Optimising local map updates in a dynamic environment using an object detection method

Abstract

Keeping maps up to date in a time-varying environment is a big challenge in the development of autonomous mobile robots. In this paper, the authors propose a new solution to update the local map of the robot in a dynamic environment in real time. This proposed method works based on a combination of lidar scan data and object recognition and positioning algorithms to allow the robot to update the map continuously. Laser scan data for the same part of the environment at different times are compared to perform change detection. When there are enough changes in a location, the map will be edited to remove old features or add new features. To optimise updates for moving objects, authors use object recognition and positioning to decide which objects should be added to the map. The results are displayed in a simulated environment, showing the ability to maintain an up-to-date map despite long-term environmental changes.