Evaluation of a simultaneous localization and mapping algorithm in a dynamic environment using a red green blue—depth camera

Abstract

Simultaneous localization and mapping (SLAM) assumes a static environment. In a dynamic environment, the localization accuracy and map quality of SLAM may be degraded by moving objects. By removing these moving objects SLAM performance may improve. Oriented FAST (Features from Accelerated Segment Test) and Rotated BRIEF (Binary Robust Independent Elementary Features) (ORB)-SLAM (Mur-Artal et al. in IEEE Trans Rob 31:1147–1163, 2015 [1]) is a state-of-the-art SLAM algorithm that has shown good performance on several Red Green Blue—Depth (RGB-D) datasets with a moving camera in static and dynamic environments. ORB-SLAM is robust to moderate dynamic changes (Mur-Artal et al. in IEEE Trans Rob 31:1147–1163, 2015 [1]). However, ORB-SLAM has not been evaluated with a moving RGB-D camera and an object moving at a range of specific linear speeds. This paper evaluates the performance of ORB-SLAM with a moving RGB-D camera in a dynamic environment that includes an object moving at a range of specific linear speeds. Results from experiments indicate that a moving object at lower speeds, in the range tested, degrades the performance of ORB-SLAM and by removing the moving object the performance of ORB-SLAM improves.