Appearance-Based Loop Closure Detection for Online Large-Scale and Long-Term Operation

--In appearance-based localization and mapping, loop closure detection is the process used to determinate if the current observation comes from a previously visited location or a new one. As the size of the internal map increases, so does the time required to compare new observations with all stored locations, eventually limiting online processing. This paper presents an online loop closure detection approach for large-scale and long-term operation. The approach is based on a memory management method, which limits the number of locations used for loop closure detection so that the computation time remains under real-time constraints. The idea consists of keeping the most recent and frequently observed locations in a Working Memory (WM) used for loop closure detection, and transferring the others into a Long-T erm Memory (L TM). When a match is found between the current location and one stored in WM, associated locations stored in L TM can be updated and remembered for additional loop closure detections. Results demonstrate the approach's adaptability and scalability using ten standard data sets from other appearance-based loop closure approaches, one custom data set using real images taken over a 2 km loop of our university campus, and one custom data set (7 hours) using virtual images from the racing video game "Need for Speed: Most Wanted". UTONOMOUS robots operating in real life settings must be able to navigate in large, unstructured, dynamic and unknown spaces. Simultaneous localization and mapping (SLAM) [1] is the capability required by robots to build and update a map of their operating environment and to localize themselves in it. A key feature in SLAM is to recognize previously visited locations. This process is also known as loop closure detection, referring to the fact that coming back to a previously visited location makes it possible to associate this location with another one recently visited. For most of the probabilistic SLAM approaches [2]-[13], loop closure detection is done locally, i.e., matches are found between new observations and a limited region of the map, determined by the uncertainty associated with the robot's Manuscript received April 23, 2012; revised October 2, 2012; accepted January 14, 2013. This paper was recommended for publication by Associate Editor P . This work was supported in part by the Natural Sciences and Engineering Research Council of Canada, the Canadian Foundation for Innovation and the Canada Research Chair program. Color versions of one or more of the figures in this paper are available online at http://ieeexplore.ieee.org. Such approaches can be processed under real-time contraints at 30 Hz [14] as long as the estimated position is valid, which cannot be guaranteed in real world situations [15].