A5061841016- Robotics and Sensor-Based Localization
- Robotic Path Planning Algorithms
- UAV Applications and Optimization
- Multimodal Machine Learning Applications
- Advanced Vision and Imaging
- Distributed Control Multi-Agent Systems
- Advanced Image and Video Retrieval Techniques
- Wireless Sensor Networks for Data Analysis
- Tactile and Sensory Interactions
- 3D Surveying and Cultural Heritage
- Neuroblastoma Research and Treatments
- Muscle activation and electromyography studies
- Gaze Tracking and Assistive Technology
- Visual perception and processing mechanisms
- Advanced Neural Network Applications
- Machine Learning and ELM
- Neural Networks and Applications
- Space Satellite Systems and Control
- Space Exploration and Technology
- Teleoperation and Haptic Systems
- Robotics and Automated Systems
- Remote Sensing and LiDAR Applications
- Domain Adaptation and Few-Shot Learning
École Polytechnique Fédérale de Lausanne
2019-2022
Institute of Electrical and Electronics Engineers
2019
Gorgias Press (United States)
2019
Vrije Universiteit Brussel
2019
South Dakota State University
2017
University of Groningen
1990
Decentralized deployment of drone swarms usually relies on inter-agent communication or visual markers that are mounted the vehicles to simplify their mutual detection. This letter proposes a vision-based detection and tracking algorithm enables groups drones navigate without markers. We employ convolutional neural network detect localize nearby agents onboard quadcopters in real-time. Rather than manually labeling dataset, we automatically annotate images train using background subtraction...
Decentralized drone swarms deployed today either rely on sharing of positions among agents or detecting swarm members with the help visual markers. This letter proposes an entirely approach to coordinate markerless based imitation learning. Each agent is controlled by a small and efficient convolutional neural network that takes raw omnidirectional images as inputs predicts 3D velocity commands match those computed flocking algorithm. We start training in simulation propose simple yet...
Mobile manipulation robots have great potential for roles in support of rescuers on disaster-response missions. Robots can operate places too dangerous humans and therefore assist accomplishing hazardous tasks while their human operators work at a safe distance. We developed system that consists the highly flexible Centauro robot suitable control interfaces, including an immersive telepresence suit support-operator controls offering different levels autonomy.
In this paper, we present a multi-sensory terrain classification algorithm with generalized representation using semantic and geometric features. We compute features from lidar point clouds extract pixel-wise labels fully convolutional network that is trained dataset strong focus on urban navigation. use data augmentation to overcome the biases of original apply transfer learning adapt model new in off-road environments. Finally, fuse visual random forest classify traversability into three...
Dynamic environments such as urban areas are still challenging for popular visual-inertial odometry (VIO) algorithms. Existing datasets typically fail to capture the dynamic nature of these environments, therefore making it difficult quantitatively evaluate robustness existing VIO methods. To address this issue, we propose three contributions: firstly, provide VIODE benchmark, a novel dataset recorded from simulated UAV that navigates in environments. The unique feature is systematic...
Vision-based drone swarms have recently emerged as a promising alternative to address the fault-tolerance and flexibility limitations of centralized communication-based aerial collective systems. Although most vision-based control algorithms rely on detection neighbors, they usually neglect critical perceptual factors such visual occlusions their effect scalability swarm. To estimate impact we propose simple but perceptually realistic neighbor selection model that discards obstructed agents....
This paper presents a data-driven approach to learning vision-based collective behavior from simple flocking algorithm. We simulate swarm of quadrotor drones and formulate the controller as regression problem in which we generate 3D velocity commands directly raw camera images. The dataset is created by simultaneously acquiring omnidirectional images computing corresponding control command show that convolutional neural network trained on visual inputs drone can learn not only robust...
People learn motor activities best when they are conscious of their errors and make a concerted effort to correct them. While haptic interfaces can facilitate training, existing often bulky do not always ensure post‐training skill retention. Herein, programmable sleeve composed textile‐based electroadhesive clutches for acquisition retention is described. Its functionality in learning study where users control drone's movement using elbow joint rotation shown. Haptic feedback used restrain...
Decentralized deployment of drone swarms usually relies on inter-agent communication or visual markers that are mounted the vehicles to simplify their mutual detection. This letter proposes a vision-based detection and tracking algorithm enables groups drones navigate without markers. We employ convolutional neural network detect localize nearby agents onboard quadcopters in real-time. Rather than manually labeling dataset, we automatically annotate images train using background subtraction...
This special issue presents cutting-edge humanoid robot applications in real-world scenarios to solve problems.
Mobile manipulation robots have high potential to support rescue forces in disaster-response missions. Despite the difficulties imposed by real-world scenarios, are promising perform mission tasks from a safe distance. In CENTAURO project, we developed system which consists of highly flexible Centauro robot and suitable control interfaces including an immersive tele-presence suit support-operator controls on different levels autonomy. this article, give overview final system. particular,...
Decentralized drone swarms deployed today either rely on sharing of positions among agents or detecting swarm members with the help visual markers. This work proposes an entirely approach to coordinate markerless based imitation learning. Each agent is controlled by a small and efficient convolutional neural network that takes raw omnidirectional images as inputs predicts 3D velocity commands match those computed flocking algorithm. We start training in simulation propose simple yet...