Ankush Chakrabarty

Volterra kernel based face recognition using artificial bee colonyoptimization

The present paper describes a novel method of implementation of a stochastic optimization technique for the face recognition problem. The method proposed divides the original images into patches in space, and seeks a non-linear functional mapping using second-order Volterra kernels. The artificial bee colony optimization technique, a modern stochastic optimization algorithm, is used to derive optimal Volterra kernels during training to simultaneously maximize inter-class distances and minimize intra-class distances in the feature space. During testing, a voting procedure is used in conjunction with a nearest neighbor classifier to decide to which class each individual patch belongs. Finally, the aggregate classification results of all patches in an image are used to determine the overall recognition outcome for the given image. The utility of the proposed scheme is aptly demonstrated by implementing it on two popular benchmark face recognition datasets, and comparing the effectiveness of the proposed approach vis-a-vis other statistical learning procedures in facial recognition and also several other methods developed so far. The effectiveness of the artificial bee colony optimization technique and its Levy-mutated variation in optimizing Volterra kernels is conclusively proven in this paper by significantly outperforming many popular contemporary algorithms.

Nonfragile Fault-Tolerant Fuzzy Observer-Based Controller Design for Nonlinear Systems

The problem of actuator fault estimation and fault-tolerant control for a class of uncertain nonlinear systems using Takagi-Sugeno fuzzy models is investigated. A design procedure for nonfragile proportional-integral (PI) observer is proposed to estimate the states of the nonlinear system and reconstruct the abrupt (modeled as step-like faults) and incipient fault signals. Subsequently, a nonfragile fault-tolerant controller is constructed, which is informed by the PI observer. Sufficient conditions of the existence of the PI observer and the fault-tolerant controller are provided in the form of linear matrix inequalities. The proposed fault-tolerant control architecture is tested on two numerical examples.

Model-based design of experiments for cellular processes

Model‐based design of experiments (MBDOE) assists in the planning of highly effective and efficient experiments. Although the foundations of this field are well‐established, the application of these techniques to understand cellular processes is a fertile and rapidly advancing area as the community seeks to understand ever more complex cellular processes and systems. This review discusses the MBDOE paradigm along with applications and challenges within the context of cellular processes and systems. It also provides a brief tutorial on Fisher information matrix (FIM)‐based and Bayesian experiment design methods along with an overview of existing software packages and computational advances that support MBDOE application and adoption within the Systems Biology community. As cell‐based products and biologics progress into the commercial sector, it is anticipated that MBDOE will become an essential practice for design, quality control, and production. WIREs Syst Biol Med 2013, 5:181–203. doi: 10.1002/wsbm.1204

Feedback linearizing indirect adaptive fuzzy control with foraging based on-line plant model estimation

The present paper describes the development of an indirect adaptive fuzzy control scheme employing feedback linearizing technique. The scheme proposes the development of a fuzzy certainty equivalence controller for controlling non-linear plants. This controller is designed on the basis of plant parameters estimated online at each sampling instant using bacterial foraging optimization (BFO) technique, a stochastic optimization technique, popularly employed in recent times. The utility of the proposed scheme is aptly demonstrated by implementing it to control the level in a surge tank under a variety of reference input commands, where the fuzzy controller could significantly out-perform the corresponding classical feedback linearizing controller and PSO-based fuzzy controller.

Robust explicit nonlinear model predictive control with integral sliding mode

A robust control strategy for stabilizing nonlinear systems in the presence of additive bounded disturbances is proposed. The proposed control architecture is a novel combination of explicit nonlinear model predictive control (EMPC) and integral sliding mode control (ISMC). Feasibility analysis of a finite-horizon optimal control problem involved in deriving the EMPC control action is performed over a polytope of interest in the state space. A sparse sampling-based boundary detection algorithm is employed to compute an approximating polynomial bounding the feasible region. A sparse-grid based interpolation scheme with Chebyshev-Gauss-Lobatto nodes and Legendre-basis polynomials are used to design the stabilizing EMPC surface. The proposed method is appealing because of the simplicity of the controller construction in conjunction with its applicability to higher-dimensional problems, which stems from the scale-ability property of sparse-grids. Robustness to the designed EMPC is provided by the ISMC. A simulated example is provided to illustrate the efficacy and performance of the proposed control strategy for the stabilization of an uncertain nonlinear dynamical system.

Support Vector Machine Informed Explicit Nonlinear Model Predictive Control Using Low-Discrepancy Sequences

In this paper, an explicit nonlinear model predictive controller (ENMPC) for the stabilization of nonlinear systems is investigated. The proposed ENMPC is constructed using tensored polynomial basis functions and samples drawn from low-discrepancy sequences. Solutions of a finite-horizon optimal control problem at the sampled nodes are used (1) to learn an inner and outer approximation of the feasible region of the ENMPC using support vector machines, and (2) to construct the ENMPC control surface on the computed feasible region using regression or sparse-grid interpolation, depending on the shape of the feasible region. The attractiveness of the proposed control scheme lies in its tractability to higher-dimensional systems with feasibility and stability guarantees, significantly small online computation times, and ease of implementation.

State and Unknown Input Observers for Nonlinear Systems With Bounded Exogenous Inputs

A systematic design methodology for state observers for a large class of nonlinear systems with bounded exogenous inputs (disturbance inputs and sensor noise) is proposed. The nonlinearities under consideration are characterized by an incremental quadratic constraint parameterized by a set of multiplier matrices. Linear matrix inequalities are developed to construct observer gains, which ensure that a performance output based on the state estimation error satisfies a prescribed degree of accuracy. Furthermore, conditions guaranteeing estimation of the unknown inputs to arbitrary degrees of accuracy are provided. The proposed scheme is illustrated with a numerical example, which does not satisfy the so-called “matching conditions.”

Intraperitoneal Insulin Delivery Provides Superior Glycemic Regulation to Subcutaneous Insulin Delivery in Model Predictive Control-based Fully-automated Artificial Pancreas in Patients with Type 1 Diabetes: A Pilot Study.

To compare intraperitoneal (IP) to subcutaneous (SC) insulin delivery in an artificial pancreas (AP).

Hyperspectral image classification incorporating bacterial foraging-optimized spectral weighting

The present paper describes the development of a hyperspectral image classification scheme using support vector machines (SVM) with spectrally weighted kernels. The kernels are designed during the training phase of the SVM using optimal spectral weights estimated using the Bacterial Foraging Optimization (BFO) algorithm, a popular modern stochastic optimization algorithm. The optimized kernel functions are then in the SVM paradigm for bi-classification of pixels in hyperspectral images. The effectiveness of the proposed approach is demonstrated by implementing it on three widely used benchmark hyperspectral data sets, two of which were taken over agricultural sites at Indian Pines, Indiana, and Salinas Valley, California, by the Airborne Visible Infra-Red Imaging Spectrometer (AVIRIS) at NASA’s Jet Propulsion Laboratory. The third dataset was acquired using the Reflective Optical System Imaging Spectrometer (ROSIS) over an urban scene at Pavia University, Italy to demonstrate the efficacy of the proposed approach in an urban scenario as well as with agricultural data. Classification errors for One-Against-One (OAO) and classification accuracies for One-Against-All (OAA) schemes were computed and compared to other methods developed in recent times. Finally, the use of the BFO-based technique is recommended owing to its superior performance, in comparison to other contemporary stochastic bio-inspired algorithms.

Delayed unknown input observers for discrete-time linear systems with guaranteed performance

Abstract In this paper, we propose a state and unknown input observer for discrete-time linear systems with bounded unknown inputs and measurement disturbances. The design procedure is formulated using a set of linear matrix inequalities, and leverages delayed (or fixed-lag) estimates. The observer error states and/or user-defined performance outputs are guaranteed to operate at certain performance bounds. Furthermore, by employing sufficiently large delays, the observer is guaranteed to provide exact asymptotic state and input estimates for minimum-phase systems. We demonstrate, via numerical examples, that the proposed observer can be used for a wider class of systems than those satisfying matching conditions.