A5029413988- Target Tracking and Data Fusion in Sensor Networks
- Particle accelerators and beam dynamics
- Magnetic confinement fusion research
- Gaussian Processes and Bayesian Inference
- Social and Educational Sciences
- Distributed Sensor Networks and Detection Algorithms
- Superconducting Materials and Applications
- Indoor and Outdoor Localization Technologies
- Autonomous Vehicle Technology and Safety
- Advanced Neural Network Applications
- Robotics and Sensor-Based Localization
- Anomaly Detection Techniques and Applications
- Fusion materials and technologies
- Underwater Acoustics Research
- Bayesian Methods and Mixture Models
- Plasma Diagnostics and Applications
- Fault Detection and Control Systems
- Education and Critical Thinking Development
- Domain Adaptation and Few-Shot Learning
- Millimeter-Wave Propagation and Modeling
- Remote Sensing and LiDAR Applications
- Evaluation of Teaching Practices
- Infrared Target Detection Methodologies
- Second Language Learning and Teaching
- Multimodal Machine Learning Applications
Chalmers University of Technology
2015-2024
Göteborgs Stads
2019-2024
Volvo (Sweden)
2024
Linköping University
2003-2023
ITER
2010-2020
Lund University
2001-2020
Commonwealth Scientific and Industrial Research Organisation
2019
University of Gothenburg
2019
University of Liverpool
2019
Autoliv (Sweden)
2017
Abstract In this article the concepts of research tradition, programme, tool and orientation are used to clarify character phenomenography. Phenomenography is said be fundamentally a characterised by delimitation an aim in relation kind object. The describe object conception. Phenomenographic also has common characteristics method general related these called approach. approach together represent specialisation. historical roots ontological, epistemological methodological assumptions...
The state of the art in semantic segmentation is steadily increasing performance, resulting more precise and reliable segmentations many different applications. However, progress limited by cost generating labels for training, which sometimes requires hours manual labor a single image. Because this, semi-supervised methods have been applied to this task, with varying degrees success. A key challenge that common augmentations used classification are less effective segmentation. We propose...
Semantic segmentation models based on convolutional neural networks have recently displayed remarkable performance for a multitude of applications. However, these typically do not generalize well when applied new domains, especially going from synthetic to real data. In this paper we address the problem unsupervised do-main adaptation (UDA), which attempts train labelled data one domain (source domain), and simultaneously learn unlabelled in interest (target domain). Existing methods seen...
This paper presents the generalized optimal sub-pattern assignment (GOSPA) metric on space of finite sets targets. Compared to well-established (OSPA) metric, GOSPA is unnormalized as a function cardinality and it penalizes errors differently, which enables us express an optimisation over assignments instead permutations. An important consequence this that allows penalize localization for detected targets due missed false targets, indicated by traditional multiple target tracking (MTT)...
In this work, a deep learning approach has been developed to carry out road detection using only LIDAR data. Starting from an unstructured point cloud, top-view images encoding several basic statistics such as mean elevation and density are generated. By considering representation, is reduced single-scale problem that can be addressed with simple fast fully convolutional neural network (FCN). The FCN specifically designed for the task of pixel-wise semantic segmentation by combining large...
In this paper, we present the Cooperative Adaptive Cruise Control (CACC) architecture, which was proposed and implemented by team from Chalmers University of Technology, Göteborg, Sweden, that joined Grand Driving Challenge (GCDC) in 2011. The CACC architecture consists following three main components, are described detail: 1) communication; 2) sensor fusion; 3) control. Both simulation experimental results provided, demonstrating system can drive within a vehicle platoon while minimizing...
A model for workplace learning is presented, which intends to integrate formal and informal with the use of e‐learning. An important underlying assumption that integration necessary in order create desirable competencies, from both an individual organisational perspective. Two case studies are presented was tested. One carried out industrial setting, other a hospital context. The results promising terms flexibility accessibility, but some problems remain be solved. These have do learning,...
The ITER Neutral Beam Test Facility (NBTF), called PRIMA (Padova Research on Megavolt Accelerator), is hosted in Padova, Italy and includes two experiments: MITICA, the full-scale prototype of heating neutral beam injector, SPIDER, full-size radio frequency negative-ions source. NBTF realization exploitation SPIDER MITICA have been recognized as necessary to make future operation injectors efficient reliable, fundamental achievement thermonuclear-relevant plasma parameters ITER. This paper...
This paper is concerned with Gaussian approximations to the posterior probability density function (PDF) in update step of Bayesian filtering nonlinear measurements. In this setting, sigma-point Kalman filter (KF) recursion are widely used due their ease implementation and relatively good performance. step, these KFs equivalent linearizing measurement by statistical linear regression (SLR) respect prior PDF. paper, we argue that should be linearized using SLR rather than take into account...
The ITER project requires additional heating by two neutral beam injectors, each accelerating to 1 MV a 40 A of negative deuterium ions, deliver the plasma power about 17 MW for one hour. As these requirements have never been experimentally met, it was recognized as necessary setup test facility, PRIMA (Padova Research on Megavolt Accelerator), in Italy, including full-size ion source, SPIDER, and prototype whole injector, MITICA, aiming develop injectors be installed ITER. This realization...
We provide a derivation of the Poisson multi-Bernoulli mixture (PMBM) filter for multi-target tracking with standard point target measurements without using probability generating functionals or functional derivatives. also establish connection \delta-generalised labelled (\delta-GLMB) filter, showing that \delta-GLMB density represents targets so it can be seen as special case PMBM. In addition, we propose an implementation linear/Gaussian dynamic and measurement models how to efficiently...
In this paper, we propose a metric on the space of finite sets trajectories for assessing multi-target tracking algorithms in mathematically sound way. The main use is to compare estimates from different with ground truth trajectories. proposed includes intuitive costs associated localization error properly detected targets, missed and false targets track switches at each time step. computation based solving multi-dimensional assignment problem. We also lower bound metric, which computable...
Masked autoencoding has become a successful pretraining paradigm for Transformer models text, images, and, recently, point clouds. Raw automotive datasets are suitable candidates self-supervised pre-training as they gener-ally cheap to collect compared annotations tasks like 3D object detection (OD). However, the development of masked autoencoders clouds focused solely on synthetic and indoor data. Consequently, existing meth-ods have tailored their representations toward small dense with...
In this article we show that when targets are closely spaced, traditional tracking algorithms can be adjusted to perform better under a performance measure disregards identity.More specifically, propose an version of the Joint Probabilistic Data Association (JPDA) filter, which call Set JPDA (SJPDA) filter.Through examples and theory motivate new approach, its possibilities.To decrease computational requirements, further SJPDA filter formulated as continuous optimization problem is fairly...
Environment perception is a key enabling technology in autonomous vehicles, and multiple object tracking an important part of this. The use high resolution sensors, such as automotive radar lidar, leads to the extended problem, with detections per tracked object. For computationally feasible tracking, data association problem must be handled. Previous work has relied on two-step approach, using clustering algorithms, together assignment achieve In this paper, we show that it possible handle...
This paper presents the probability hypothesis density filter (PHD) and cardinality PHD (CPHD) for sets of trajectories, which are referred to as trajectory (TPHD) CPHD (TCPHD) filters. Contrary PHD/CPHD filters, TPHD/TCPHD filters able produce estimates from first principles. The TPHD is derived by recursively obtaining best Poisson multitrajectory approximation posterior over alive trajectories minimising Kullback-Leibler divergence. TCPHD in same way but propagating an independent...
This paper presents two trajectory Poisson multi-Bernoulli (TPMB) filters for multi-target tracking: one to estimate the set of alive trajectories at each time step and another all trajectories, which includes dead step. The are based on propagating a (PMB) density corresponding through filtering recursion. After update step, posterior is PMB mixture (PMBM) so, in order obtain density, Kullback-Leibler divergence minimisation an augmented space performed. developed computationally lighter...
Heating neutral beam (HNB) injectors, necessary to achieve burning conditions and control plasma instabilities in ITER, are characterized by such demanding parameters that a test facility (NBTF) dedicated their development optimization is being realized Padua (Italy) with direct contributions from the Italian government (through Consorzio RFX as host entity) ITER international organization (with kind domestic agencies of Europe, Japan India) technical scientific support various European...
We propose a solution of the multiple target tracking (MTT) problem based on sets trajectories and random finite set framework. A full Bayesian approach to MTT should characterise distribution given measurements, as it contains all information about trajectories. attain this by considering multi-object density functions in which objects are For standard models, we also describe conjugate family multitrajectory functions.
5G communication systems operating above 24 GHz have promising properties for user localization and environment mapping. Existing studies either relied on simplified abstract models of the signal propagation measurements, or are based direct positioning approaches, which directly map received waveform to a position. In this study, we consider an intermediate approach, consists four phases—downlink data transmission, multi-dimensional channel estimation, parameter clustering, simultaneous...