Extended kalman filter (EKF)-based local SLAM in dynamic environments: A framework

Abstract

In the domain of mobile robots local maps of environments are used as knowledge base for decisions allowing reactive control in order to prevent collisions when following a global trajectory. These maps are normally discrete and updated at relatively high frequency, but with no dynamic information. The proposed framework uses a sparse description of clustered scan points from a laser range scanner. These features and the system odometry are used to predict the agent’s ego motion as well as feature motion using an Extended Kalman Filter. This approach is similar to the Simultaneous Localization and Mapping (SLAM) algorithm but with low-constraint features. The presented local Simultaneous Localization and Mapping (LSLAM) approach creates a decision base, holding a dynamic description which relaxes the requirement of high update rates. Simulated results demonstrate environment classification and tracking as well as self-pose correction in static and in dynamic environments.