Towards Life-Long Autonomy of Mobile Robots Through Feature-Based Change Detection

Abstract

Autonomous mobile robots are becoming increasingly important in many industrial and domestic environments. Dealing with unforeseen situations is a difficult problem that must be tackled in order to move closer to the ultimate goal of lifelong autonomy. In computer vision-based methods employed on mobile robots, such as localization or navigation, one of the major issues is the dynamics of the scenes. The autonomous operation of the robot may become unreliable if the changes that are common in dynamic environments are not detected and managed. Moving chairs, opening and closing doors or windows, replacing objects on the desks and other changes make many conventional methods fail. To deal with that, we present a novel method for change detection based on the similarity of local visual features. The core idea of the algorithm is to distinguish important stable regions of the scene from the regions that are changing. To evaluate the change detection algorithm, we have designed a simple visual localization framework based on feature matching and we have performed a series of real-world localization experiments. The results have shown that the change detection method substantially improves the accuracy of the robot localization, compared to using the baseline localization method without change detection.

Uncertainty-based localization in a topological robot navigation system

“…Works dealing with low dynamics, such as objects that are moved or change in their appearance, have been mainly approached from a feature-based perspective [126,45,49]. In the work by Patel et al [126], semantically enhanced features are used for pose estimation.…”

“…Features can only be matched if they belong to the same semantic type. Other works such as the work by Dayoub et al [45] and Derner et al [49] assign weights to each feature according to their stability in the environment. In the work presented by Stachniss and Burgard [146], laser information is used to estimate robot pose in a non-static environment through a Rao-Blackwellized particle filter.…”

“…In our work, we present a probabilistic localization algorithm for indoor environments that includes semantic information in the estimation process, similarly to [15,147]. Unlike most approaches for non-static indoor environments [45,49,126], we exploit semantic segmentation for every pixel and 3D point in the environment. However, we share with Dayoub et al [45] and Derner et al [49] a weighting method in the estimation process.…”

“…Unlike most approaches for non-static indoor environments [45,49,126], we exploit semantic segmentation for every pixel and 3D point in the environment. However, we share with Dayoub et al [45] and Derner et al [49] a weighting method in the estimation process. In the case of the previous works, each feature is weighted according to its stability.…”

PlaneSDF-Based Change Detection for Long-Term Dense Mapping

“…Classical 2D feature descriptors, e.g. SURF, ORB, and BRISK [12], [13], were extracted from the grey scale input images and the visual database, respectively. Next, the Euclidean distance between the two features is computed to determine if changes have taken place.…”