Papers by Christophe Wagter
Controlling over-actuated Unmanned Aerial Vehicle (UAV) is an important task to achieve reliable ... more Controlling over-actuated Unmanned Aerial Vehicle (UAV) is an important task to achieve reliable fail-safe autonomous flight. Incremental Non-linear Control Allocation or INCA has been proposed to solve the platform’s control allocation problem by minimizing a set of objective functions with a method known as the Active Set Method. This work proposes an extension to INCA to control the outer loop of a quadplane UAV, an in-plane combination between a quadrotor and a conventional fixed-wing. The novel controller is called Extended INCA or XINCA and optimizes a mix of physical actuator commands and angular control setpoints fed to the vehicle’s inner loop. It does so while adapting to varying flight phases, conditions, and vehicle states, and taking into account the aerodynamic properties of the main wing. XINCA has low dependence on accurate vehicle models and requires only several optimization parameters. Flight simulations and experimental flights are performed to prove the performa...
Monocular optical flow has been widely used to detect obstacles in Micro Air Vehicles (MAVs) duri... more Monocular optical flow has been widely used to detect obstacles in Micro Air Vehicles (MAVs) during visual navigation. However, this approach requires significant movement, which reduces the efficiency of navigation and may even introduce risks in narrow spaces. In this paper, we introduce a novel setup of self-supervised learning (SSL), in which optical flow cues serve as a scaffold to learn the visual appearance of obstacles in the environment. We apply it to a landing task, in which initially 'surface roughness' is estimated from the optical flow field in order to detect obstacles. Subsequently, a linear regression function is learned that maps appearance features represented by texton distributions to the roughness estimate. After learning, the MAV can detect obstacles by just analyzing a still image. This allows the MAV to search for a landing spot without moving. We first demonstrate this principle to work with offline tests involving images captured from an on-board c...
Flying Into the Wind: Insects and Bio-Inspired Micro-Air-Vehicles With a Wing-Stroke Dihedral Steer Passively Into Wind-Gusts
Frontiers in Robotics and AI, 2022
Natural fliers utilize passive and active flight control strategies to cope with windy conditions... more Natural fliers utilize passive and active flight control strategies to cope with windy conditions. This capability makes them incredibly agile and resistant to wind gusts. Here, we study how insects achieve this, by combining Computational Fluid Dynamics (CFD) analyses of flying fruit flies with freely-flying robotic experiments. The CFD analysis shows that flying flies are partly passively stable in side-wind conditions due to their dorsal-ventral wing-beat asymmetry defined as wing-stroke dihedral. Our robotic experiments confirm that this mechanism also stabilizes free-moving flapping robots with similar asymmetric dihedral wing-beats. This shows that both animals and robots with asymmetric wing-beats are dynamically stable in sideways wind gusts. Based on these results, we developed an improved model for the aerodynamic yaw and roll torques caused by the coupling between lateral motion and the stroke dihedral. The yaw coupling passively steers an asymmetric flapping flyer into t...
Autonomous flight of pocket drones is challenging due to the severe limitations on on-board energ... more Autonomous flight of pocket drones is challenging due to the severe limitations on on-board energy, sensing, and processing power. However, tiny drones have great potential as their small size allows maneuvering through narrow spaces while their small weight provides significant safety advantages. This paper presents a computationally efficient algorithm for determining optical flow, which can be run on an STM32F4 microprocessor (168 MHz) of a 4 gram stereo-camera. The optical flow algorithm is based on edge histograms. We propose a matching scheme to determine local optical flow. Moreover, the method allows for sub-pixel flow determination based on time horizon adaptation. We demonstrate velocity measurements in flight and use it within a velocity control-loop on a pocket drone.
Replication Data for: "Enhancing optical flow-based control by learning visual appearance cues for flying robots
This repository contains all data and code necessary to reproduce the experiments and figures in ... more This repository contains all data and code necessary to reproduce the experiments and figures in the article: "Enhancing optical flow-based control by learning visual appearance cues for flying robots". It allows to reproduce both the experiments in simulation and the real-world experiments with the Parrot Bebop 2 drone. Please see the README in the repository for a detailed explanation. Please note that the Paparazzi code included in this data set is subject to a GNU left license. See https://github.com/paparazzi/paparazzi/blob/master/LICENSE for more details.
International Journal of Micro Air Vehicles, 2015
Flapping Wing Micro Air Vehicles (FWMAVs) hold the potential to both cover large distances and pe... more Flapping Wing Micro Air Vehicles (FWMAVs) hold the potential to both cover large distances and perform precision flights when arrived at destination. However, flying at different speeds leads to a complex control problem for attitude stabilization. Inspired by nature, we present a morphing mechanism that allows tailed FW- MAVs to have a passively stabilized attitude both in fast forward flight and in slow hovering flight. The mechanism displaces the wings and hence aerodynamic center. It is implemented on the DelFly II and tested in-flight in a motion tracking arena. The experimental tests show that the morphing mechanism indeed allows to fly passively stable in multiple flight modes. Just changing the aerodynamic center allows the DelFly II to fly fast forward (∼ 6 m/s, pitch attitude of 10°), transition to slow forward flight (∼ 0.8 m/s, pitch attitude of 55°), and back. The proposed mechanism paves the way for FWMAVs performing long range missions such as search-and-rescue.
The concept of an aircraft capable of both hover as well as fast forward flight (hybrid) has re-c... more The concept of an aircraft capable of both hover as well as fast forward flight (hybrid) has re-cently been implemented on unmanned aerial vehicles (UAV). Hybrid UAVs combine hover capability with long range and endurance. As UAVs are often required to operate without hu-man intervention, there is a call for autonomous guidance of hybrid UAVs. Because the dynam-ics in the transition region from hover to for-ward flight are not well known, this research focusses on the development of a longitudinal model for a hybrid UAV based on test flight data. The same approach can be used for other types of airframes as well, allowing cheap and easy modelling. Sensors were logged for nine dif-ferent flights and a Kalman filter was used for state estimation. The system was excited by dou-blet inputs on the commanded pitch and thrust. From the input-output response a piecewise lin-ear model was estimated. This model was ver-ified by comparing the measured doublet input response to the simulated re...
Journal of Guidance, Control, and Dynamics, 2012
a Exteroceptive sensors observe entities external to the spacecraft, while proprioceptive sensors... more a Exteroceptive sensors observe entities external to the spacecraft, while proprioceptive sensors measure quantities within the spacecraft "body".
A challenging problem in the research field of Micro Air Vehicles is to achieve vision-based auto... more A challenging problem in the research field of Micro Air Vehicles is to achieve vision-based autonomous indoor flight. Approaches to this problem currently hardly make use of image appearance features, because these features generally are computationally expensive. In this article, we demonstrate that the broadly applicable strategy of random sampling can render the extraction of appearance features computationally efficient enough for use in autonomous flight. Random sampling is applied to a height control algorithm that estimates the height at which an image is taken by processing small image patches. The patches are extracted at random locations in the image. We vary the specific number of image patches to directly influence the trade-off between processing time and the accuracy of the height estimation. The algorithm is first tested on image sets and then on videos taken from a real platform. Subsequently, the algorithm is tested on a 15-gram ornithopter in an office room. The experiments show that very few image patches ( 0.56% of all possible patches) are already sufficient for the task of height control.
International Journal of Micro Air Vehicles, 2022
Wireless ranging measurements have been proposed for enabling multiple Micro Air Vehicles (MAVs) ... more Wireless ranging measurements have been proposed for enabling multiple Micro Air Vehicles (MAVs) to localize with respect to each other. However, the high-dimensional relative states are weakly observable due to the scalar distance measurement. Hence, the MAVs have degraded relative localization and control performance under unobservable conditions as can be deduced by the Lie derivatives. This paper presents a nonlinear model predictive control (NMPC) by maximizing the determinant of the observability matrix to generate optimal control inputs, which also satisfy constraints including multi-robot tasks, input limitation, and state bounds. Simulation results validate the localization and control efficacy of the proposed MPC method for range-based multi-MAV systems with weak observability, which has faster convergence time and more accurate localization compared to previously proposed random motions. A real-world experiment on two Crazyflies indicates the optimal states and control be...
3 Lockheed Martin Assistant Professor of Avionics Integration
Building aircraft with navigation and control systems that can complete flight tasks is complex, ... more Building aircraft with navigation and control systems that can complete flight tasks is complex, and often involves integrating information from multiple sensors to estimate the state of the vehicle. This paper describes a method, in which a glider can fly from a starting point to a predetermined end location (target) precisely using vision only. Using vision to control an aircraft represents a unique challenge, partially due to the high rate of images required in order to maintain tracking and to keep the glider on target in a moving air mass. Second, absolute distance and angle measurements to the target are not readily available when the glider does not have independent measurements of its own position. The method presented here uses an integral image representation of the video input for the analysis. The integral image, which is obtained by integrating the pixel intensities across the image, is reduced to a probable target location by performing a cascade of feature matching fu...
Hybrid UAVs with hovering as well as fast forward flight capab ility or enhanced maneuverability ... more Hybrid UAVs with hovering as well as fast forward flight capab ility or enhanced maneuverability are expected to become increasin gly important. To approach the complex problem of autonomous flight in the full fli ght envelope of these transitioning or reconfiguring vehicles, a simple but power ful approach is presented. A traditional rotorcraft control strategy consisting of an ttitude innerloop and position outerloop is enhanced with a lift allocation control ler in between. By running several sub-controllers per lift-device, simplicity is ke pt while allowing sustained flight at any transitioning percentage for any number of lift ing devices. The applications of this approach range from hover of fixedwings, or al l wing easier fast forward flight of conventional rotorcraft to autonomous flight o f most types of hybrid or reconfiguring UAVs. Flight test results are presented usi ng the ATMOS hybrid UAV. Christophe De Wagter Micro Aerial Vehicle Lab, Faculty of Aerospace Engineerin...
The Delfly II Flapping Wing Micro Air Vehicle was flown in an external tracking chamber. It was p... more The Delfly II Flapping Wing Micro Air Vehicle was flown in an external tracking chamber. It was possible to perform controlled flight-test maneuvers with an ornithopter that is capable of hover and forward flight, for system identification purposes. This was achieved by programming its autopilot to deflect the a control surface, while keeping the other surfaces at trimmed condition. Step, doublet and triplet inputs of 1/3, 2/3 and 4/3 of a second on the elevator, rudder and flapping frequency actuators were performed to excite the Delfly's eigenmodes. These tests were carried out at different flight speeds, ranging from-0.5 to 8 m/s and with the ornithopter's center of gravity at 83%, 74%, 44% and 42% of the wing root chord. As a result, it was possible to cover the Delfly's flight envelope and collect data that will be used to build a dynamic and aerodynamic model of the Delfly. The selected inputs have shown to excite the Delfly in dampened oscillatory modes that can be compared to phugoid and short period for the longitudinal dynamics. The Delfly is highly affected by the rudder deflections. The results also reveal an unstable lateral mode similar to a spiral.
As autonomous GPS waypoint tracking has been achieved for 50 cm 500 gram MAVs, this paper focuses... more As autonomous GPS waypoint tracking has been achieved for 50 cm 500 gram MAVs, this paper focuses on autonomous interaction with the world and higher level interaction with the operators. Image Analysis is performed on down-streamed images in order to es-timate altitude above ground, map the surroundings and detect targets of interest. Detected targets are collected, ordered by certainty and pre-sented to the user for validation. Furthermore information from all levels of sophistication ranging from flight level to mission level are merged in a single intelligence map allowing the operator to chose the level of inter-action. The resulting system enables extremely easy task definitions such as search this area for, determine the location of this visual target very pre-cisely identify target at this location. With no user intervention other than the launch of the vehicle this mission can be performed. However during tasks where the MAV has less certainty such as for instance during se...
Robotics and Autonomous Systems
Hear-and-Avoid: Acoustic Detection of General Aviation Aircraft for UAV
This dataset consists of audio recordings of overflying general aviation propeller aircraft which... more This dataset consists of audio recordings of overflying general aviation propeller aircraft which overfly the microphone array. The data is labeled to identify where the plane overflies the microphone.
This is the first feasibility study of hear-andavoid on Micro Air Vehicles with acoustic vector s... more This is the first feasibility study of hear-andavoid on Micro Air Vehicles with acoustic vector sensors. The Microflown MEMS technology based sensor is used for this purpose. The threedimensional acoustic vector sensor consists of three orthogonally placed particle velocity sensors and one sound pressure microphone. The usage of this sensor is explored for detecting the direction to other sound sources in the sky. The experiments involve adapting and mounting the MEMS sensor on a fixed wing MAV. The empirical results show that loud sounds as could be produced by civil aircrafts are clearly detectable by the sensor. In addition, they indicate that the detection of other MAVs is possible. In general, the long-range detection properties and the small weight of the sensors hold an important promise for enhancing the sense-and-avoid capabilities of MAVs.
Flapping Wing Micro Air Vehicles (FWMAVs) hold the potential to both cover large distances and pe... more Flapping Wing Micro Air Vehicles (FWMAVs) hold the potential to both cover large distances and perform precision flights when arrived at destination. However, flying at different speeds leads to a complex control problem for attitude stabilization. Inspired by nature, we present a morphing mechanism that allows tailed FW- MAVs to have a passively stabilized attitude both in fast forward flight and in slow hovering flight. The mechanism displaces the wings and hence aerodynamic center. It is implemented on the DelFly II and tested in-flight in a motion tracking arena. The experimental tests show that the morphing mechanism indeed allows to fly passively stable in multiple flight modes. Just changing the aerodynamic center allows the DelFly II to fly fast forward (~ 6 m/s, pitch attitude of 10˚), transition to slow forward flight (~ 0.8 m/s, pitch attitude of 55˚), and back. The proposed mechanism paves the way for FWMAVs performing long range missions such as search-and-rescue.
Document Analysis Systems, 2003
Building aircraft with navigation and control systems that can complete flight tasks is complex, ... more Building aircraft with navigation and control systems that can complete flight tasks is complex, and often involves integrating information from multiple sensors to estimate the state of the vehicle. This paper describes a method, in which a glider can fly from a starting point to a predetermined and location (target) precisely using vision only. Using vision to control an aircraft
ArXiv, 2021
Relative localization is an important ability for multiple robots to perform cooperative tasks. T... more Relative localization is an important ability for multiple robots to perform cooperative tasks. This paper presents a deep neural network (DNN) for monocular relative localization between multiple tiny flying robots. This approach does not require any ground-truth data from external systems or manual labelling. Our system is able to label real-world images with 3D relative positions between robots by another onboard relative estimation technology. After the training from scratch in this self-supervised way, the DNN can predict the relative positions of peer robots by purely using the monocular image. This deep-learning based visual relative localization is scalable, distributed and autonomous. Simulation shows the pipeline for synthetic image generation for multiple robots with Blender and 3D rendering, which allows for preliminary validation of the designed network. Experiments are conducted on two Crazyflie quadrotors for dataset collection with random attitude and velocity. Train...
Uploads
Papers by Christophe Wagter