IEEE/RSJ IROS'13 International Workshop on
Vision-based Closed-Loop Control and Navigation of
Micro Helicopters in GPS-denied Environments
November 7, 2013 - Room 607 - Tokyo Big Sight, Japan
- 09:05 - The Role of Computer Vision for Autonomous Quadrotor MAV's
Friedrich Fraundorfer, Lionel Heng, Marc Pollefeys - TU Munich and ETH Zurich
The talk will discuss our camera based autonomous quadrotor MAV, the Pixhawk system. A special focus will be on the sensor systems that facilitate autonomous flight, in particular the optical flow camera and the stereo system. Furthermore this talk will highlight the role that computer vision played in the development of the system and discuss the role and importance of computer vision for MAV's in general.
- 09:35 - Vision Based Navigation, Grasping and Localization for Micro Aerial Vehicles
Giuseppe Loianno and Vijay Kumar - University of Pennsylnavia
This talk will provide an overview of different approaches for vision based autonomous navigation, grasping and cooperative mapping by aerial robots with an emphasis on micro UAVs. Scientific and technological challenges directions, for future research in these fields, will be discussed.
- 10:05 - Optic-Flow Based Control of a 46g Quadrotor [PDF]
Adrien Briod, Jean-Christophe Zufferey, Dario Floreano - EPFL
INCLUDES INDOOR FLIGHT DEMONSTRATION
This talk will describe a novel ego-motion estimation algorithm for hovering robots equipped with inertial and optic-flow sensors that runs in realtime on a microcontroller. Unlike many vision-based methods, this algorithm does not rely on feature tracking, structure estimation, additional distance sensors or assumptions about the environment. Key to this method is the introduction of the translational optic-flow direction constraint, which does not use the optic-flow scale, but only its direction to correct for inertial sensor drift during changes of direction. This solution requires comparatively much simpler electronics and sensors and works in environments of any geometries. We demonstrate the implementation of this algorithm on a miniature 46g quadrotor for closed-loop position control
- 10:25 - Poster and Coffee Session
- 11:00 - Sensing and actuating with unmanned aerial vehicles; from single to networked systems
Fernando Caballero and Anibal Ollero - University of Seville
This talk will present the application of unmanned aerial vehicles to different tasks involving environments perception and system actuation. Applications such as perception for aerial grasping, automatic detection of events or multi-UAV cooperation will be presented. Thus, algorithms for distributed data fusion, robot localization involving multiple sensors and robot collaboration will be presented in detail. The talk will discuss about the challenges and the opportunities that networked aerial vehicles could bring to the robotics community. Outdoor and indoor experimental results involving real UAVs will be presented and discussed.
- 11:30 - Next Generation Autonomous Micro Helicopters
Stephan Weiss and Larry Matthies - California Institute of Technology
The research on autonomous MAVs and in particular on autonomous micro helicopters rapidly advanced in the last years. A few years ago, designing a simple low level attitude controller for this type of platforms was state-of-the-art. Nowadays flying such vehicles position controlled only based on visual cues is the standard. In this talk we focus on the necessary steps to take in order to bring true autonomy and robustness to such MAVs. We will first discuss the failure modes of the current vision based MAV navigation approaches and present a novel approach for state estimation to maintain the MAV airborne in a fail-safe manner. We will then shed light to a novel vision based local obstacle avoidance approach. While fail-safe state estimation and obstacle avoidance are the core ingredients for keeping an MAV airborne, we will also show initial results on vision based path planning and landing site detection for the next generation of autonomous MAVs.
- 12:00 - Air-ground Collaboration, Dense Mapping, and Aggressive Manueuvers: Challenges and Next Generation Sensors Davide Scaramuzza - University of Zurich
INCLUDES INDOOR FLIGHT DEMONSTRATION
This talk will give an overview of my group's research activities on visual navigation of MAVs, such as, multi-MAV navigation and mapping, ground-aerial robot collaboration, real-time dense 3D reconstruction, and low-latency sensing. Most of the talk will be devoted to low-latency sensing: enabling vision algorithms running at several kilohertz, which could one day allow achieving aggressive maneuvers, so far demonstrated only in motion-capture systems (see Kumar at UPenn or D'Andrea at ETH). The talk will and with a live flight demo showing dense 3D reconstruction over a cluetterd scene.
- 12:30 - Lunch
- 13:30 - Image Based Visual Control of Aerial Robotic Vehicles
Robert Mahony - The Australian National University
Vision systems provide a low-cost, light-weight and information rich sensing paradigm that offers significant potential for control of aerial robotic vehicles. Processing vision data to extract state and environment information requires significant computational power. Although recent results using visual SLAM and similar techniques are impressive, there will always be a place for low level visual control of aerial vehicles using simple visual cues that can be computed by embedded vision systems. In this talk I discuss several examples of image based visual control for aerial vehicles, including the use of optic flow for a range of terrain tracking, obstacle avoidance and landing manoeuvres, as well as using more classical image features for position and path control as well as formation control.
- 14:00 - High-speed Monocular Vision for Close-quarters Flying
Inkyu Sa and Peter Corke - Queensland University of Technology
This talk will describe recent work in using high-speed monocular vision plus inertial sensors to enable close flying to solid infrastructure. We do not use GPS and consider a scenario where an unskilled remote pilot maneuvers the vehicle with respect to a local task frame.
- 14:30 - Closing the MAV navigation loop: from control to localization and path-planning
Markus Achtelik, Sammy Omari, Roland Siegwart - ETH Zurich
INCLUDES SENSOR DEMONSTRATION
With localization and control of Micro Aerial Vehicles (MAVs) reaching a certain maturity, these fields are often studied separately from path planning and generally outside the complete navigation loop. In this workshop, we present our recent achievements in control and state estimation for our rotocraft, and finally show how these can be incorporated within the planning process. In terms of position control of the MAV, we show a computationally efficient and novel approach requiring only two nested control loops, while eliminating the need for a dedicated attitude controller, commonly used in multiply-nested control structures for MAVs. This controller requires accurate knowledge of the MAV's state, which we obtain with our recent self-calibrating state estimation framework. This is able to handle multiple (potentially delayed) measurement updates, which can be either absolute or relative - the latter is usually the case in camera-based localisation. Having both visual perception and state estimation in the control loop, neither the uncertainty in the MAV's perception nor the MAV's dynamics can be ignored if a realistic path is to be computed. With our state estimation framework for MAVs depending on motion to render all states observable, the challenge lies in acquiring informative measurements, such that the MAV always remains in a fully observable mode. We show how to formulate this problem and produce realistic paths for the MAV, using the same navigation pipeline (state estimation, localization, control) within the planning phase as the one used by the MAV in real scenarios.
- 15:00 - Poster and Coffee Session
- 15:30 - Vision-Based Bilateral Shared Control and Autonomous Flight for Single and Multiple quadrotor UAVs
Paolo Robuffo Giordano - CNRS at IRISA and INRIA Rennes Bretagne Atlantique, France
This talk will cover some recent results in the field of motion control and estimation for quadrotor Unmanned Aerial Vehicles (UAVs) when relying on onboard monocular cameras as main exteroceptive sensor modality. We will first discuss an example of bilateral shared control of a group of UAVs by a human user: in this application, the onboard cameras are exploited as a source of "position feedback" for reaching and then maintaining a given UAV "bearing-formation". However, the UAV velocities over ground are still regulated by resorting to an external tracking system (Vicon). The next part of the talk will then focus on a recently proposed nonlinear estimation machinery able to cope with this issue by providing a metric body-frame velocity estimation from the perceived optical flow and the onboard IMU readings, thus allowing for a fully autonomous ("vicon-free") UAV flight in highly unstructured environments.
- 16:00 - RGB-D based Haptic Teleoperation of UAVs with Onboard Sensors: Development and Preliminary Results [PDF]
Antonio Franchi, Paolo Stegagno, Massimo Basile, Heinrich H. Buelthoff - Max Planck Institute for Biological Cybernetics
This talk will present the design of a platform for autonomous navigation of a quadrotor UAV based on RGB-D technology. The proposed platform can safely navigate in an unknown environment while self-stabilization is done relying only on its own sensor perception. An estimation system based IMU and RGB-D integration computes the velocity of the quadrotor in its body frame. Experimental tests conducted as teleoperation experiments show the effectiveness of our approach in an unstructured environment.
- 16:20 - Demo of Fotokite: the Flying Steadycam
Sergei Lupashin - ETH Zurich and University of Zurich
INCLUDES INDOOR FLIGHT DEMONSTRATION
- 16:30 - Panel Discussion
The poster sessions will be held during the coffee breaks
- Smartphone Quadrotor Flight Controllers and Algorithms for Autonomous Flight [PDF] Tyler Ryan, H. Jin Kim - Seoul National University
- Neural sensorimotor primitives for vision-controlled flying robots [PDF]
Oswald Berthold, Verena V. Hafner - Humboldt University of Berlin
- Monocular Parallel Tracking and Mapping with Odometry Fusion for MAV Navigation in Feature-lacking Environments [PDF]
Duy-Nguyen Ta, Kyel Ok, Frank Dellaert - Georgia Institute of Technology