A5031768178- Robotics and Sensor-Based Localization
- Advanced Vision and Imaging
- Robotic Locomotion and Control
- Advanced Image and Video Retrieval Techniques
- Optical measurement and interference techniques
- Robotic Path Planning Algorithms
- Water Quality Monitoring Technologies
- Image Processing Techniques and Applications
- Image Enhancement Techniques
- Fish biology, ecology, and behavior
- Prosthetics and Rehabilitation Robotics
- Robot Manipulation and Learning
- Underwater Vehicles and Communication Systems
- Image and Signal Denoising Methods
- Ichthyology and Marine Biology
- 3D Surveying and Cultural Heritage
- Bat Biology and Ecology Studies
- Image and Object Detection Techniques
- Target Tracking and Data Fusion in Sensor Networks
- Gait Recognition and Analysis
- Maritime and Coastal Archaeology
- Medical Image Segmentation Techniques
- Video Surveillance and Tracking Methods
- Distributed Control Multi-Agent Systems
- Visual Attention and Saliency Detection
Université de Haute-Alsace
2019-2024
École nationale supérieure de techniques avancées Bretagne
2023
Centre de Recherche en Informatique
2018-2023
Institut Icare
2019-2020
École Nationale Supérieure des Mines de Paris
2017
Laboratoire des Sciences du Numérique de Nantes
2017
Institut de Recherches en Technologies et Sciences pour le Vivant
2017
IMT Atlantique
2017
Nantes Université
2017
Institut Mines-Télécom
2017
Reliable state estimation is crucial for stable planning and control of legged locomotion. A fundamental component a estimator in platforms Leg Odometry, which only requires information about kinematics contacts. Many robots use dedicated sensors on each foot to detect ground However, this choice impractical many agile field operations, as these often degrade break. Instead, paper focuses the development robust Odometry module, does not require contact sensors. The module estimates...
We present a framework for quadrupedal locomotion over highly challenging terrain where the choice of appropriate footholds is crucial success behaviour. use path planning approach which shares many similarities with results DARPA Learning Locomotion challenge and extend it to allow more flexibility increased robustness. During execution we incorporate an on-line force-based foothold adaptation mechanism that updates planned motion according perceived state environment. This way exploit...
This paper presents a framework developed to increase the autonomy and versatility of large (~75kg) hydraulically actuated quadrupedal robot. It combines onboard perception with two locomotion strategies, dynamic trot static crawl gait. way robot can perceive its environment arbitrate between behaviours according situation at hand. All computations are performed on-board carried out in separate computers, one handles high-level processes while other is concerned low-level hard real-time...
We present a mobile robot whose goal is to autonomously explore an unknown indoor environment and build semantic map containing high-level information similar those extracted by humans. This includes the rooms, their connectivity, objects they contain material of walls ground. was developed in order participate French exploration mapping contest called CAROTTE produce easily interpretable maps environment. In particular we our object detection approach based on color+depth camera that fuse...
The human perception of the external world appears as a natural, immediate and effortless task. It is achieved through number "low-level" sensory-motor processes that provide high-level representation adapted to complex reasoning decision. Compared these representations, mobile robots usually only low-level obstacle maps lack such highlevel information. We present robot whose goal autonomously explore an unknown indoor environment build semantic map containing information similar those...
We address the problem of simultaneous localization and mapping by combining visual loop-closure detection with metrical information given robot odometry. The proposed algorithm builds in real-time topo-metric maps an unknown environment, a monocular or omnidirectional camera odometry gathered motors encoders. A dedicated improved version our previous work on purely appearance-based [1] is used to extract potential locations. Potential locations are then verified classified using new...
Legged robots have the potential to navigate in more challenging terrain than wheeled do. Unfortunately, their control is difficult because they deal with traditional mapping and path planning problems, as well foothold computation, leg trajectories posture order achieve successful navigation. Many parameters need be adjusted real time keep robot stable safe while it moving. In this paper, we will present a new framework for quadruped robot, which performs goal-oriented navigation on unknown...
Recently, biologists have shown that the weakly electric fish are able to estimate nature, localization and 3D geometric properties of an object using active sense. Incredibly, Gnathonemus petersii performs this task in dark only by moving towards around object, its vision is not required. In paper, we proposed address challenging issue shape estimation a real underwater robot equipped with artificial To end, used corrected version dipolar tensor dedicated small objects [Ammari et al., 2014]...
We present a real-time SLAM system that combines an improved version of the Iterative Closest Point (ICP) and inertial dead reckoning to localize our dynamic quadrupedal machine in local map. Despite strong fast motions induced by 80 kg hydraulic legged robot, is robust enough keep position error below 5% within map surrounds robot. The 3D terrain, computed at camera frame rate suitable for vision based planned locomotion. measurements are used before after ICP registration, provide good...
In this article we address the issue of shape estimation using electric sense inspired by active fish. These fish can perceive their environment measuring perturbations in a self-generated field caused nearby objects. The approach proceeded three stages. Firstly, object was detected and its properties (insulator or conductor) identified. Secondly, localized multiple signal classification algorithm, which originally developed to localize radio wave emitter network antennas. Thirdly, relied on...
We address the problem of simultaneous localization and mapping (SLAM) by combining visual loop-closure detection with metrical information given a robot odometry. The proposed algorithm extends purely appearance-based method based on bags words [A. Angeli, D. Filliat, S. Doncieux J.-A. Meyer, IEEE Transactions On Robotics, Special Issue Visual SLAM 24 (2008) 1027–1037], which is able to detect when has returned back previously visited place. An efficient optimization used integrate odometry...
Agile robot locomotion on rough terrain is highly dependent the ability to perceive environment. In this paper, we show how interaction between a reactive control framework and an online mapping system can significantly improve trotting performance irregular terrain. particular, new controller increases stability of reduces frontal leg shin collisions with obstacles by correcting in realtime foothold locations. The uses RGB-D sensor motion capture build three dimensional map surroundings...
In the 50s, biologists discovered that some electric hsh is capable of discriminating pose as well and geometric properties surrounding objects by navigating measuring distortions a self-generated held. this article, we address challenging issue ellipsoidal size estimation for underwater robots equipped with artihcial sense. Unlike current methods, approach can estimate both position in parallel single straight trajectory. Neither multipolarization nor reactive self-alignment control are...
In computer vision, the epipolar geometry embeds geometrical relationship between two views of a scene. This is degenerated for planar scenes as they do not provide enough constraints to estimate it without ambiguity. Nearly can necessary resolve But classic estimators such 5-point or 8-point algorithm combined with random sampling strategy are likely fail in this case because large part scene and requires lots trials get nondegenerated sample. However, be associated homographic model...
SUMMARY Legged robots have the potential to navigate in more challenging terrains than wheeled robots. Unfortunately, their control is demanding, because they deal with common tasks of mapping and path planning as well specific issues legged locomotion, like balancing foothold planning. In this paper, we present integration development a stabilized vision system on fully torque-controlled hydraulically actuated quadruped robot (HyQ). The active head added onto composed fast pan tilt unit...