Efficient and scalable 3D surface reconstruction from range data remains a core challenge in computer graphics and vision, particularly in real-time and resource-constrained scenarios. Traditional volumetric methods based on fixed-resolution voxel grids or hierarchical structures like octrees often suffer from memory inefficiency, computational overhead, and a lack of GPU support. We propose a novel variance-adaptive, multi-resolution voxel grid that dynamically adjusts voxel size based on the local variance of signed distance field (SDF) observations. Unlike prior multi-resolution approaches that rely on recursive octree structures, our method leverages a flat spatial hash table to store all voxel blocks, supporting constant-time access and full GPU parallelism. This design enables high memory efficiency and real-time scalability. We further demonstrate how our representation supports GPU-accelerated rendering through a parallel quad-tree structure for Gaussian Splatting, enabling effective control over splat density. Our open-source CUDA/C++ implementation achieves up to 13× speedup and 4× lower memory usage compared with fixed-resolution baselines, while maintaining on par results in terms of reconstruction accuracy, offering a practical and extensible solution for high-performance 3D reconstruction.

LiDARs provide accurate geometric measurements, making them valuable for ego-motion estimation and reconstruction tasks. Although its success, managing an accurate and lightweight representation of the environment still poses challenges. Both classic and NeRF-based solutions have to trade off accuracy over memory and processing times. In this work, we build on recent advancements in Gaussian Splatting methods to develop a novel LiDAR odometry and mapping pipeline that exclusively relies on Gaussian primitives for its scene representation. Leveraging spherical projection, we drive the refinement of the primitives uniquely from LiDAR measurements. Experiments show that our approach matches the current registration performance, while achieving SOTA results for mapping tasks with minimal GPU requirements. This efficiency makes it a strong candidate for further exploration and potential adoption in real-time robotics estimation tasks.

Robust and accurate localization and mapping of an environment using laser scanners, so-called LiDAR SLAM, is essential to many robotic applications. Early 3D LiDAR SLAM methods often exploited additional information from IMU or GNSS sensors to enhance localization accuracy and mitigate drift. Later, advanced systems further improved the estimation at the cost of a higher runtime and complexity. This paper explores the limits of what can be achieved with a LiDAR-only SLAM approach while following the "Keep It Small and Simple" (KISS) principle. By leveraging this minimalist design principle, our system, KISS-SLAM, achieves state-of-the-art performance in pose accuracy while requiring little to no parameter tuning for deployment across diverse environments, sensors, and motion profiles. We follow best practices in graph-based SLAM and build upon LiDAR odometry to compute the relative motion between scans and construct local maps of the environment. To correct drift, we match local maps and optimize the trajectory in a pose graph optimization step. The experimental results demonstrate that this design achieves competitive performance while reducing complexity and reliance on additional sensor modalities. By prioritizing simplicity, this work provides a new strong baseline for LiDAR-only SLAM and a high-performing starting point for future research. Furthermore, our pipeline builds consistent maps that can be used directly for downstream tasks like navigation. Our open-source system operates faster than the sensor frame rate in all presented datasets and is designed for real-world scenarios.

This paper presents a vision and perception research dataset collected in Rome, featuring RGB data, 3D point clouds, IMU, and GPS data. We introduce a new benchmark targeting visual odometry and SLAM, to advance the research in autonomous robotics and computer vision. This work complements existing datasets by simultaneously addressing several issues, such as environment diversity, motion patterns, and sensor frequency. It uses up-to-date devices and presents effective procedures to accurately calibrate the intrinsic and extrinsic of the sensors while addressing temporal synchronization. During recording, we cover multi-floor buildings, gardens, urban and highway scenarios. Combining handheld and car-based data collections, our setup can simulate any robot (quadrupeds, quadrotors, autonomous vehicles). The dataset includes an accurate 6-dof ground truth based on a novel methodology that refines the RTK-GPS estimate with LiDAR point clouds through Bundle Adjustment (BA). All sequences divided in training and testing are accessible at www.rvp-group.net/datasets/slam.

Agricultural robots have the prospect to enable more efficient and sustainable agricultural production of food, feed, and fiber. Perception of crops and weeds is a central component of agricultural robots that aim to monitor fields and assess the plants as well as their growth stage in an automatic manner. Semantic perception mostly relies on deep learning using supervised approaches, which require time and qualified workers to label fairly large amounts of data. In this paper, we look into the problem of reducing the amount of labels without compromising the final segmentation performance. For robots operating in the field, pre-training networks in a supervised way is already a popular method to reduce the number of required labeled images. We investigate the possibility of pre-training in a self-supervised fashion using data from the target domain. To better exploit this data, we propose a set of domain-specific augmentation strategies. We evaluate our pre-training on semantic segmentation and leaf instance segmentation, two important tasks in our domain. The experimental results suggest that pre-training with domain-specific data paired with our data augmentation strategy leads to superior performance compared to commonly used pre-trainings. Furthermore, the pre-trained networks obtain similar performance to the fully supervised with less labeled data.

Simultaneous Localization and Mapping (SLAM) systems are fundamental building blocks for any autonomous robot navigating in unknown environments. The SLAM implementation heavily depends on the sensor modality employed on the mobile platform. For this reason, assumptions on the scene's structure are often made to maximize estimation accuracy. This paper presents a novel direct 3D SLAM pipeline that works independently for RGB-D and LiDAR sensors. Building upon prior work on multi-cue photometric frame-to-frame alignment [4], our proposed approach provides an easy-to-extend and generic SLAM system. Our pipeline requires only minor adaptations within the projection model to handle different sensor modalities. We couple a position tracking system with an appearance-based relocalization mechanism that handles large loop closures. Loop closures are validated by the same direct registration algorithm used for odometry estimation. We present comparative experiments with state-of-the-art approaches on publicly available benchmarks using RGB-D cameras and 3D LiDARs. Our system performs well in heterogeneous datasets compared to other sensor-specific methods while making no assumptions about the environment. Finally, we release an open-source C++ implementation of our system.

Robots and autonomous systems need to know where they are within a map to navigate effectively. Thus, simultaneous localization and mapping or SLAM is a common building block of robot navigation systems. When building a map via a SLAM system, robots need to re-recognize places to find loop closure and reduce the odometry drift. Image-based place recognition received a lot of attention in computer vision, and in this work, we investigate how such approaches can be used for 3D LiDAR data. Recent LiDAR sensors produce high-resolution 3D scans in combination with comparably stable intensity measurements. Through a cylindrical projection, we can turn this information into a 360° panoramic range image. As a result, we can apply techniques from visual place recognition to LiDAR intensity data. The question of how well this approach works in practice has only partially been investigated. This paper provides an analysis of how such visual techniques can be with LiDAR data, and we provide an evaluation on different datasets. Our results suggest that this form of place recognition is possible and an effective means for determining loop closures.

The state-of-the-art modern pose-graph optimization (PGO) systems are vertex based. In this context, the number of variables might be high, albeit the number of cycles in the graph (loop closures) is relatively low. For sparse problems particularly, the cycle space has a significantly smaller dimension than the number of vertices. By exploiting this observation, in this article, we propose an alternative solution to PGO that directly exploits the cycle space. We characterize the topology of the graph as a cycle matrix, and reparameterize the problem using relative poses, which are further constrained by a cycle basis of the graph. We show that by using a minimum cycle basis, the cycle-based approach has superior convergence properties against its vertex-based counterpart, in terms of convergence speed and convergence to the global minimum. For sparse graphs, our cycle-based approach is also more time efficient than the vertex-based. As an additional contribution of this work, we present an effective algorithm to compute the minimum cycle basis. Albeit known in computer science, we believe that this algorithm is not familiar to the robotics community. All the claims are validated by experiments on both standard benchmarks and simulated datasets. To foster the reproduction of the results, we provide a complete open-source C++ implementation1of our approach.

Simultaneous Localization and Mapping (SLAM) is considered a mature research field with numerous applications and publicly available open-source systems. Despite this maturity, existing SLAM systems often rely on ad-hoc implementations or are tailored to predefined sensor setups. In this work, we tackle these issues, proposing a novel unified SLAM architecture specifically designed to standardize the SLAM problem and to address heterogeneous sensor configurations. Thanks to its modularity and design patterns, the presented framework is easy to extend, maximizes code reuse and improves computational efficiency. We show in our experiments with a variety of typical sensor configurations that these advantages come without compromising state-of-the-art SLAM performance. The result demonstrates the architecture's relevance for facilitating further research in (multi-sensor) SLAM and its transfer into practical applications.

In this paper we propose an approach to embed multi-dimensional continuous cues in binary feature descriptors used for visual place recognition. The embedding is achieved by extending each feature descriptor with a binary string that encodes a cue and supports the Hamming distance metric. Augmenting the descriptors in such a way has the advantage of being transparent to the procedure used to compare them. We present a concrete application of our methodology, demonstrating the considered type of continuous cue. Additionally, we conducted a broad quantitative and comparative evaluation on that application, covering five benchmark datasets and several state-of-the-art image retrieval approaches in combination with various binary descriptor types.

A Simultaneous Localization and Mapping (SLAM) system is a complex program consisting of several interconnected components with different functionalities such as optimization, tracking or loop detection. Whereas the literature addresses in detail how enhancing the algorithmic aspects of the individual components improves SLAM performance, the modal aspects, such as when to localize, relocalize or close a loop, are usually left aside. In this paper, we address the modal aspects of a SLAM system and show that the design of the modal controller has a strong impact on SLAM performance in particular in terms of robustness against unforeseen events such as sensor failures, perceptual aliasing or kidnapping. We preset a novel taxonomy for the components of a modern SLAM system, investigate their interplay and propose a highly modular architecture of a generic SLAM system using the Unified Modeling LanguageTM(UML) state machine formalism. The result, called SLAM state machine, is compared to the modal controller of several state-of-the-art SLAM systems and evaluated in two experiments. We demonstrate that our state machine handles unforeseen events much more robustly than the state-of-the-art systems.

In this paper, we present a sonar-based navigation system, designed to deploy a fleet of autonomous mobile platforms at a reasonable cost. In educational and hobbyist contexts, a large number of robots is required. By means of classical navigation approaches, every robot should be provided with accurate vision or range sensors. This limits the maximum number of robots in the fleet, due to the unaffordable cost of these sensors. In contrast to that, our system requires a single platform equipped with a higher quality sensor, used to perform calibration and mapping tasks. The rest of the fleet, able to localize and navigate, is equipped solely with low-cost sonars, providing a notable reduction in the overall cost. We achieve this task by presenting a novel calibration procedure to estimate the sonars extrinsic and adapting a classical monte-carlo localization algorithm to the sonar model, focusing on efficiency. We release an open source implementation of the system to the community.

Mobile robots can be considered completely autonomous if they embed active algorithms for Simultaneous Localization And Mapping (SLAM). This means that the robot is able to autonomously, or actively, explore and create a reliable map of the environment, while simultaneously estimating its pose. In this paper, we propose a novel framework to robustly solve the active SLAM problem, in scenarios in which some prior information about the environment is available in the form of a topo-metric graph. This information is typically available or can be easily developed in industrial environments, but it is usually affected by uncertainties. In particular, the distinguishing features of our approach are: the inclusion of prior information for solving the active SLAM problem; the exploitation of this information to pursue active loop closure; the on-line correction of the inconsistencies in the provided data. We present some experiments, that are performed in different simulated environments: the results suggest that our method improves on state-of-the-art approaches, as it is able to deal with a wide variety of possibly large uncertainties.

The ability to build maps is a key functionality for the majority of mobile robots. A central ingredient to most mapping systems is the registration or alignment of the recorded sensor data. In this paper, we present a general methodology for photometric registration that can deal with multiple different cues. We provide examples for registering RGBD as well as 3D LIDAR data. In contrast to popular point cloud registration approaches such as ICP our method does not rely on explicit data association and exploits multiple modalities such as raw range and image data streams. Color, depth, and normal information are handled in an uniform manner and the registration is obtained by minimizing the pixel-wise difference between two multi-channel images. We developed a flexible and general framework and implemented our approach inside that framework. We also released our implementation as open source C++ code. The experiments show that our approach allows for an accurate registration of the sensor data without requiring an explicit data association or model-specific adaptations to datasets or sensors. Our approach exploits the different cues in a natural and consistent way and the registration can be done at framerate for a typical range or imaging sensor.

In this paper we present ProSLAM, a lightweight open-source stereo visual SLAM system designed with simplicity in mind. This work stems from the experience gathered by the authors while teaching SLAM and aims at providing a highly modular system that can be easily implemented and understood. Rather than focusing on the well known mathematical aspects of stereo visual SLAM, we highlight the data structures and the algorithmic aspects required to realize such a system. We implemented ProSLAM using the C++ programming language in combination with a minimal set of standard libraries. The results of a thorough validation performed on several standard benchmark datasets show that ProSLAM achieves precision comparable to state-of-the-art approaches, while requiring substantially less computation.

As autonomous robots are increasingly being introduced in real-world environments operating for long periods of time, the difficulties of long-term mapping are attracting the attention of the robotics research community. This paper proposes a full SLAM system capable of handling the dynamics of the environment across a single or multiple mapping sessions. Using the pose graph SLAM paradigm, the system works on local maps in the form of 2D point cloud data which are updated over time to store the most up-to-date state of the environment. The core of our system is an efficient ICP-based alignment and merging procedure working on the clouds that copes with non-static entities of the environment. Furthermore, the system retains the graph complexity by removing out-dated nodes upon robust inter- and intra-session loop closure detections while graph coherency is preserved by using condensed measurements. Experiments conducted with real data from longterm SLAM datasets demonstrate the efficiency, accuracy and effectiveness of our system in the management of the mapping problem during long-term robot operation.

Selective weeding is one of the key challenges in the field of agriculture robotics. To accomplish this task, a farm robot should be able to accurately detect plants and to distinguish them between crop and weeds. Most of the promising state-of-the-art approaches make use of appearance-based models trained on large annotated datasets. Unfortunately, creating large agricultural datasets with pixel-level annotations is an extremely time consuming task, actually penalizing the usage of data-driven techniques. In this paper, we face this problem by proposing a novel and effective approach that aims to dramatically minimize the human intervention needed to train the detection and classification algorithms. The idea is to procedurally generate large synthetic training datasets randomizing the key features of the target environment (i.e., crop and weed species, type of soil, light conditions). More specifically, by tuning these model parameters, and exploiting a few real-world textures, it is possible to render a large amount of realistic views of an artificial agricultural scenario with no effort. The generated data can be directly used to train the model or to supplement real-world images. We validate the proposed methodology by using as testbed a modern deep learning based image segmentation architecture. We compare the classification results obtained using both real and synthetic images as training data. The reported results confirm the effectiveness and the potentiality of our approach.

This paper describes an unsupervised approach to retrieve the kinematic parameters of a wheeled mobile robot. The robot chooses which action to take in order to minimize the uncertainty in the parameter estimate and to fully explore the parameter space.

In this paper we present an approach for efficiently retrieving the most similar image, based on point-to-point correspondences, within a sequence that has been acquired through continuous camera movement. Our approach is entailed to the use of standardized binary feature descriptors and exploits the temporal form of the input data to dynamically adapt the search structure. While being straightforward to implement, our method exhibits very fast response times and its Precision/Recall rates compete with state of the art approaches. Our claims are supported by multiple large scale experiments on publicly available datasets.

Matching 3D point clouds, a critical operation in map building and localization, is difficult with Velodyne-type sensors due to the sparse and non-uniform point clouds that they produce. Standard methods from dense 3D point clouds are generally not effective. In this paper, we describe a feature-based approach using Principal Components Analysis (PCA) of neighborhoods of points, which results in mathematically principled line and plane features. The key contribution in this work is to show how this type of feature extraction can be done efficiently and robustly even on non-uniformly sampled point clouds. The resulting detector runs in real-time and can be easily tuned to have a low false positive rate, simplifying data association. We evaluate the performance of our algorithm on an autonomous car at the MCity Test Facility using a Velodyne HDL-32E, and we compare our results against the state-of-the-art NARF keypoint detector.

In this paper we present a novel on-line method to recursively align point clouds. By considering each point together with the local features of the surface (normal and curvature), our method takes advantage of the 3D structure around the points for the determination of the data association between two clouds. The algorithm relies on a least squares formulation of the alignment problem, that minimizes an error metric depending on these surface characteristics. We named the approach Normal Iterative Closest Point (NICP in short). Extensive experiments on publicly available benchmark data show that NICP outperforms other state-of-the-art approaches.

Non-linear error minimization methods became widespread approaches for solving the simultaneous localization and mapping problem. If the initial guess is far away from the global minimum, converging to the correct solution and not to a local one can be challenging and sometimes even impossible. This paper presents an experimental analysis of dynamic covariance scaling, a recently proposed method for robust optimization of SLAM graphs, in the context of a poor initialization. Our evaluation shows that dynamic covariance scaling is able to mitigate the effects of poor initializations. In contrast to other methods that first aim at finding a good initial guess to seed the optimization, our method is more elegant because it does not require an additional method for initialization. Furthermore, it can robustly handle data association outliers. Experiments performed with real world and simulated datasets show that dynamic covariance scaling outperforms existing methods, both in the presence and absence of data association outliers.

In this paper we describe a Simultaneous Localization and Mapping (SLAM) approach specifically designed to address the communication and computational issues that affect multi-robot systems. Our method utilizes condensed measurements to exchange map information between the robots. These measurements can effectively compress relevant portions of a map in a few data. This results in a substantial reduction of both the data to be transmitted and processed, that renders the system more robust and efficient. As documented by our simulated and real world experiments, these advantages come with a very little decrease in accuracy compared to ideal (but not realistic) methods that share the full data among all the robots.

In this paper, we propose an approach to obtain highly accurate 3D models from range data. The key idea of our method is to jointly optimize the poses of the sensor and the positions of the surface points measured with a range scanning device. Our approach applies a physical model of the underlying range sensor. To solve the optimization task it employs a state-of-the-art graph-based optimizer and iteratively refines the structure of the error function by recomputing the data associations after each optimization. We present our approach and evaluate it on data recorded in different real world environments with a RGBD camera and a laser range scanner. The experimental results demonstrate that our method is able to substantially improve the accuracy of SLAM results and that it compares favorable over the moving least squares method.

Popular problems in robotics and computer vision like simultaneous localization and mapping (SLAM) or structure from motion (SfM) require to solve a least-squares problem that can be effectively represented by factor graphs. The chance to find the global minimum of such problems depends on both the initial guess and the non-linearity of the sensor models. In this paper we propose an approach to determine an approximation of the original problem that has a larger convergence basin. To this end, we employ a divide-and-conquer approach that exploits the structure of the factor graph. Our approach has been validated on real-world and simulated experiments and is able to succeed in finding the global minimum in situations where other state-of-the-art methods fail.

Recently, there has been increased interest in the development of autonomous flying vehicles. However, as most of the proposed approaches are suitable for outdoor operation, only a few techniques have been designed for indoor environments, where the systems cannot rely on the Global Positioning System (GPS) and, therefore, have to use their exteroceptive sensors for navigation. In this paper, we present a general navigation system that enables a small-sized quadrotor system to autonomously operate in indoor environments. To achieve this, we systematically extend and adapt techniques that have been successfully applied on ground robots. We describe all algorithms and present a broad set of experiments, which illustrate that they enable a quadrotor robot to reliably and autonomously navigate in indoor environments.

Many popular problems in robotics and computer vision including various types of simultaneous localization and mapping (SLAM) or bundle adjustment (BA) can be phrased as least squares optimization of an error function that can be represented by a graph. This paper describes the general structure of such problems and presents g2o, an open-source C++ framework for optimizing graph-based nonlinear error functions. Our system has been designed to be easily extensible to a wide range of problems and a new problem typically can be specified in a few lines of code. The current implementation provides solutions to several variants of SLAM and BA. We provide evaluations on a wide range of real-world and simulated datasets. The results demonstrate that while being general g2o offers a performance comparable to implementations of state of-the-art approaches for the specific problems.

In this paper we describe an algorithm for learning highly accurate laser-based maps that treats the overall mapping problem as a joint optimization problem over robot poses and laser points. We assume that a laser range finder senses points sampled from a regular surface and we utilize an improved likelihood function that accounts for two phenomena affecting the laser measurements that are often neglected: the conic shape of the laser beam and the incidence angle. To solve the entire problem we apply an optimization procedure that jointly adjusts the position of all the robot poses and all points in the scans. As a result, we obtain highly accurate maps. We evaluated our approach using simulated and real-world data and we show that utilizing the estimated maps greatly improves the localization accuracy of robots. The results furthermore suggest that the accuracy of the resulting map can exceed the resolution of the laser sensors used.

In graph-based SLAM, the pose graph encodes the poses of the robot during data acquisition as well as spatial constraints between them. The size of the pose graph has a substantial influence on the runtime and the memory requirements of a SLAM system, which hinders long-term mapping. In this paper, we address the problem of efficient information-theoretic compression of pose graphs. Our approach estimates the expected information gain of laser measurements with respect to the resulting occupancy grid map. It allows for restricting the size of the pose graph depending on the information that the robot acquires about the environment or based on a given memory limit, which results in an any-space SLAM system. When discarding laser scans, our approach marginalizes out the corresponding pose nodes from the graph. To avoid a densely connected pose graph, which would result from exact marginalization, we propose an approximation to marginalization that is based on local Chow-Liu trees and maintains a sparse graph. Real world experiments suggest that our approach effectively reduces the growth of the pose graph while minimizing the loss of information in the resulting grid map.

The calibration parameters of a mobile robot play a substantial role in navigation tasks. Often these parameters are subject to variations that depend either on environmental changes or on the wear of the devices. In this paper, we propose an approach to simultaneously estimate a map of the environment, the position of the on-board sensors of the robot, and its kinematic parameters. Our method requires no prior knowledge about the environment and relies only on a rough initial guess of the platform parameters. The proposed approach performs on-line estimation of the parameters and it is able to adapt to non-stationary changes of the configuration. We tested our approach in simulated environments and on a wide range of real world data using different types of robotic platforms.

A large number of applications use motion capture systems to track the location and the body posture of people. For instance, the movie industry captures actors to animate virtual characters that perform stunts. Today's tracking systems either operate with statically mounted cameras and thus can be used in confined areas only or rely on inertial sensors that allow for free and large-scale motion but suffer from drift in the pose estimate. This paper presents a novel tracking approach that aims to provide globally aligned full body posture estimates by combining a mobile robot and an inertial motion capture system. In our approach, a mobile robot equipped with a laser scanner is used to anchor the pose estimates of a person given a map of the environment. It uses a particle filter to globally localize a person wearing a motion capture suit and to robustly track the person's position. To obtain a smooth and globally aligned trajectory of the person, we solve a least squares optimization problem formulated from the motion capture suite and tracking data. Our approach has been implemented on a real robot and exhaustively tested. As the experimental evaluation shows, our system is able to provide locally precise and globally aligned estimates of the person's full body posture.

In this paper, we present a new hierarchical optimization solution to the graph-based simultaneous localization and mapping (SLAM) problem. During online mapping, the approach corrects only the coarse structure of the scene and not the overall map. In this way, only updates for the parts of the map that need to be considered for making data associations are carried out. The hierarchical approach provides accurate non-linear map estimates while being highly efficient. Our error minimization approach exploits the manifold structure of the underlying space. In this way, it avoids singularities in the state space parameterization. The overall approach is accurate, efficient, designed for online operation, overcomes singularities, provides a hierarchical representation, and outperforms a series of state-of-the-art methods.

The problem of place recognition appears in different mobile robot navigation problems including localization, SLAM, or change detection in dynamic environments. Whereas this problem has been studied intensively in the context of robot vision, relatively few approaches are available for three-dimensional range data. In this paper, we present a novel and robust method for place recognition based on range images. Our algorithm matches a given 3D scan against a database using point features and scores potential transformations by comparing significant points in the scans. A further advantage of our approach is that the features allow for a computation of the relative transformations between scans which is relevant for registration processes. Our approach has been implemented and tested on different 3D data sets obtained outdoors. In several experiments we demonstrate the advantages of our approach also in comparison to existing techniques.

Pose graphs have become a popular representation for solving the simultaneous localization and mapping (SLAM) problem. A pose graph is a set of robot poses connected by nonlinear constraints obtained from observations of features common to nearby poses. Optimizing large pose graphs has been a bottleneck for mobile robots, since the computation time of direct nonlinear optimization can grow cubically with the size of the graph. In this paper, we propose an efficient method for constructing and solving the linear subproblem, which is the bottleneck of these direct methods. We compare our method, called Sparse Pose Adjustment (SPA), with competing indirect methods, and show that it outperforms them in terms of convergence speed and accuracy. We demonstrate its effectiveness on a large set of indoor real-world maps, and a very large simulated dataset. Open-source implementations in C++, and the datasets, are publicly available.

Accurate and robust localization is essential for the successful navigation of autonomous mobile robots. The majority of existing localization approaches, however, is based on the assumption that the environment is static which does not hold for most practical application domains. In this paper, we present a localization framework that can robustly track a robot's pose even in non-static environments. Our approach keeps track of the observations caused by unexpected objects in the environment using temporary local maps. It relies both on these temporary local maps and on a reference map of the environment for estimating the pose of the robot. Experimental results demonstrate that by exploiting the observations caused by unexpected objects our approach outperforms standard localization methods for static environments.

In this paper we describe a novel algorithm for constructing a compact representation of 3D laser range data. Our approach extracts an alphabet of local scans from the scene. The words of this alphabet are used to replace recurrent local 3D structures, which leads to a substantial compression of the entire point cloud. We optimize our model in terms of complexity and accuracy by minimizing the Bayesian information criterion (BIC). Experimental evaluations on large real-world data show that our method allows robots to accurately reconstruct environments with as few as 70 words.

Recently there has been increasing research on the development of autonomous flying vehicles.Whereas most of the proposed approaches are suitable for outdoor operation, only a few techniques have been designed for indoor environments. In this paper we present a general system consisting of sensors and algorithms which enables a small sized flying vehicle to operate indoors. This is done by adapting techniques which have been successfully applied on ground robots. We released our system as open-source with the intention to provide the community with a new framework for building applications for indoor flying robots. We present a set of experiments to validate our system on an open source quadrotor.

We present an algorithm for learning 3D object models from partial object observations. The input to our algorithm is a sequence of 3D laser range scans. Models learned from the objects are represented as point clouds. Our approach can deal with partial views and it can robustly learn accurate models from complex scenes. It is based on an iterative matching procedure which attempts to recursively merge similar models. The alignment between models is determined using a novel scan registration procedure based on range images. The decision about which models to merge is performed by spectral clustering of a similarity matrix whose entries represent the consistency between different models.

In this paper, we address the problem of creating an objective benchmark for comparing SLAM approaches. We propose a framework for analyzing the results of SLAM approaches based on a metric for measuring the error of the corrected trajectory. The metric uses only relative relations between poses and does not rely on a global reference frame. The idea is related to graph-based SLAM approaches in the sense that it considers the energy needed to deform the trajectory estimated by a SLAM approach to the ground truth trajectory. Our method enables us to compare SLAM approaches that use different estimation techniques or different sensor modalities since all computations are made based on the corrected trajectory of the robot. We provide sets of relative relations needed to compute our metric for an extensive set of datasets frequently used in the SLAM community. The relations have been obtained by manually matching laser-range observations. We believe that our benchmarking framework allows the user an easy analysis and objective comparisons between different SLAM approaches.

A mobile robot that accomplishes high level tasks needs to be able to classify the objects in the environment and to determine their location. In this paper, we address the problem of online object detection in 3D laser range data. The object classes are represented by 3D point-clouds that can be obtained from a set of range scans. Our method relies on the extraction of point features from range images that are computed from the point-clouds. Compared to techniques that directly operate on a full 3D representation of the environment, our approach requires less computation time while retaining the robustness of full 3D matching. Experiments demonstrate that the proposed approach is even able to deal with partially occluded scenes and to fulfill the runtime requirements of online applications.

Learning models of the environment is one of the fundamental tasks of mobile robots since maps are needed for a wide range of robotic applications, such as navigation and transportation tasks, service robotic applications, and several others. In the past, numerous efficient approaches to map learning have been proposed. Most of them, however, assume that the robot lives on a plane. In this paper, we present a highly efficient maximum-likelihood approach that is able to solve 3-D and 2-D problems. Our approach addresses the so-called graph-based formulation of simultaneous localization and mapping (SLAM) and can be seen as an extension of Olson's algorithm toward non-flat environments. It applies a novel parameterization of the nodes of the graph that significantly improves the performance of the algorithm and can cope with arbitrary network topologies. The latter allows us to bound the complexity of the algorithm to the size of the mapped area and not to the length of the trajectory. Furthermore, our approach is able to appropriately distribute the roll, pitch, and yaw error over a sequence of poses in 3-D mapping problems. We implemented our technique and compared it with multiple other graph-based SLAM solutions. As we demonstrate in simulated and real-world experiments, our method converges faster than the other approaches and yields accurate maps of the environment.

In this paper, we address the problem of incrementally optimizing constraint networks for maximum likelihood map learning. Our approach allows a robot to efficiently compute configurations of the network with small errors while the robot moves through the environment. We apply a variant of stochastic gradient descent and use a tree-based parameterization of the nodes in the network. By integrating adaptive learning rates in the parameterization of the network, our algorithm can use previously computed solutions to determine the result of the next optimization run. Additionally, our approach updates only the parts of the network which are affected by the newly incorporated measurements and starts the optimization approach only if the new data reveals inconsistencies with the network constructed so far. These improvements yield an efficient solution for this class of online optimization problems. Our approach has been implemented and tested on simulated and on real data. We present comparisons to recently proposed online and offline methods that address the problem of optimizing constraint network. Experiments illustrate that our approach converges faster to a network configuration with small errors than the previous approaches.

Humanoids have recently become a popular research platform in the robotics community. Such robots offer various fields for new applications. However, they have several drawbacks compared to wheeled vehicles such as stability problems, limited payload capabilities, violation of the flat world assumption, and they typically provide only very rough odometry information, if at all. In this paper, we investigate the problem of learning accurate grid maps with humanoid robots. We present techniques to deal with some of the above-mentioned difficulties. We describe how an existing approach to the simultaneous localization and mapping (SLAM) problem can be adapted to robustly learn accurate maps with a humanoid equipped with a laser range finder. We present an experiment in which our mapping system builds a highly accurate map with a size of around 20 m by 20 m using data acquired with a humanoid in our office environment containing two loops. The resulting maps have a similar accuracy as maps built with a wheeled robot.

The ability to learn a map of the environment is important for numerous types of robotic vehicles. In this paper, we address the problem of learning a visual map of the ground using flying vehicles. We assume that the vehicles are equipped with one or two low-cost downlooking cameras in combination with an attitude sensor. Our approach is able to construct a visual map that can later on be used for navigation. Key advantages of our approach are that it is comparably easy to implement, can robustly deal with noisy camera images, and can operate either with a monocular camera or a stereo camera system. Our technique uses visual features and estimates the correspondences between features using a variant of the progressive sample consensus (PROSAC) algorithm. This allows our approach to extract spatial constraints between camera poses that can then be used to address the simultaneous localization and mapping (SLAM) problem by applying graph methods. Furthermore, we address the problem of efficiently identifying loop closures. We performed several experiments with flying vehicles that demonstrate that our method is able to construct maps of large outdoor and indoor environments.

Learning maps is one of the fundamental tasks of mobile robots. In the past, numerous efficient approaches to map learning have been proposed. Most of them, however, assume that the robot lives on a plane. In this paper, we consider the problem of learning maps with mobile robots that operate in non-flat environments and apply maximum likelihood techniques to solve the graph-based SLAM problem. Due to the non-commutativity of the rotational angles in 3D, major problems arise when applying approaches designed for the two-dimensional world. The non-commutativity introduces serious difficulties when distributing a rotational error over a sequence of poses. In this paper, we present an efficient solution to the SLAM problem that is able to distribute a rotational error over a sequence of nodes. Our approach applies a variant of gradient descent to solve the error minimization problem. We implemented our technique and tested it on large simulated and real world datasets. We furthermore compared our approach to solving the problem by LU-decomposition. As the experiments illustrate, our technique converges significantly faster to an accurate map with low error and is able to correct maps with bigger noise than existing methods.

Particle filters are a frequently used filtering technique in the robotics community. They have been successfully applied to problems such as localization, mapping, or tracking. The particle filter framework allows the designer to freely choose the proposal distribution which is used to obtain the next generation of particles in estimating dynamical processes. This choice greatly influences the performance of the filter. Many approaches have achieved good performance through informed proposals which explicitly take into account the current observation. A popular approach is to approximate the desired proposal distribution by a Gaussian. This paper presents a statistical analysis of the quality of such Gaussian approximations. We also propose a way to obtain the optimal proposal in a non-parametric way and then identify the error introduced by the Gaussian approximation. Furthermore, we present an alternative sampling strategy that better deals with situations in which the target distribution is multi-modal. Experimental results indicate that our alternative sampling strategy leads to accurate maps more frequently that the Gaussian approach while requiring only minimal additional computational overhead.

In this paper, we address the problem of learning 3D maps of the environment using a cheap sensor setup which consists of two standard web cams and a low cost inertial measurement unit. This setup is designed for lightweight or flying robots. Our technique uses visual features extracted from the web cams and estimates the 3D location of the landmarks via stereo vision. Feature correspondences are estimated using a variant of the PROSAC algorithm. Our mapping technique constructs a graph of spatial constraints and applies an efficient gradient descent-based optimization approach to estimate the most likely map of the environment. Our approach has been evaluated in comparably large outdoor and indoor environments. We furthermore present experiments in which our technique is applied to build a map with a blimp.

Smoothing and optimization approaches are an effective means for solving the simultaneous localization and mapping (SLAM) problem. Most of the existing techniques focus mainly on determining the most likely map and leave open how to efficiently compute the marginal covariances. These marginal covariances, however, are essential for solving the data association problem. In this paper we present a novel algorithm for computing an approximation of the marginal. In experiments we demonstrate that our approach outperforms two commonly used techniques, namely loopy belief propagation and belief propagation on a spanning tree. Compared to these approaches, our algorithm yields better estimates while preserving the same time complexity.

Recently, Rao-Blackwellized particle filters (RBPF) have been introduced as an effective means to solve the simultaneous localization and mapping problem. This approach uses a particle filter in which each particle carries an individual map of the environment. Accordingly, a key question is how to reduce the number of particles. In this paper, we present adaptive techniques for reducing this number in a RBPF for learning grid maps. We propose an approach to compute an accurate proposal distribution, taking into account not only the movement of the robot, but also the most recent observation. This drastically decreases the uncertainty about the robot's pose in the prediction step of the filter. Furthermore, we present an approach to selectively carry out resampling operations, which seriously reduces the problem of particle depletion. Experimental results carried out with real mobile robots in large-scale indoor, as well as outdoor, environments illustrate the advantages of our methods over previous approaches

Recently, Rao-Blackwellized particle filters have become a popular tool to solve the simultaneous localization and mapping problem. This technique applies a particle filter in which each particle carries an individual map of the environment. Accordingly, a key issue is to reduce the number of particles and/or to make use of compact map representations. This paper presents an approximative but highly efficient approach to mapping with Rao-Blackwellized particle filters. Moreover, it provides a compact map model. A key advantage is that the individual particles can share large parts of the model of the environment. Furthermore, they are able to re-use an already computed proposal distribution. Both techniques substantially speed up the overall process and reduce the memory requirements. Experimental results obtained with mobile robots in large-scale indoor environments and based on published, standard datasets illustrate the advantages of our methods over previous Rao-Blackwellized mapping approaches

Scan matching is used as a building block in many robotic applications, for localization and simultaneous localization and mapping (SLAM). Although many techniques have been proposed for scan matching in the past years, more efficient and effective scan matching procedures allow for improvements of such associated problems. In this paper we present a new scan matching method that, exploiting the properties of the Hough domain, allows for combining advantages of dense scan matching algorithms with feature-based ones.

Recently Rao-Blackwellized particle filters have been introduced as effective means to solve the simultaneous localization and mapping (SLAM) problem. This approach uses a particle filter in which each particle carries an individual map of the environment. Accordingly, a key question is how to reduce the number of particles. In this paper we present adaptive techniques to reduce the number of particles in a Rao-Blackwellized particle filter for learning grid maps. We propose an approach to compute an accurate proposal distribution taking into account not only the movement of the robot but also the most recent observation. This drastically decrease the uncertainty about the robot's pose in the prediction step of the filter. Furthermore, we present an approach to selectively carry out re-sampling operations which seriously reduces the problem of particle depletion. Experimental results carried out with mobile robots in large-scale indoor as well as in outdoor environments illustrate the advantages of our methods over previous approaches.

Acquiring models of the environment belongs to the fundamental tasks of mobile robots. Approaches addressing the problem of simultaneous localization and mapping (SLAM) typically process the perceived sensor data and do not influence the motion of the mobile robot. In this paper, we present an approach to actively closing loops during exploration. It applies a Rao-Blackwellized particle filter to maintain multiple hypotheses about potential trajectories of the robot and corresponding maps. To prevent the particle filter from becoming overly confident, we present a technique to recover the particle diversity after successfully closing a loop. This way the particle depletion problem is avoided. The combination of our approach with the active loop closing strategy allows to deal with multiple nested loops. Experimental results presented in this paper illustrate the advantage of our method over pervious approaches to mapping with Rao-Blackwellized particle filters.

The activities of search and rescue of victims in large-scale disasters are very relevant social problems, and from a scientific viewpoint raise many different technical problems in the fields of artificial intelligence, robotics and multi agent systems. In this paper we describe the development of a multi agent system based on the RoboCup Rescue simulator to allow monitoring and decision support, that are needed in a rescue operation. Two significant accomplishments are reported in this paper: the first is a framework for cognitive agent development that provides for the capabilities of information fusion, planning and coordination; the second one is a methodology for evaluation of multi-agent systems in this scenario that aims at measuring not only the efficiency of a system, but also its robustness when conditions in the environment change.

Knowing the position of a mobile robot in the environment in which it operates is an important element for effectively accomplishing complex tasks requiring autonomous navigation. Among several existing techniques for robot self-localization, a new approach called Hough localization was proposed for map matching in the Hough domain, that turned out to be reliable and efficient for position tracking in polygonal environments. In this paper we present an extension of Hough localization which is able to deal with the global localization problem, in which the robot does not know its initial position in the environment.