Browsing by Author "Lee, Der-Hau"

Now showing 1 - 5 of 5

Deep learning and control algorithms of direct perception for autonomous driving
(1970-01-01)
Lee, Der-Hau
;
Chen, Kuan-Lin
;
Liou, Kuan-Han
;
Liu, Chang-Lun
;
Liu, Jinn-Liang
We propose an end-to-end machine learning model that integrates multi-task (MT) learning, convolutional neural networks (CNNs), and control algorithms to achieve efficient inference and stable driving for self-driving cars. The CNN-MT model can simultaneously perform regression and classification tasks for estimating perception indicators and driving decisions, respectively, based on the direct perception paradigm of autonomous driving. The model can also be used to evaluate the inference efficiency and driving stability of different CNNs on the metrics of CNN's size, complexity, accuracy, processing speed, and collision number, respectively, in a dynamic traffic. We also propose new algorithms for controllers to drive a car using the indicators and its short-range sensory data to avoid collisions in real-time testing. We collect a set of images from a camera of The Open Racing Car Simulator in various driving scenarios, train the model using this dataset, test it in unseen traffics, and find that it outperforms earlier models in highway traffic. The stability of end-to-end learning and self driving depends crucially on the dynamic interplay between CNN and control algorithms. The source code and data of this work are available on our website, which can be used as a simulation platform to evaluate different learning models on equal footing and quantify collisions precisely for further studies on autonomous driving.
Kane-Mele Hubbard model on a zigzag ribbon: Stability of the topological edge states and quantum phase transitions
(2014-07-14)
Chung, Chung-Hou
;
Lee, Der-Hau
;
Chao, Sung-Po
;
電子物理學系
;
Department of Electrophysics
We study the quantum phases and phase transitions of the Kane-Mele Hubbard (KMH) model on a zigzag ribbon of honeycomb lattice at a finite size via the weak-coupling renormalization group (RG) approach. In the noninteracting limit, the Kane-Mele (KM) model is known to support topological edge states where electrons show helical property with orientations of the spin and momentum being locked. The effective interedge hopping terms are generated due to finite-size effect. In the presence of an on-site Coulomb (Hubbard) interaction and the interedge hoppings, special focus is put on the stability of the topological edge states (TI phase) in the KMH model against (i) the charge and spin gaped (II) phase, (ii) the charge gaped but spin gapless (IC) phase, and (iii) the spin gaped but charge gapless (CI) phase depending on the number (even/odd) of the zigzag ribbons, doping level (electron filling factor) and the ratio of the Coulomb interaction to the interedge tunneling. We discuss different phase diagrams for even and odd numbers of zigzag ribbons. We find the TI-CI, II-IC, and II-CI quantum phase transitions are of the Kosterlitz-Thouless (KT) type. By computing various correlation functions, we further analyze the nature and leading instabilities of these phases. The relevance of our results for graphene is discussed.
Non-Centrosymmetric Superconductors on Honeycomb Lattice
(2018-09-01)
Lee, Der-Hau
;
Chung, Chung-Hou
;
電子物理學系
;
Department of Electrophysics
Non-centrosymmetric topological superconductivity is studied in correlated doped quantum spin-Hall insulators (QSHI) on honeycomb lattice without inversion symmetry where the intrinsic (Kane-Mele) and Rashba spin-orbit couplings can in general exist. The generic topologically non-trivial superconducting phase diagram of the model system is explored. Over a certain parameter range, the parity-mixing superconducting state with co-existing spin-singlet d+id and spin-triplet p+ip-wave pairing is found. On a zigzag nanoribbon, the parity-mixing superconducting state shows co-existing helical and chiral Majorana fermions at edges. The relevance of our results for experiments is discussed.
Short-time power-law blinking statistics of single quantum dots and a test of the diffusion-controlled electron transfer model
(2009-08-14)
Tang, Jau
;
Lee, Der-Hau
;
Yeh, Yi-Cheun
;
Yuan, Chi-Tsu
;
光電工程學系
In this work we analyzed the blinking statistics data of single CdSe/ZnS quantum dots at vey short times to test some predictions of the diffusion-controlled electron transfer (DCET) model. Using autocorrelation function (ACF) approach we could extract the exponent of the inverse power-law blinking statistics down to 1 mu s. Such an approach also minimizes human subjectivity in choosing a bin time and an on-off threshold. We showed that the observed stretched exponential decay in the ACF and its relationship to the blinking statistics are consistent with the DCET model, and we set an upper bound for the characteristic time constant t(c).
Unusually large exponent for the inverse power-law blinking of single chromophores
(2008-10-09)
Lee, Der-Hau
;
Tang, Jau
;
光電工程學系
;
Department of Photonics
In this study of single-particle fluorescence intermittency, we investigate some Interesting blinking phenomena of chromophores by Yeow et al. (J. Phys. Chem. A 2006, 110, 1726), showing behavior very different from the usual quantum dots. Using Monte Carlo simulations we demonstrate that the short-time blinking statistics are generally nonexponential but could be characterized by an inverse power law with an effective exponent m far exceeding 2, as observed in some single-chromophores experiments. We show that in. for the on-events decreases with the effective diffusion correlation time which increases with light intensity. This work offers an explanation to why m could change from 2.8 to 2.0 in their experiments as the light intensity is quadrupled.