A novel system for reducing turbo-lag by injection of compressed gas into the exhaust manifold

A key challenge in achieving good transient performance of highly boosted engines is the difficulty of accelerating the turbocharger from low air flow conditions (turbo lag). Multi-stage turbocharging, electric turbocharger assistance, electric compressors and hybrid powertrains are helpful in the mitigation of this deficit, but these technologies add significant cost and integration effort. Air-assist systems have the potential to be more cost-effective. Injecting compressed air into the intake manifold has received considerable attention, but the performance improvement offered by this concept is severely constrained by the compressor surge limit. The literature describes many schemes for generating the compressed gas, often involving significant mechanical complexity and/or cost. In this paper we demonstrate a novel exhaust assist system in which a reservoir is charged during braking. Experiments have been conducted using a 2.0 litre light-duty Diesel engine equipped with exhaust gas recirculation (EGR) and variable geometry turbine (VGT) coupled to an AC transient dynamometer, which was controlled to mimic engine load during in-gear braking and acceleration. The experimental results confirm that the proposed system reduces the time to torque during the 3rd gear tip-in by around 60%. Such a significant improvement was possible due to the increased acceleration of turbocharger immediately after the tip-in. Injecting the compressed gas into the exhaust manifold circumvents the problem of compressor surge and is the key enabler of the superior performance of the proposed concept. Copyright © 2013 SAE International.

Manifold covering theory in biomimetic pattern recognition

We describe a new model which is based on the concept of cognizing theory. The method identifies subsets of the data which are embedded in arbitrary oriented lower dimensional space. We definite manifold covering in biomimetic pattern recognition, and study its property. Furthermore, we propose this manifold covering algorithm based on Biomimetic Pattern Recognition. At last, the experimental results for face recognition demonstrates that the correct rejection rate of the test samples excluded in the classes of training samples is very high and effective.

Simultaneous Learning of Nonlinear Manifold and Dynamical Models for High-dimensional Time Series

The goal of this work is to learn a parsimonious and informative representation for high-dimensional time series. Conceptually, this comprises two distinct yet tightly coupled tasks: learning a low-dimensional manifold and modeling the dynamical process. These two tasks have a complementary relationship as the temporal constraints provide valuable neighborhood information for dimensionality reduction and conversely, the low-dimensional space allows dynamics to be learnt efficiently. Solving these two tasks simultaneously allows important information to be exchanged mutually. If nonlinear models are required to capture the rich complexity of time series, then the learning problem becomes harder as the nonlinearities in both tasks are coupled. The proposed solution approximates the nonlinear manifold and dynamics using piecewise linear models. The interactions among the linear models are captured in a graphical model. By exploiting the model structure, efficient inference and learning algorithms are obtained without oversimplifying the model of the underlying dynamical process. Evaluation of the proposed framework with competing approaches is conducted in three sets of experiments: dimensionality reduction and reconstruction using synthetic time series, video synthesis using a dynamic texture database, and human motion synthesis, classification and tracking on a benchmark data set. In all experiments, the proposed approach provides superior performance.

Simultaneous Learning of Non-Linear Manifold and Dynamical Models for High-Dimensional Time Series

The goal of this work is to learn a parsimonious and informative representation for high-dimensional time series. Conceptually, this comprises two distinct yet tightly coupled tasks: learning a low-dimensional manifold and modeling the dynamical process. These two tasks have a complementary relationship as the temporal constraints provide valuable neighborhood information for dimensionality reduction and conversely, the low-dimensional space allows dynamics to be learnt efficiently. Solving these two tasks simultaneously allows important information to be exchanged mutually. If nonlinear models are required to capture the rich complexity of time series, then the learning problem becomes harder as the nonlinearities in both tasks are coupled. The proposed solution approximates the nonlinear manifold and dynamics using piecewise linear models. The interactions among the linear models are captured in a graphical model. The model structure setup and parameter learning are done using a variational Bayesian approach, which enables automatic Bayesian model structure selection, hence solving the problem of over-fitting. By exploiting the model structure, efficient inference and learning algorithms are obtained without oversimplifying the model of the underlying dynamical process. Evaluation of the proposed framework with competing approaches is conducted in three sets of experiments: dimensionality reduction and reconstruction using synthetic time series, video synthesis using a dynamic texture database, and human motion synthesis, classification and tracking on a benchmark data set. In all experiments, the proposed approach provides superior performance.

Efficient tracking of human poses using a manifold hierarchy

In this paper a 3D human pose tracking framework is presented. A new dimensionality reduction method (Hierarchical Temporal Laplacian Eigenmaps) is introduced to represent activities in hierarchies of low dimensional spaces. Such a hierarchy provides increasing independence between limbs, allowing higher flexibility and adaptability that result in improved accuracy. Moreover, a novel deterministic optimisation method (Hierarchical Manifold Search) is applied to estimate efficiently the position of the corresponding body parts. Finally, evaluation on public datasets such as HumanEva demonstrates that our approach achieves a 62.5mm-65mm average joint error for the walking activity and outperforms state-of-the-art methods in terms of accuracy and computational cost.

Croissance et ensemble nodal de fonctions propres du laplacien sur des surfaces

Dans cette thèse, nous étudions les fonctions propres de l'opérateur de Laplace-Beltrami - ou simplement laplacien - sur une surface fermée, c'est-à-dire une variété riemannienne lisse, compacte et sans bord de dimension 2. Ces fonctions propres satisfont l'équation $\Delta_g \phi_\lambda + \lambda \phi_\lambda = 0$ et les valeurs propres forment une suite infinie. L'ensemble nodal d'une fonction propre du laplacien est celui de ses zéros et est d'intérêt depuis les expériences de plaques vibrantes de Chladni qui remontent au début du 19ème siècle et, plus récemment, dans le contexte de la mécanique quantique. La taille de cet ensemble nodal a été largement étudiée ces dernières années, notamment par Donnelly et Fefferman, Colding et Minicozzi, Hezari et Sogge, Mangoubi ainsi que Sogge et Zelditch. L'étude de la croissance de fonctions propres n'est pas en reste, avec entre autres les récents travaux de Donnelly et Fefferman, Sogge, Toth et Zelditch, pour ne nommer que ceux-là. Notre thèse s'inscrit dans la foulée du travail de Nazarov, Polterovich et Sodin et relie les propriétés de croissance des fonctions propres avec la taille de leur ensemble nodal dans l'asymptotique $\lambda \nearrow \infty$. Pour ce faire, nous considérons d'abord les exposants de croissance, qui mesurent la croissance locale de fonctions propres et qui sont obtenus à partir de la norme uniforme de celles-ci. Nous construisons ensuite la croissance locale moyenne d'une fonction propre en calculant la moyenne sur toute la surface de ces exposants de croissance, définis sur de petits disques de rayon comparable à la longueur d'onde. Nous montrons alors que la taille de l'ensemble nodal est contrôlée par le produit de cette croissance locale moyenne et de la fréquence $\sqrt{\lambda}$. Ce résultat permet une reformulation centrée sur les fonctions propres de la célèbre conjecture de Yau, qui prévoit que la mesure de l'ensemble nodal croît au rythme de la fréquence. Notre travail renforce également l'intuition répandue selon laquelle une fonction propre se comporte comme un polynôme de degré $\sqrt{\lambda}$. Nous généralisons ensuite nos résultats pour des exposants de croissance construits à partir de normes $L^q$. Nous sommes également amenés à étudier les fonctions appartenant au noyau d'opérateurs de Schrödinger avec petit potentiel dans le plan. Pour de telles fonctions, nous obtenons deux résultats qui relient croissance et taille de l'ensemble nodal.