List of all recorded talks

  • Design and Experimental Characterization of an Omnidirectional Unmanned Ground Vehicle for Outdoor Terrain Authors: Nie, chenghui; Hauschka, Guillaume; Spenko, Matthew
    This paper presents the design and experimental characterization of an omnidirectional unmanned ground vehicle built to operate on a wide variety of real-world terrains. The vehicle can change its orientation and direction of travel regardless of its current kinematic configuration and without significantly decreasing its speed. This gives it the advantage of having high mobility in relatively tight and confined spaces compared to vehicles that utilize skid or Ackermann type steering mechanisms. The vehicle described here utilizes conventional wheels, which gives it several advantages over other omnidirectional vehicle designs that use specialized wheels with small, slender rollers that can become clogged with dirt and debris commonly encountered in outdoor environments. The focus of the paper is on how the concept of kinematic isotropy affects the mechanical design of the system and the experimental results used to validate the design.
  • Unsupervised Incremental Learning for Long-Term Autonomy Authors: Ott, Lionel; Ramos, Fabio
    Abstract— We present an approach to automatically learn the visual appearance of an environment in terms of object classes. The procedure is totally unsupervised, incremental, and can be executed in real time. The traversability property of an unseen object is also learnt without human supervision by the interaction between the robot and the environment. An incremental version of affinity propagation, a state-of-the- art clustering procedure, is used to cluster image patches into groups of similar visual appearance. For each of these clusters, we obtain the probability of representing an obstacle through the interaction of the robot with the environment. This information then allows the robot to navigate safely through the environment based solely on visual information. Experimental results show that our method extracts meaningful clusters from the images and learns the appearance of objects efficiently. We show that the approach generalises well to both indoor and outdoor environments and that the amount of learning reduces as the robot explores the environment. This is a fundamental property for autonomous adaptation and long-term autonomy.
  • A Psychological Scale for General Impressions of Humanoids Authors: Kamide, Hiroko; Mae, Yasushi; Kawabe, Koji; Shigemi, Satoshi; Hirose, Masato; Arai, Tatsuo
    This study identifies the basic general perspectives that ordinary people use to evaluate humanoids (Study 1). In addition, it develops a new psychological scale to quantify general impressions regarding humanoids based on these basic perspectives (Study 2). In Study 1, to discover the basic perspectives toward humanoids, we used 11 humanoids and collected data from 919 Japanese people ranging from teenagers to people in their 70s in three big cities. We asked people to describe their impressions in free text about the robots. Five psychologists analyzed the qualitative data to categorize all the descriptions into several categories. Then, we made the items based on the obtained descriptions to construct a new psychological scale for evaluating general impressions regarding humanoids. In Study 2, we asked 2,624 Japanese who did not participate in Study 1 to evaluate 11 humanoids on the developed scale. Factor analysis showed that nine factors should be used for evaluating the general impressions regarding humanoids: Familiarity, Repulsion, Performance, Utility, Motion, Sound, Voice, Humanness, and Entitativity. The factor structure is clear and its reliability as a psychological scale is satisfactorily high. Finally, we discuss the usability of the new scale.
  • What Could Move? Finding Cars, Pedestrians and Bicyclists in 3D Laser Data Authors: Wang, Dominic Zeng; Posner, Ingmar; Newman, Paul
    This paper tackles the problem of segmenting things that could move from 3D laser scans of urban scenes. In particular, we wish to detect instances of classes of interest in autonomous driving applications - cars, pedestrians and bicyclists - amongst significant background clutter. Our aim is to provide the layout of an end-to-end pipeline which, when fed by a raw stream of 3D data, produces distinct groups of points which can be fed to downstream classifiers for categorisation. We postulate that, for the specific classes considered in this work, solving a binary classification task (i.e. separating the data into foreground and background first) outperforms approaches that tackle the multi-class problem directly. This is confirmed using custom and third-party datasets gathered of urban street scenes. While our system is agnostic to the specific clustering algorithm deployed we explore the use of a Euclidean Minimum Spanning Tree for an end-to-end segmentation pipeline and devise a RANSAC-based edge selection criterion.
  • Scaled-Up Helical Nanobelt Modeling and Simulation at Low Reynolds Numbers Authors: Xu, Tiantian; Hwang, Gilgueng; Régnier, Stéphane; Andreff, Nicolas
    Micro and nanorobots which enable targeted diagnosis and therapy, minimal invasive surgery, can change many aspects of medicine. A helical nanobelt with a magnetic head was proposed as a microrobot driven by rotating magnetic field in prior works. Magnetically coated tails were already shown in some works. However the control of such surface magnetic tails is not clearly realized yet. This paper aims to obtain control parameters with modeling and simulation the influences of surface magnets onto the swimming performances. For this, we created scaled-up helical nanobelts and the experimental testbed to get the control parameters and to achieve a closed-loop control in the future.
  • A Comprehensive Pressure-Sinkage Model for Small-Wheeled Unmanned Ground Vehicles on Dilative, Deformable Terrain Authors: Meirion-Griffith, Gareth; Spenko, Matthew
    This paper details a novel pressure-sinkage model for small-diameter wheels on dilative soils. Pressure-sinkage models are fundamental to the prediction of UGV mobility on deformable terrains. The proposed model builds on previous work in which the flat-plate pressure-sinkage assumption of classical terramechanics was shown to yield diminished accuracy for UGVs with wheels less than 50 cm in diameter. It has been shown that classical pressure-sinkage models can be modified by a diameter dependent term, yielding greatly increased accuracy. Here, an investigation into the effect of wheel width on the diameter-dependent model is detailed. Results from over 250 pressure-sinkage tests on three soils using 85 wheel geometries are summarized. The results of this investigation are used to create a comprehensive pressure-sinkage model for dilative soils that includes wheel width and diameter parameters. The physics of the model are visually validated with X-ray images of sub-surface soil deformation during the wheel indentation process. A comparison between the dilative soil pressure-sinkage model and a previously obtained model for compactive soils is also presented. The pressure-sinkage model presented here can be used to improve the accuracy of the terramechanics framework and UGV mobility predictions.
  • How Was Your Day? Online Visual Workspace Summaries Using Incremental Clustering in Topic Space Authors: Paul, Rohan; Rus, Daniela; Newman, Paul
    Someday mobile robots will operate continually. Day after day, they will be in receipt of a never ending stream of images. In anticipation of this, this paper is about having a mobile robot generate apt and compact summaries of its life experience. We consider a robot moving around its environment both revisiting and exploring, accruing images as it goes. We describe how we can choose a subset of images to summarise the robot's cumulative visual experience. Moreover we show how to do this such that the time cost of generating an summary is largely independent of the total number of images processed. No one day is harder to summarise than any other.
  • Semantic Map Segmentation Using Function-Based Energy Maximization Authors: Sjöö, Kristoffer
    This work describes the automatic segmentation of 2-dimensional indoor maps into semantic units along lines of spatial function, such as connectivity or objects used for certain tasks. Using a conceptually simple and readily extensible energy maximization framework, segmentations similar to what a human might produce are demonstrated on several real-world datasets. In addition, it is shown how the system can perform reference resolution by adding corresponding potentials to the energy function, yielding a segmentation that responds to the context of the spatial reference.
  • Computing Occupancy Grids from Multiple Sensors Using Linear Opinion Pools Authors: Adarve, Juan David; Perrollaz, Mathias; Makris, Alexandros; Laugier, Christian
    Perception is a key component for any robotic system. In this paper we present a method to construct occupancy grids by fusing sensory information using Linear Opinion Pools. We used lidar sensors and a stereo-vision system mounted on a vehicle to make the experiments. To perform the validation, we compared the proposed method with the fusion method previously used in the Bayesian Occupancy Filter framework, using real data taken from road and urban scenarios. The results show that our method is better at dealing with conflicting information coming from the sensors. We propose an implementation on parallel hardware which allows real-time execution.
  • Rotation of Bacteria Sheet Driven Micro Gear in Open Micro Channel Authors: Miyamoto, Tatsuya; Kojima, Masaru; Nakajima, Masahiro; Homma, Michio; Fukuda, Toshio
    Recently, micro-nano robots intended for application to various fields are developed. However, motors which are robot's power are not yet practical. In this paper, to realize the motor which can be applied to micro-nano robot's power source, we established bio-motor by using surface swarming of Vibrio alginolyticus. First, we succeeded in driving the micro gear in the closed micro channel. Next, to transmit power of rotational movement, we fabricated the open micro channel, and succeeded in driving the gear in the open micro channel similarly. In addition, we revealed that ratchet type gear rotated faster than other type. Finally, we assembled the micro gear with shaft. Therefore, it became possible transmitting the power from rotational movement to outside. Thus, we achieved to construct base of bio-motor.