to recover from extreme situations where kinematic model failures are observed - robot slipping and movement obstruction - that result in erroneous localizations, leading to complete SLAM method failures.