Abstract — This paper describes a 3D SLAM system using information from an actuated laser scanner and camera installed on a mobile robot.The laser samples the local geometry of the environment and is used to incrementally build a 3D point-cloud map of the workspace. Sequences of images from the camera are used to detect loop closure events (without reference to the internal estimates of vehicle location) using a novel appearancebased retrieval system. The loop closure detection is robust to repetitive visual structure and provides a probabilistic measure of confidence. The images suggesting loop closure are then further processed with their corresponding local laser scans to yield putative Euclidean image-image transformations. We show how naive application of this transformation to effect the loop closure can lead to catastrophic linearization errors and go on to describe a way in which gross, pre-loop closing errors can be successfully annulled. We demonstrate our system working in a challenging, outdoor setting containing substantial loops and beguiling, gently curving traversals. The results are overlaid on an aerial image to provide a ground truth comparison with the estimated map. The paper concludes with an extension into the multi-robot domain in which 3D maps resulting from distinct SLAM sessions (no common reference frame) are combined without recourse to mutual observation. I.

Abstract — This paper 1 presents a method for Simultaneous Localization and Mapping (SLAM) relying on a monocular camera as the only sensor which is able to build outdoor, closedloop maps much larger than previously achieved with such input. Our system, based on the Hierarchical Map approach [1], builds independent local maps in real-time using the EKF-SLAM technique and the inverse depth representation proposed in [2]. The main novelty in the local mapping process is the use of a data association technique that greatly improves its robustness in dynamic and complex environments. A new visual map matching algorithm stitches these maps together and is able to detect large loops automatically, taking into account the unobservability of scale intrinsic to pure monocular SLAM. The loop closing constraint is applied at the upper level of the Hierarchical Map in near real-time. We present experimental results demonstrating monocular SLAM as a human carries a camera over long walked trajectories in outdoor areas with people and other clutter, even in the more difficult case of forward-looking camera, and show the closing of loops of several hundred meters. I.

The problem studied in this paper is a mobile robot that autonomously navigates in a domestic environment, builds a map as it moves along and localizes its position in it. In addition, the robot detects predefined objects, estimates their position in the environment and integrates this with the localization module to automatically put the objects in the generated map. Thus, we demonstrate one of the possible strategies for the integration of spatial and semantic knowledge in a service robot scenario where a simultaneous localization and mapping (SLAM) and object detection/recognition system work in synergy to provide a richer representation of the environment than it would be possible with either of the methods alone. Most SLAM systems build maps that are only used for localizing the robot. Such maps are typically based on grids or different types of features such as point and lines. The novelty is the augmentation of this process with an object recognition system that detects objects in the environment and puts them in the map generated by the SLAM system. The metric map is also split into topological entities corresponding to rooms. In this way the user can command the robot to retrieve a certain object from a certain room. We present the results of map building and an extensive evaluation of the object detection algorithm performed in an indoor setting. 1

This article presents an approach to SLAM that takes advantage of panoramic images. Landmarks are interest points detected and matched in the images and mapped according to a bearings-only SLAM approach. As they are acquired and processed, the panoramic images are also indexed and stored into a database. A database query procedure, independent of the robot and landmark position estimates, is able to detect loop closures by retrieving memorized images that are close to the current robot position. The bearings-only estimation process is described, and results over a trajectory of a few hundreds of meters are presented and discussed. 1

www.robots.ox.ac.uk/∼mobile Abstract — We consider the task of processing 3D laser data for use in the Simultaneous Localization and Mapping Problem. The motivation for using 3D data comes in part from the impracticality of relying on 2D laser scanners when the vehicle operates on undulating terrain and in part from a desire to produce 3D maps of arbitrary, a priori unknown environments. We use an information-theoretic derived measure of local saliency to partition the raw 3D data stream into spatially distinct point-clusters. These clusters are natural features in measurement space that capture the geometry of intrinsically interesting surface patches. In common with “scan-matching ” methods in SE 2, the SE 3 relationship between consecutive vehicle poses is calculated using an iterative point-wise registration scheme operating on the reduced data set. The saliency driven decimation process not only substantially reduces the computational burden of registration but also provides the registration process with data that is geometrically diverse. This characteristic improves registration performance. We present initial results showing our methods working on both outdoor and indoor data.

www.elsevier.com/locate/robot In this paper we describe a system for use on a mobile robot that detects potential loop closures using both visual and spatial appearance of local scenes. Loop closing is the act of correctly asserting that a vehicle has returned to a previously visited location. Current approaches rely heavily on vehicle pose estimates to prompt loop closure. Paradoxically, these approaches are least reliable when the need for accurate loop closure detection is the greatest. Our underlying approach relies instead upon matching distinctive ‘signatures ’ of individual local scenes to prompt loop closure. A key advantage of this method is that it is entirely independent of the navigation and or mapping process and so is entirely unaffected by gross errors in pose estimation. Another advantage, which is explored in this paper, is the possibility to enhance robustness of loop closure detection by incorporating heterogeneous sensory observations. We show how a description of local spatial appearance (using laser rangefinder data) can be combined with visual descriptions to form multi-sensory signatures of local scenes which enhance loop-closure detection. c ○ 2006 Elsevier B.V. All rights reserved.

I propose a notion of visual information as the complexity not of the raw images, but of the images after the effects of nuisance factors such as viewpoint and illumination are discounted. It is rooted in ideas of J. J. Gibson, and stands in contrast to traditional information as entropy or coding length of the data regardless of its use, and regardless of the nuisance factors affecting it. The non-invertibility of nuisances such as occlusion and quantization induces an “information gap ” that can only be bridged by controlling the data acquisition process. Measuring visual information entails early vision operations, tailored to the structure of the nuisances so as to be “lossless ” with respect to visual decision and control tasks (as opposed to data transmission and storage tasks implicit in traditional Information Theory). I illustrate these ideas on visual exploration, whereby a “Shannonian Explorer ” guided by the entropy of the data navigates unaware of the structure of the physical space surrounding it, while a “Gibsonian Explorer ” is guided by the topology of the environment, despite measuring only images of it, without performing 3D reconstruction. The operational definition of visual information suggests desirable properties that a visual representation should possess to best accomplish vision-based decision and control tasks. 1.

This paper is centered around landmark detection, tracking and matching for visual SLAM (Simultaneous Localization And Mapping) using a monocular vision system with active gaze control. We present a system specialized in creating and maintaining a sparse set of landmarks based on a biologically motivated feature selection strategy. A visual attention system detects salient features which are highly discriminative, ideal candidates for visual landmarks which are easy to redetect. Features are tracked over several frames to determine stable landmarks and to estimate their 3D position in the environment. Matching of current landmarks to database entries enables loop closing. Active gaze control allows us to overcome some of the limitations of using a monocular vision system with a relatively small field of view. It supports (i) the tracking of landmarks which enable a better pose estimation, (ii) the exploration of regions without landmarks to obtain a better distribution of landmarks in the environment, and (iii) the active redetection of landmarks to enable loop closing in situations in which a fixed camera fails to close the loop. Several real-world experiments show that accurate pose estimation is obtained with the presented system and that active camera control outperforms the passive approach.

Abstract. Visual attention regions are useful for many applications in the field of computer vision and robotics. Here, we introduce an application to simultaneous robot localization and mapping. A biologically motivated attention system finds regions of interest which serve as visual landmarks for the robot. The regions are tracked and matched over consecutive frames to build stable landmarks and to estimate the 3D position of the landmarks in the environment. Matching of current landmarks to database entries enables loop closing and global localization. Additionally, the system is equipped with an active camera control, which supports the system with a tracking, a re-detection, and an exploration behaviour. We present experiments which show the applicability of the system in a real-world scenario. A comparison between the system operating in active and in passive mode shows the advantage of active camera control: we achieve a better distribution of landmarks as well as a faster and more reliable loop closing. 1

In this paper, we propose a new, hierarchical approach to landmark selection for simultaneous robot localization and mapping based on visual sensors: a biologically motivated attention system finds salient regions of interest (ROIs) in images, and within these regions, Harris corners are detected. This combines the advantages of the ROIs (reducing complexity, enabling good redetactability of regions) with the advantages of the Harris corners (high stability). Reducing complexity is important to meet real-time requirements and stability of features is essential to compute the depth of landmarks from structure from motion with a small baseline. We show that the number of landmarks is highly reduced compared to all Harris corners while maintaining the stability of features for the mapping task. 1.