Two-level Hierarchical Scheme Research Articles

A Hierarchical Scheme for Video‐Based Person Re‐identification Using Lightweight PCANet and Handcrafted LOMO Features

A two-level hierarchical scheme for videobased person re-identification (re-id) is presented, with the aim of learning a pedestrian appearance model through more complete walking cycle extraction. Specifically, given a video with consecutive frames, the objective of the first level is to detect the key frame with lightweight Convolutional neural network (CNN) of PCANet to reflect the summary of the video content. At the second level, on the basis of the detected key frame, the pedestrian walking cycle is extracted from the long video sequence. Moreover, local features of Local maximal occurrence (LOMO) of the walking cycle are extracted to represent the pedestrian' s appearance information. In contrast to the existing walking-cycle-based person re-id approaches, the proposed scheme relaxes the limit on step number for a walking cycle, thus making it flexible and less affected by noisy frames. Experiments are conducted on two benchmark datasets: PRID 2011 and iLIDS-VID. The experimental results demonstrate that our proposed scheme outperforms the six state-of-art video-based re-id methods, and is more robust to the severe video noises and variations in pose, lighting, and camera viewpoint.

Multidisciplinary Design of Electric Vehicles Based on Hierarchical Multi-Objective Optimization

A method for the optimal design of complex systems is developed by effectively combining multi-objective optimization and analytical target cascading techniques. The complex systems with high dimensionality are partitioned into manageable subsystems that can be optimized using dedicated algorithms. The multiple objective functions in each subsystem are treated simultaneously, and the interactions between subsystems are managed using linking variables and shared variables. The analytical target cascading algorithm ensures the convergence of the optimal solution that meets the system level targets while complying with the subsystem level constraints. A design optimization of electric vehicles with in-wheel motors is formulated as a two-level hierarchical scheme where the top level has a model representing the electric vehicle and the bottom level contains models of battery and suspension. The vehicle model includes an electric motor model and a power electronics model. Pareto-optimal solutions are derived holistically. The effectiveness of the proposed method for optimizing the complex systems is compared against the conventional all-in-one optimization approach.

Color Compressive Sensing Imaging of Arbitrary-Shaped Scatterers

The problem of imaging arbitrary-shaped targets is addressed through a methodological generalization of the compressive sensing (CS) paradigm. The Color CS (C-CS) methodology consists of a two-level hierarchical scheme where, at the first level and using a basis dictionary , several sparsity-regularized inversions are performed in parallel, while, at the second level, the retrieved expansion coefficients are compared to select the most reliable reconstruction according to a sparsity-rewarding criterion. Within the Born-approximated formulation of the microwave imaging problem, a Bayesian solver and a filtered $\ell _{0}$ -norm criterion are employed to implement the first and the second C-CS levels, respectively. Selected numerical results, representative of an extensive validation, are presented to illustrate the features, the advantages, and the limitations of C-CS also in comparison with some competitive state-of-the-art inversion techniques.

Hierarchical Co-Ordinated Wide Area and Local Controls of DFIG Wind Turbine and PSS for Robust Power Oscillation Damping

In this paper, the two-level hierarchical scheme, which consists of wide area centralized and local controls of the power oscillation damper (POD) installed with the doubly-fed induction generator (DFIG) wind turbine and the power system stabilizer (PSS) has been proposed for robust power oscillation damping. In the wide area level, the centralized POD and PSS has received the input signals from synchronized phasor measurement units (PMUs). The geometric measures of controllability and observability have been applied to select the suitable DFIG and synchronous generator (SG) for stabilizing the target oscillation modes, the proper input signals of the centralized POD and PSS, and the location of PMUs. In the local level, the suitable DFIG and SG have been equipped with POD and PSS, respectively. In the parameters optimization of POD and PSS, the practical issues such as damping performance, controller structure, communication latency, and robustness against system uncertainties have been considered. The controller efficiency and resiliency of the proposed controller have been evaluated in comparison with other controllers by eigenvalue analysis and nonlinear simulation for a wide range of operating conditions, line outage contingencies, severe faults, and communication failure.

An optimization method for metamorphic mechanisms based on multidisciplinary design optimization

The optimization of metamorphic mechanisms is different from that of the conventional mechanisms for its characteristics of multi-configuration. There exist complex coupled design variables and constraints in its multiple different configuration optimization models. To achieve the compatible optimized results of these coupled design variables, an optimization method for metamorphic mechanisms is developed in the paper based on the principle of multidisciplinary design optimization (MDO). Firstly, the optimization characteristics of the metamorphic mechanism are summarized distinctly by proposing the classification of design variables and constraints as well as coupling interactions among its different configuration optimization models. Further, collaborative optimization technique which is used in MDO is adopted for achieving the overall optimization performance. The whole optimization process is then proposed by constructing a two-level hierarchical scheme with global optimizer and configuration optimizer loops. The method is demonstrated by optimizing a planar five-bar metamorphic mechanism which has two configurations, and results show that it can achieve coordinated optimization results for the same parameters in different configuration optimization models.

A new hierarchical parallelization scheme: generalized distributed data interface (GDDI), and an application to the fragment molecular orbital method (FMO).

A two-level hierarchical scheme, generalized distributed data interface (GDDI), implemented into GAMESS is presented. Parallelization is accomplished first at the upper level by assigning computational tasks to groups. Then each group does parallelization at the lower level, by dividing its task into smaller work loads. The types of computations that can be used with this scheme are limited to those for which nearly independent tasks and subtasks can be assigned. Typical examples implemented, tested, and analyzed in this work are numeric derivatives and the fragment molecular orbital method (FMO) that is used to compute large molecules quantum mechanically by dividing them into fragments. Numeric derivatives can be used for algorithms based on them, such as geometry optimizations, saddle-point searches, frequency analyses, etc. This new hierarchical scheme is found to be a flexible tool easily utilizing network topology and delivering excellent performance even on slow networks. In one of the typical tests, on 16 nodes the scalability of GDDI is 1.7 times better than that of the standard parallelization scheme DDI and on 128 nodes GDDI is 93 times faster than DDI (on a multihub Fast Ethernet network). FMO delivered scalability of 80-90% on 128 nodes, depending on the molecular system (water clusters and a protein). A numerical gradient calculation for a water cluster achieved a scalability of 70% on 128 nodes. It is expected that GDDI will become a preferred tool on massively parallel computers for appropriate computational tasks.

An expert approach for control of fermentation processes as variable structure plants

Structural variability is a fundamental property of biological systems. Fermentation processes, in particular, are often accompanied by physiological phenomena which have the characteristics of structural alterations. The conventional control approach does not have the competence to handle such processes, because it is based on the general assumption for a single-structure plant. In order to overcome this limitation, we consider under some conditions fermentation processes as variable structure plants and propose a two-level hierarchical scheme for their control. At the higher hierarchical level, which performs the organizing functions and operates in the structural space of the plant, the current plant structure is recognized as an element of a finite set of structures defined on the basis of ‘deep’ expert knowledge of the cell physiology and/or ‘shallow’ expert knowledge of the behavioral characteristics of the microbial population. The recognition is based on fuzzy decomposition of the structural space of the plant into several subspaces, mapped isomorphically into a respective set of rational control strategies. A set of expert rules is used for the actual synthesis of the recognition procedure, which is finally tuned by supervised learning. The adopted methodology at this hierarchical level provides the system with a kind of intelligence. To the lower hierarchical level, two distinct functions are assigned. The first one performs the global task of control of the plant structure and operates in its structural space. The second function has a local character and consists of control of the plant in a particular subspace of constant structure. It works in the state space of the plant. Under supervision of the higher level, the control strategy relevant to the current plant structure is picked out from a predefined pool and the corresponding control action is calculated. Extensive use of expert knowledge in the conceptual formulation of proper control strategies is assumed. Compared with the conventional single-structure approach, the variable structure concept results in rational and physiologically motivated control of fermentation processes. Instead of the development of a single but complicated process model and control structure, construction of several simple control structures, often in a ‘model-less’ form, is required. This method naturally incorporates and theoretizes the widely accepted policy of artificial induction of structural transformations in the plant, which is essential for enhanced productivity in many modern fermentation technologies. The proposed approach has been realized as a real-time software system, provided with enough flexibility to be used with different fermentation processes and equipment. This system is orientated to IBM personal computers and works under the QNX multi-tasking real-time operating system.