Alexandre Saidi

Reliability assessment of complex mechatronic systems using a modified nonparametric belief propagation algorithm

Various parametric skewed distributions are widely used to model the time-to-failure (TTF) in the reliability analysis of mechatronic systems, where many items are unobservable due to the high cost of testing. Estimating the parameters of those distributions becomes a challenge. Previous research has failed to consider this problem due to the difficulty of dependency modeling. Recently the methodology of Bayesian networks (BNs) has greatly contributed to the reliability analysis of complex systems. In this paper, the problem of system reliability assessment (SRA) is formulated as a BN considering the parameter uncertainty. As the quantitative specification of BN, a normal distribution representing the stochastic nature of TTF distribution is learned to capture the interactions between the basic items and their output items. The approximation inference of our continuous BN model is performed by a modified version of nonparametric belief propagation (NBP) which can avoid using a junction tree that is inefficient for the mechatronic case because of the large treewidth. After reasoning, we obtain the marginal posterior density of each TTF model parameter. Other information from diverse sources and expert priors can be easily incorporated in this SRA model to achieve more accurate results. Simulation in simple and complex cases of mechatronic systems demonstrates that the posterior of the parameter network fits the data well and the uncertainty passes effectively through our BN based SRA model by using the modified NBP.

A dynamic-reliable multiple model adaptive controller for active vehicle suspension under uncertainties

The inherent uncertainties of vehicle suspension systems challenge not only the capability of ride comfort and handling performance, but also the reliability requirement. In this research, a dynamic-reliable multiple model adaptive (MMA) controller is developed to overcome the difficulty of suspension uncertainties while considering performance and reliability at the same time. The MMA system consists of a finite number of optimal sub-controllers and employs a continuous-time based Markov chain to guide the jumping among the sub-controllers. The failure mode considered is the bottoming and topping of suspension components. A limitation on the failure probability is imposed to penalize the performance of the sub-controllers and a gradient-based genetic algorithm yields their optimal feedback gains. Finally, the dynamic reliability of the MMA controller is approximated by using the integration of state covariances and a judging condition is induced to assert that the MMA system is dynamic-reliable. In numerical simulation, a long scheme with piecewise time-invariant parameters is employed to examine the performance and reliability under the uncertainties of sprung mass, road condition and driving velocity. It is shown that the dynamic-reliable MMA controller is able to trade a small amount of model performance for extra reliability.

Learning-Based Driving Events Recognition and Its Application to Digital Roads

Automatic recognition of driving events, e.g., approaching roundabouts, is important both for the truck design process based on simulated road data and for advanced driver assistance systems. However, the problem faced is extremely challenging as only in-vehicle driving data must be used, whereas the number of driving events is usually quite large. In this paper, we propose a learning-based driving events classification method, which is trained and tested with a real driving events database. The proposed method includes definition of driving events relevant to our final application, selection of discriminating features, and classification, using two machine-learning techniques, namely, decision trees and linear logistic regression. We then introduce the digital road concept. This consists of simulated road data used in the truck design process to quantify the behavior of a truck, particularly in terms of fuel consumption. While a digital road typically contains far less driving information, we show that we can still apply the proposed driving events recognition models learnt on real driving data and pave the way for a more realistic assessment of truck characteristics via simulation tools.

Learning-based driving events classification

Drivers typically depict different behavior with respect to various driving events. The modeling of their behavior enables an accurate estimation of fuel consumption during the truck design process and is also helpful for ADAS in order to give relevant advices. In this paper, we propose a learning-based approach to the automatic recognition of driving events, e.g., roundabouts or stops, which impact the driver behavior. We first synthesize and categorize meaningful driving events and then study a set of features potentially sensitive to the driver behavior. These features were experimented on real truck driver data using two machine-learning techniques, i.e., decision tree and linear logic regression, to evaluate their relevance and ability to recognize driving events.

Discriminative Transfer Learning Using Similarities and Dissimilarities

Transfer learning (TL) aims at solving the problem of learning an effective classification model for a target category, which has few training samples, by leveraging knowledge from source categories with far more training data. We propose a new discriminative TL (DTL) method, combining a series of hypotheses made by both the model learned with target training samples and the additional models learned with source category samples. Specifically, we use the sparse reconstruction residual as a basic discriminant and enhance its discriminative power by comparing two residuals from a positive and a negative dictionary. On this basis, we make use of similarities and dissimilarities by choosing both positively correlated and negatively correlated source categories to form additional dictionaries. A new Wilcoxon–Mann–Whitney statistic-based cost function is proposed to choose the additional dictionaries with unbalanced training data. Also, two parallel boosting processes are applied to both the positive and negative data distributions to further improve classifier performance. On two different image classification databases, the proposed DTL consistently outperforms other state-of-the-art TL methods while at the same time maintaining very efficient runtime.

Using Grammatical Inference For Structure Induction

Given the huge quantity of the current available textual information, text mining process tackles the task of searching useful knowledge in a natural language document. When dealing with a free-format textual corpus (e.g. a job announcement) where the linguistic rules are not respected, the time consuming morpho-syntactic analysis is not of a great help. However, text mining techniques process may exploit linguistic sub-structures in the text. In this paper, we present an applications of grammatical inference (GI) in a machine learning system applied to a text corpus. We specify and use the process of the grammatical inference as an instance of the constraint satisfaction problem that instantiates automata in a (language inclusion) lattice

Learning visual categories through a sparse representation classifier based cross-category knowledge transfer

To solve the challenging task of learning effective visual categories with limited training samples, we propose a new sparse representation classifier based transfer learning method, namely SparseTL, which propagates the cross-category knowledge from multiple source categories to the target category. Specifically, we enhance the target classification task in learning a both generative and discriminative sparse representation based classifier using pairs of source categories most positively and most negatively correlated to the target category. We further improve the discriminative ability of the classifier by choosing the most discriminative bins in the feature vector with a feature selection process. The experimental results show that the proposed method achieves competitive performance on the NUS-WIDE Scene database compared to several state of the art transfer learning algorithms while keeping a very efficient runtime.

Using CLP to characterise linguistic lattice boundaries in a text mining process

In this paper, we expose the use of CLP in a Textual Data Mining Task. Text Mining process is here applied to a corpus of semi-structured documents like seminary and job announcement. Such documents contain semi-structured sections each of which will be recognised by an automaton whose language is characterised by a set of CLP rules.