Fakulta elektrotechnická ČVUT uspěla v hodnocení vybraných výsledků modulu M1 metodiky hodnocení výzkumných organizací M17+ v segmentu vysokých škol
Aktualizace 18.10.2024
Hodnocení výzkumných organizací (VO) v segmentu vysokých škol (VŠ) podle metodiky M17+ provádí MŠMT na základě hodnocení v modulech M1 (vybrané výsledky), M2 (bibliometrie) a zprávy vypracované mezinárodním evaluačním panelem posuzující VŠ ve třech dalších modulech M3 Společenská relevance, M4 Viabilita a M5 Strategie a koncepce. Výsledky v M1 posuzuje odborný panel z hlediska jejich kvality, originality a významnosti ve srovnání s mezinárodní úrovní. Naše fakulta svými výsledky v Hodnocení 2023 (H23) modulu M1 opět obhájila přední pozici nejen mezi fakultami ČVUT, ale i mezi konkurenčními elektrotechnickými fakultami dalších univerzit v ČR.
Počty známek na škále 1 až 5 udělené RVVI fakultním výsledků z období let 2018 - 2022 vybraným na základě jejich prokázaných dopadů jsou 3-7-5-1-0, přičemž 15 z našich 16 navržených výsledků dosahuje dle definice M17+ z hlediska originality, významu a obtížnosti získání výsledku mezinárodně uznávané kvality. Jednotlivé známky představují: “1” špičkovou světovou úroveň (world-leading), “2” vynikající mezinárodní úroveň, “3” mezinárodně uznávanou úroveň, “4” národně uznatelnou úroveň, “5” nesplňuje národně uznatelný standard.
Fakulta děkuje všem autorům, kteří své výsledky každoročně nominují do sběru kvalitních výsledků pro modul M1. Výběrová komise pracuje dvoukolově, přičemž zdůvodnění kvality a dopadu výsledků se v procesu výběru precizuje. Výsledky se známkou “1” a “2”jsou národně vnímány jako prestižní a ve fakultních kritériích hodnocení vědecko-výzkumné činnosti jsou bodovány (oddíl Q - vybrané výsledky).
Představujeme vám výsledky ohodnocené “jedničkou” a “dvojkou”. Celkové Hodnocení 2023 naleznete zde.
Jedničku získali:
prof. Petr Hájek: Hilbert generated Banach spaces need not have a norming Markushevich basis
We construct a zero dimensional uniform Eberlein compact space K such that C(K) is a Hilbert generated Banach space which does not admit any norming Markushevich basis. This solves a classical problem from the seventies.
prof. Zdeněk Bečvář, Ing. Pavel Mach, PhD.: Hierarchical resource scheduling method of wireless communication system
A resource scheduling method of a wireless communication system is provided. The resource scheduling method includes the following steps. Each of the user equipment (UEs) is classified by a centralized scheduler as a cell-edge UE or a non cell-edge UE. A first scheduling is performed by the centralized scheduler by allocating a first resource for the cell-edge UEs, a second resource for the non cell-edge UEs, and a third resource for retransmission of at least one of the cell-edge UEs. A second scheduling is performed by a distributed scheduler by allocating a first part of the second resource for at least one of the non cell-edge UEs.
prof. Tomáš Polcar, Menghzhou Liao, PhD., Paolo Nicolini, PhD., Victor Claernout, PhD.: UItra-low friction and edge-pinning effect in large-lattice-mismatch van der Waals heterostructures
Two-dimensional heterostructures are excellent platforms to realize twist-angle-independent ultra-low friction due to their weak interlayer van der Waals interactions and natural lattice mismatch. However, for finite-size interfaces, the effect of domain edges on the friction process remains unclear. Here we report the superlubricity phenomenon and the edge-pinning effect at MoS2/graphite and MoS2/hexagonal boron nitride van der Waals heterostructure interfaces. We found that the friction coefficients of these heterostructures are below 10−6. Molecular dynamics simulations corroborate the experiments, which highlights the contribution of edges and interface steps to friction forces. Our experiments and simulations provide more information on the sliding mechanism of finite low-dimensional structures, which is vital to understand the friction process of laminar solid lubricants.
Dvojku obdrželi:
doc. Jiří Bittner, ing Daniel Meister, PhD.: Parallel Locally-Ordered Clustering for Bounding Volume Hierarchy Construction
We propose a novel massively parallel construction algorithm for Bounding Volume Hierarchies (BVHs) based on locally-ordered agglomerative clustering. Our method builds the BVH iteratively from bottom to top by merging a batch of cluster pairs in each iteration. To efficiently find the neighboring clusters, we keep the clusters ordered along the Morton curve. This ordering allows us to identify approximate nearest neighbors very efficiently and in parallel. We implemented our algorithm in CUDA and evaluated it in the context of GPU ray tracing. For complex scenes, our method achieves up to a twofold reduction of build times while providing up to 17 percent faster trace times compared with the state-of-the-art methods.
prof. Jiří Matas, Dmytro Mishkin, PhD., Filip Radenovic, PhD.: Repeatability Is Not Enough: Learning Affine Regions via Discriminability
A method for learning local affine-covariant regions is presented. We show that maximizing geometric repeatability does not lead to local regions, a.k.a features, that are reliably matched and this necessitates descriptor-based learning. We explore factors that influence such learning and registration: the loss function, descriptor type, geometric parametrization and the trade-off between matchability and geometric accuracy and propose a novel hard negative-constant loss function for learning of affine regions. The affine shape estimator – AffNet – trained with the hard negative-constant loss outperforms the state-of-the-art in bag-of-words image retrieval and wide baseline stereo. The proposed training process does not require precisely geometrically aligned patches. The source codes and trained weights are available at https://github.com/ducha-aiki/affnet.
prof. Jiří Matas, Daniel Baráth, PhD.: Graph-Cut RANSAC
A novel method for robust estimation, called Graph-Cut RANSAC1, GC-RANSAC in short, is introduced. To separate inliers and outliers, it runs the graph-cut algorithm in the local optimization (LO) step which is applied when a so-far-the-best model is found. The proposed LO step is conceptually simple, easy to implement, globally optimal and efficient. GC-RANSAC is shown experimentally, both on synthesized tests and real image pairs, to be more geometrically accurate than state-of-the-art methods on a range of problems, e.g. line fitting, homography, affine transformation, fundamental and essential matrix estimation. It runs in real-time for many problems at a speed approximately equal to that of the less accurate alternatives (in milliseconds on standard CPU).
ing. Tomáš Báča, PhD., RNDr. Petr Štěpán, PhD., ing. Vojtěch Spurný, PhD., ing. Daniel Heřt, ing. Robert Pěnička, PhD., doc.Martin Saska: Autonomous landing on a moving vehicle with an unmanned aerial vehicle
This paper addresses the perception, control, and trajectory planning for an aerial platform to identify and land on a moving car at 15 km/hr. The hexacopter unmanned aerial vehicle (UAV), equipped with onboard sensors and a computer, detects the car using a monocular camera and predicts the car future movement using a nonlinear motion model. While following the car, the UAV lands on its roof, and it attaches itself using magnetic legs. The proposed system is fully autonomous from takeoff to landing. Numerous field tests were conducted throughout the year-long development and preparations for the Mohamed Bin Zayed International Robotics Challenge (MBZIRC) 2017 competition, for which the system was designed. We propose a novel control system in which a model predictive controller is used in real time to generate a reference trajectory for the UAV, which are then tracked by the nonlinear feedback controller. This combination allows to track predictions of the car motion with minimal position error. The evaluation presents three successful autonomous landings during the MBZIRC 2017, where our system achieved the fastest landing among all competing teams.
prof. Ondřej Chum, doc. Georgios Tolias, Ahmet Iscen, PhD.: Label Propagation for Deep Semi-Supervised Learning
Semi-supervised learning is becoming increasingly important because it can combine data carefully labeled by humans with abundant unlabeled data to train deep neural networks. Classic methods on semi-supervised learning that have focused on transductive learning have not been fully exploited in the inductive framework followed by modern deep learning. The same holds for the manifold assumption-that similar examples should get the same prediction. In this work, we employ a transductive label propagation method that is based on the manifold assumption to make predictions on the entire dataset and use these predictions to generate pseudo-labels for the unlabeled data and train a deep neural network. At the core of the transductive method lies a nearest neighbor graph of the dataset that we create based on the embeddings of the same network. Therefore our learning process iterates between these two steps. We improve performance on several datasets especially in the few labels regime and show that our work is complementary to current state of the art.
doc. Jan Rusz, ing. Jan Hlavnička, PhD., ing. Michal Novotný, PhD., ing. Tereza Tykalová, PhD.: Speech Biomarkers in Rapid Eye Movement Sleep Behavior Disorder and Parkinson Disease
This multilanguage study used simple speech recording and high-end pattern analysis to provide sensitive and reliable noninvasive biomarkers of prodromal versus manifest α-synucleinopathy in patients with idiopathic rapid eye movement sleep behavior disorder (iRBD) and early-stage Parkinson disease (PD).
doc. Radim Šára + CIIRC: Off-line LiDAR and Camera System Calibration Software
Automation of individual transport systems is considered an up-and-coming prospect with the potential of greatly mitigating many of the challenges associated with intensified urbanization, while at the same time offering additional benefits for the citizens and drastically increasing overall street safety. However, due to the lack of maturity of involved key technologies and persisting legal limitations, full automation of on-road driving remains a longer-term vision, particularly in urban environments.
The goal and ambition of UP-Drive is to address these technological challenges through the development of an automated valet parking service for city environments, aimed at relieving a car driver from the burden of finding a parking space in city centers. Instead, the fully automated car navigates on its own through urban neighborhoods, finds a parking space and returns on-demand.
Creating such a system requires mastering all key technologies essential to automated urban driving beyond the current state-of-the-art: complete round-view perception of the vehicle environment, robust lifelong localization and mapping, sophisticated understanding of complex scenes as well as aggregation and integration of long-term semantic data over a cloud-based infrastructure. With this, we ensure that the research and development carried out in this project will directly be applicable to other urban driving use-cases such as driver assistance and safety systems on the one hand, and on the other hand to the transportation for elderly and citizens with handicaps, last-mile delivery of goods - and ultimately fully automated urban driving in general.
The consortium will continuously integrate the research and development from all partners into a fully functional vehicle platform and will showcase the end-product in its full extent to the general public.
