2014 IEEE International Conference on Robotics and Automation

ICRA14 Workshop on Modelling, Estimation, Perception and Control of All Terrain Mobile Robots

Full Day Workshop, Track T11, Room 425, 8:25-17:30

June 1st, 2014, Hong-Kong, China

Workshop Proceedings

Contact : Professor Philippe Martinet
IRCCyN-CNRS Laboratory, Ecole Centrale de Nantes,
Email: Philippe.Martinet@irccyn.ec-nantes.fr,
Home page: http://www.irccyn.ec-nantes.fr/~martinet



Final program

Introduction to the workshop 8:25

Session I: Enhanced mobility & Modeling 8:30
Chairman: Kasuya Yoshida (Tohoku University, Japan)

  • Invited Talk: Genya Ishigami (Dept. of Mechanical Engineering, Keio University, Japan)
    Invited Talk: Rough Terrain Mobility: key issues and approaches for dynamics simulation of rough terrain mobile robot
    8:30
    Keynote speaker: Genya Ishigami (Dept. of Mechanical Engineering, Keio University, Japan) 35min + 5min questions
    Presentation Video1, Video2, Video3, Video4, Video5, Video6, Video7, Video8, Video9, Video10, Video11, Video12, Video13

    Abstract: Rough-terrain mobile robots are always subject to complicated dynamic interaction between their running gears (tire, wheel, or track) and ground. A well-defined mechanics for the robot-terrain interaction is of importance to the following technical aspects of the mobile robot: (1) mobility analysis such as slope traversability or obstacle crossing; (2) robot navigation, planning, and traction control; and (3) design of vehicle dimensions, suspension, and actuators. This presentation focuses on a topic related to a dynamics simulation of rough-terrain mobile robot using wheel contact mechanics. An overview of research and development of the dynamics simulation are described along with its application to mobility analysis, control, and design. In the presentation, typical issues and key approaches towards a next generation of dynamics simulation tools for rough-terrain mobile robot are also discussed.

  • Regular Talk: Analyzing the Impact of Learning Inputs on Near-to-Far Terrain Traversability Estimation 9:10
    Authors: K. Ho, T. Peynot and S. Sukkarieh 15min + 3min questions
    Paper, Presentation

    Abstract: With the increasing need to adapt to new environments, data-driven approaches have been developed to estimate terrain traversability by learning the roverís response on the terrain based on experience. Multiple learning inputs are often used to adequately describe the various aspects of terrain traversability. In a complex learning framework, it can be difficult to identify the relevance of each learning input to the resulting estimate. This paper addresses the suitability of each learning input by systematically analyzing the impact of each input on the estimate. Sensitivity Analysis (SA) methods provide a means to measure the contribution of each learning input to the estimate variability. Using a variance-based SA method, we characterize how the prediction changes as one or more of the input changes, and also quantify the prediction uncertainty as attributed from each of the inputs in the framework of dependent inputs. We propose an approach built on Analysis of Variance (ANOVA) decomposition to examine the prediction made in a near-to-far learning framework based on multi-task GP regression. We demonstrate the approach by analyzing the impact of driving speed and terrain geometry on the prediction of the roverís attitude and chassis configuration in a Marsanalogue terrain using our prototype rover Mawson.

  • Regular Talk: Terrain mapping with a pan and tilt stereo camera for locomotion on a quadruped robot 9:28
    Authors: S. Bazeille, M. Camurri, J. Ortiz, I. Havoutis, D. G. Caldwell, and C. Semini 15min + 3min questions
    Paper, Presentation Video1, Video2, Video3

    Abstract: Legged robots are expected to have superior mobility on rough terrain than wheeled robots. The main reason is that legged locomotion is more adaptable to a wide range of terrain types as the robot can decompose its path into a sequence of footholds and can use different locomotion strategies. In order to accomplish most of the locomotion tasks the robot requires high level control (i.e., to adjust the locomotion parameters and to choose optimal footholds) which depends on real-time localization and accurate terrain mapping. In this paper, we propose a SLAM solution using a pan and tilt stereo camera mounted on an hydraulically actuated quadruped robot that builds a map and keeps track of the robotís position. Since the computation needs to be carried out on board and the robot is subject to considerable motion during its locomotion (regular vibrations, impacts or slippages), we developed a dedicated implementation based on fast stereo depth computation, GPU based map building and mechanical motion compensation. Combined with a foothold planning framework presented in our previous work [1], this localization and mapping ability allows to perform locomotion in a fully planned manner. Successful results of foothold planning with our quadruped robot show the effectiveness of our method.

Session II: Perception in outdoor environment (1/2) 9:46
Chairman: Juan I. Nieto (University of Sydney, Australia)
  • Invited Talk: Paul Furgale (ETH Zurich, Switzerland)
    Invited Talk: There and back again: Dealing with highly-dynamic scenes and long-term change during topological/metric route following 9:46
    Keynote speaker: Paul Furgale (ETH Zurich, Switzerland) 35min + 5min questions
    Presentation
    Co-authors: P. KrŁsi, F. Pomerleau, U. Schwesinger, F. Colas, and R. Siegwart

    Abstract: Topological/metric route following, also called teach and repeat (T&R), enables long-range autonomous navigation even without globally consistent localization. This renders T&R ideal for applications where a global positioning system may not be available, such as navigation through street canyons or forests in search and rescue, reconnaissance in underground structures, surveillance, or planetary exploration. This Talk will present our efforts to develop a T&R system suitable for long-term robot autonomy in highly dynamic, unstructured environments. We use the fast iterative closest point (ICP) algorithms from libpointmatcher (https://github.com/ethz-asl/libpointmatcher) to build a T&R system based on a spinning laser range finder. The system deals with dynamic elements in two ways. First, we employ a system-compliant local motion planner to react to dynamic elements in the scene during route following. Second, the system infers the static or dynamic state of each 3D point in the environment based on repeated observations. The velocity of each dynamic point is estimated without requiring object models or explicit clustering of the points. At any time, the system is able to produce a most-likely representation of underlying static scene geometry. By storing the time history of velocities, we can infer the dominant motion patterns within the map. The result is an online mapping and localization system specifically designed to enable long-term autonomy within highly dynamic environments. We validate the approach using data collected around the campus of ETH Zurich over seven months and at an outdoor 3D test site in Thun, Switzerland.

  • Coffee Break 10:26-10:50

  • Regular Talk: Monocular Vision: A Real-Time Perception Toolkit for Mobile Robots in Outdoor Environments 10:50
    Authors: A. Miranda Neto, A. C. Victorino and J. V. Ferreira 15min + 3min questions
    Paper, Presentation Video1, Video2, Video3, Video4, Video5, Video6, Video7, Video8, Video9

    Abstract: Many applications for control of autonomous platform are being developed and some important aspects are: (a) the estimation of drivable image area and (b) the excess of information, frequently redundant, that imposes a great computational cost in data processing. In this way, we have proposed (i) a robust algorithm for detecting the horizon line to generate (ii) the navigable area. It permits to investigate dynamically only a small portion of the image (road) ahead of the vehicle. Moreover, taking into account the temporal coherence between consecutive frames, we also have proposed a set of tools based on