Output error MISO system identification using fractional models

This paper deals with system identification for continuous-time multiple-input single-output (MISO) fractional differentiation models. An output error optimization algorithm is proposed for estimating all parameters, namely the coefficients and the differentiation orders. Given the high number of parameters to be estimated, the output error method can converge to a local minimum. Therefore, an initialization procedure is proposed to help the convergence to the optimum by using three variants of the algorithm. Moreover, a new definition of structured-commensurability (or S-commensurability) has been introduced to cope with the differentiation order estimation. First, a global S-commensurate order is estimated for all subsystems. Then, local S-commensurate orders are estimated (one for each subsystem). Finally the S-commensurability constraint being released, all differentiation orders are further adjusted. Estimating a global S-commensurate order greatly reduces the number of parameters and helps initializing the second variant, where local S-commensurate orders are estimated which, in turn, are used as a good initial hit for the last variant. It is known that such an initialization procedure progressively increases the number of parameters and provides good efficiency of the optimization algorithm. Monte Carlo simulation analysis are provided to evaluate the performances of this algorithm.

Composite optimization for robust rank one bilinear sensing

We consider the task of recovering a pair of vectors from a set of rank one bilinear measurements, possibly corrupted by noise. Most notably, the problem of robust blind deconvolution can be modeled in this way. We consider a natural nonsmooth formulation of the rank one bilinear sensing problem and show that its moduli of weak convexity, sharpness and Lipschitz continuity are all dimension independent, under favorable statistical assumptions. This phenomenon persists even when up to half of the measurements are corrupted by noise. Consequently, standard algorithms, such as the subgradient and prox-linear methods, converge at a rapid dimension-independent rate when initialized within a constant relative error of the solution. We complete the paper with a new initialization strategy, complementing the local search algorithms. The initialization procedure is both provably efficient and robust to outlying measurements. Numerical experiments, on both simulated and real data, illustrate the developed theory and methods.

Optimization of the launcher ascent trajectory leading to the global optimum without any initialization: the breakthrough of the Hamilton–Jacobi–Bellman approach

The resolution of the launcher ascent trajectory problem by the so-called Hamilton–Jacobi–Bellman (HJB) approach, relying on the Dynamic Programming Principle, has been investigated. The method gives a global optimum and does not need any initialization procedure. Despite these advantages, this approach is seldom used because of the dicculties of computing the solution of the HJB equation for high dimension problems. The present study shows that an eccient resolution is found. An illustration of the method is proposed on a heavy class launcher, for a typical GEO (Geostationary Earth Orbit) mission. This study has been performed in the frame of the Centre National d’Etudes Spatiales (CNES) Launchers Research & Technology Program.

Shrimp Feed Formulation via Evolutionary Algorithm with Power Heuristics for Handling Constraints

Formulating feed for shrimps represents a challenge to farmers and industry partners. Most previous studies selected from only a small number of ingredients due to cost pressures, even though hundreds of potential ingredients could be used in the shrimp feed mix. Even with a limited number of ingredients, the best combination of the most appropriate ingredients is still difficult to obtain due to various constraint requirements, such as nutrition value and cost. This paper proposes a new operator which we call Power Heuristics, as part of an Evolutionary Algorithm (EA), which acts as a constraint handling technique for the shrimp feed or diet formulation. The operator is able to choose and discard certain ingredients by utilising a specialized search mechanism. The aim is to achieve the most appropriate combination of ingredients. Power Heuristics are embedded in the EA at the early stage of a semirandom initialization procedure. The resulting combination of ingredients, after fulfilling all the necessary constraints, shows that this operator is useful in discarding inappropriate ingredients when a crucial constraint is violated.

Wavelet Network: Online Sequential Extreme Learning Machine for Nonlinear Dynamic Systems Identification

A single hidden layer feedforward neural network (SLFN) with online sequential extreme learning machine (OSELM) algorithm has been introduced and applied in many regression problems successfully. However, using SLFN with OSELM as black-box for nonlinear system identification may lead to building models for the identified plant with inconsistency responses from control perspective. The reason can refer to the random initialization procedure of the SLFN hidden node parameters with OSELM algorithm. In this paper, a single hidden layer feedforward wavelet network (WN) is introduced with OSELM for nonlinear system identification aimed at getting better generalization performances by reducing the effect of a random initialization procedure.