The International Journal of Robotics Research

This paper introduces a new urban point cloud dataset for automatic segmentation and classification acquired by mobile laser scanning (MLS). We describe how the dataset is obtained from acquisition to post-processing and labeling. This dataset can be used to train pointwise classification algorithms; however, given that a great attention has been paid to the split between the different objects, this dataset can also be used to train the detection and segmentation of objects. The dataset consists of around 2 k m of MLS point cloud acquired in two cities. The number of points and range of classes mean that it can be used to train deep-learning methods. In addition, we show some results of automatic segmentation and classification. The dataset is available at: http://caor-mines-paristech.fr/fr/paris-lille-3d-dataset/.

This paper presents a dataset of natural language instructions for object reference in manipulation scenarios. It comprises 1582 individual written instructions, which were collected via online crowdsourcing. This dataset is particularly useful for researchers who work in natural language processing, human–robot interaction, and robotic manipulation. In addition to serving as a rich corpus of domain-specific language, it provides a benchmark of image–instruction pairs to be used in system evaluations and uncovers inherent challenges in tabletop object specification. Example code is provided for easy access via Python.

The ability to model the surrounding space and determine which areas are occupied is of key importance in many robotic applications, ranging from grasping and manipulation to path planning and obstacle avoidance. Occupancy modeling is often hindered by several factors, such as: real-time constraints, that require quick updates and access to estimates; quality of available data, that may contain gaps and partial occlusions; and memory requirements, especially for large-scale environments. In this work we propose a novel framework that elegantly addresses all these issues, by producing an efficient non-stationary continuous occupancy function that can be efficiently queried at arbitrary resolutions. Furthermore, we introduce techniques that allow the learning of individual features for different areas of the input space, that are better able to model its contained information and promote a higher-level understanding of the observed scene. Experimental tests were conducted on both simulated and real large-scale datasets, showing how the proposed framework rivals current state-of-the-art techniques in terms of computational speed while achieving a substantial decrease (of orders of magnitude) in memory requirements and demonstrating better interpolative powers, that are able to smooth out sparse and noisy information.

In this paper we propose a simultaneous localization and mapping (SLAM) back-end solution called the exactly sparse delayed state filter on Lie groups (LG-ESDSF). We derive LG-ESDSF and demonstrate that it retains all the good characteristics of the classic Euclidean ESDSF, the main advantage being the exact sparsity of the information matrix. The key advantage of LG-ESDSF in comparison with the classic ESDSF lies in the ability to respect the state space geometry by negotiating uncertainties and employing filtering equations directly on Lie groups. We also exploit the special structure of the information matrix in order to allow long-term operation while the robot is moving repeatedly through the same environment. To prove the effectiveness of the proposed SLAM solution, we conducted extensive experiments on two different publicly available datasets, namely the KITTI and EuRoC datasets, using two front-ends: one based on the stereo camera and the other on the 3D LIDAR. We compare LG-ESDSF with the general graph optimization framework ( g 2 o ) when coupled with the same front-ends. Similarly to g 2 o the proposed LG-ESDSF is front-end agnostic and the comparison demonstrates that our solution can match the accuracy of g 2 o , while maintaining faster computation times. Furthermore, the proposed back-end coupled with the stereo camera front-end forms a complete visual SLAM solution dubbed LG-SLAM. Finally, we evaluated LG-SLAM using the online KITTI protocol and at the time of writing it achieved the second best result among the stereo odometry solutions and the best result among the tested SLAM algorithms.

This paper presents a new class of animatronic soft robots created by a desktop fabrication mechanism called additive folding. In this method, two-dimensional (2D) slices are threaded by multiple strings, accordion-folded by flexure hinges and finally stacked into a predefined three-dimensional (3D) structure. As the 3D assembly of the slices is controlled by embedded strings, it becomes an animatronic soft robot that moves like a biological creature and that shows life-like movements. We create a computational design algorithm that takes as input a desired 3D geometry of the robot, and that produces a 2D surface with built-in folds and string-based actuators. This paper describes the entire robot design process and demonstrates various animatronic motions, highlighting the vision of desktop fabrication technology and its potential applications in animatronics and robotic art.

Origami shape transformation is dictated by predefined folding patterns and their folding sequence. The working principle of robotic origami is based on the same principle: we design quasi-two-dimensional tiles and connecting hinges and define and program their folding sequences. Since the tiles are often of uniform shape and size, their final configuration is governed by the kinematic relationship. Mathematicians, computer scientists and even architects have studied a wide range of origami algorithms. However, for multiple shape transformations, the origami design parameters and consequently sequence planning become more challenging. In this work, we present a reconfigurable interactive interface, a physics-based modeling control interface to explore the design space of origami robots. We developed two interactive modes for proof of concept of a bidirectional communication interface between virtual and physical environments. The first interaction mode is origami-inspired, foldable surfaces with distributed sensors that can recreate folding sequences and shape transformations in a virtual environment via hardware-in-loop simulation. Its complementary digital transcription lays the foundation for a robotic origami design tool that provides visual representation of various design formulations as well as an intuitive controller for robotic origami. In the second interaction mode, we construct a physics-based modeling interface for intuitive user manipulation of robotic origami in a virtual environment. Algorithms for graphical representation and command transformation were developed for robotic interaction. Lastly, we tested the efficacy of the algorithms on prototypes to discover the applications and capacities of the reconfigurable interactive interface.

A unified formulation that accounts for the dynamics of a general class of aquatic multi-body, soft-structured robots is presented. The formulation is based on a Cosserat formalism where the description of the ensemble of geometrical entities, such as shells and beams, gives rise to a multi-soft-body system capable of simulating both manipulation and locomotion. Conceived as an advanced tool for a priori hardware development, n-degree-of-freedom dynamics analysis and control design of underwater, soft, multi-body, vehicles, the model is validated against aquatic locomotion experiments of an octopus-inspired soft unmanned underwater robot. Upon validation, the general applicability of the model is demonstrated by predicting the self-propulsion dynamics of a diverse range of new viable combinations of multi-soft-body aquatic system.