A Neural Network Architecture for Autonomous Learning, Recognition, and Prediction in a Nonstationary World

In a constantly changing world, humans are adapted to alternate routinely between attending to familiar objects and testing hypotheses about novel ones. We can rapidly learn to recognize and narne novel objects without unselectively disrupting our memories of familiar ones. We can notice fine details that differentiate nearly identical objects and generalize across broad classes of dissimilar objects. This chapter describes a class of self-organizing neural network architectures--called ARTMAP-- that are capable of fast, yet stable, on-line recognition learning, hypothesis testing, and naming in response to an arbitrary stream of input patterns (Carpenter, Grossberg, Markuzon, Reynolds, and Rosen, 1992; Carpenter, Grossberg, and Reynolds, 1991). The intrinsic stability of ARTMAP allows the system to learn incrementally for an unlimited period of time. System stability properties can be traced to the structure of its learned memories, which encode clusters of attended features into its recognition categories, rather than slow averages of category inputs. The level of detail in the learned attentional focus is determined moment-by-moment, depending on predictive success: an error due to over-generalization automatically focuses attention on additional input details enough of which are learned in a new recognition category so that the predictive error will not be repeated. An ARTMAP system creates an evolving map between a variable number of learned categories that compress one feature space (e.g., visual features) to learned categories of another feature space (e.g., auditory features). Input vectors can be either binary or analog. Computational properties of the networks enable them to perform significantly better in benchmark studies than alternative machine learning, genetic algorithm, or neural network models. Some of the critical problems that challenge and constrain any such autonomous learning system will next be illustrated. Design principles that work together to solve these problems are then outlined. These principles are realized in the ARTMAP architecture, which is specified as an algorithm. Finally, ARTMAP dynamics are illustrated by means of a series of benchmark simulations.

A benchmark comparison between reconfigurable, intelligent and autonomous wireless inertial measurement and photonic technologies in rehabilitation

Advanced sensory systems address a number of major obstacles towards the provision for cost effective and proactive rehabilitation. Many of these systems employ technologies such as high-speed video or motion capture to generate quantitative measurements. However these solutions are accompanied by some major limitations including extensive set-up and calibration, restriction to indoor use, high cost and time consuming data analysis. Additionally many do not quantify improvement in a rigorous manner for example gait analysis for 5 minutes as opposed to 24 hour ambulatory monitoring. This work addresses these limitations using low cost, wearable wireless inertial measurement as a mobile and minimal infrastructure alternative. In cooperation with healthcare professionals the goal is to design and implement a reconfigurable and intelligent movement capture system. A key component of this work is an extensive benchmark comparison with the 'gold standard' VICON motion capture system.

DC/DC converter 3D assembly for autonomous sensor nodes

This paper reports on the design and the manufacturing of an integrated DCDC converter, which respects the specificity of sensor node network: compactness, high efficiency in acquisition and transmission modes, and compatibility with miniature Lithium batteries. A novel integrated circuit (ASIC) has been designed and manufactured to provide regulated Voltage to the sensor node from miniaturized, thin film Lithium batteries. Then, a 3D integration technique has been used to integrate this ASIC in a 3 layers stack with high efficiency passives components, mixing the wafer level technologies from two different research institutions. Electrical results have demonstrated the feasibility of this integrated system and experiments have shown significant improvements in the case of oscillations in regulated voltage. However, stability of this output voltage toward the input voltage has still to be improved.

Bridge-vehicle interaction for structural health monitoring: potentials, applications, and limitations

Structural Health Monitoring (SHM) is an integral part of infrastructure maintenance and management systems due to socio-economic, safety and security reasons. The behaviour of a structure under vibration depends on structure characteristics. The change of structure characteristics may suggest the change in system behaviour due to the presence of damage(s) within. Therefore the consistent, output signal guided, and system dependable markers would be convenient tool for the online monitoring, the maintenance, rehabilitation strategies, and optimized decision making policies as required by the engineers, owners, managers, and the users from both safety and serviceability aspects. SHM has a very significant advantage over traditional investigations where tangible and intangible costs of a very high degree are often incurred due to the disruption of service. Additionally, SHM through bridge-vehicle interaction opens up opportunities for continuous tracking of the condition of the structure. Research in this area is still in initial stage and is extremely promising. This PhD focuses on using bridge-vehicle interaction response for SHM of damaged or deteriorating bridges to monitor or assess them under operating conditions. In the present study, a number of damage detection markers have been investigated and proposed in order to identify the existence, location, and the extent of an open crack in the structure. The theoretical and experimental investigation has been conducted on Single Degree of Freedom linear system, simply supported beams. The novel Delay Vector Variance (DVV) methodology has been employed for characterization of structural behaviour by time-domain response analysis. Also, the analysis of responses of actual bridges using DVV method has been for the first time employed for this kind of investigation.

Resonant power conversion topologies for inductive charging of electrical vehicle batteries

This thesis is concerned with inductive charging of electric vehicle batteries. Rectified power form the 50/60 Hz utility feeds a dc-ac converter which delivers high-frequency ac power to the electric vehicle inductive coupling inlet. The inlet configuration has been defined by the Society of Automotive Engineers in Recommended Practice J-1773. This thesis studies converter topologies related to the series resonant converter. When coupled to the vehicle inlet, the frequency-controlled series-resonant converter results in a capacitively-filtered series-parallel LCLC (SP-LCLC) resonant converter topology with zero voltage switching and many other desirable features. A novel time-domain transformation analysis, termed Modal Analysis, is developed, using a state variable transformation, to analyze and characterize this multi-resonant fourth-orderconverter. Next, Fundamental Mode Approximation (FMA) Analysis, based on a voltage-source model of the load, and its novel extension, Rectifier-Compensated FMA (RCFMA) Analysis, are developed and applied to the SP-LCLC converter. The RCFMA Analysis is a simpler and more intuitive analysis than the Modal Analysis, and provides a relatively accurate closed-form solution for the converter behavior. Phase control of the SP-LCLC converter is investigated as a control option. FMA and RCFMA Analyses are used for detailed characterization. The analyses identify areas of operation, which are also validated experimentally, where it is advantageous to phase control the converter. A novel hybrid control scheme is proposed which integrates frequency and phase control and achieves reduced operating frequency range and improved partial-load efficiency. The phase-controlled SP-LCLC converter can also be configured with a parallel load and is an excellent option for the application. The resulting topology implements soft-switching over the entire load range and has high full-load and partial-load efficiencies. RCFMA Analysis is used to analyze and characterize the new converter topology, and good correlation is shown with experimental results. Finally, a novel single-stage power-factor-corrected ac-dc converter is introduced, which uses the current-source characteristic of the SP-LCLC topology to provide power factor correction over a wide output power range from zero to full load. This converter exhibits all the advantageous characteristics of its dc-dc counterpart, with a reduced parts count and cost. Simulation and experimental results verify the operation of the new converter.

CloudNeg: An autonomous multi issue negotiation system, with preference elicitation component, for trading cloud services

Cloud services provide its users with flexible resource provisioning. But in the current market, a user has to choose from a limited set of configurations at a fixed price. This paper presents an autonomous negotiation system termed CloudNeg for negotiating cloud services. CloudNeg provides buyers and sellers of cloud services with autonomous agents to negotiate on the specifications of a cloud instance, including price, on their behalf. These agents elicit their buyers’ time preferences and use them in negotiations. Further, this paper presents two artifacts: a negotiation algorithm and a prototype which together form CloudNeg.

Adenovirus F protein as a delivery vehicle for botulinum B.

BACKGROUND: Immunization with recombinant carboxyl-terminal domain of the heavy chain (Hc domain) of botulinum neurotoxin (BoNT) stimulates protective immunity against native BoNT challenge. Most studies developing a botulism vaccine have focused on the whole Hc; however, since the principal protective epitopes are located within beta-trefoil domain (Hcbetatre), we hypothesize that immunization with the Hcbetatre domain is sufficient to confer protective immunity. In addition, enhancing its uptake subsequent to nasal delivery prompted development of an alternative vaccine strategy, and we hypothesize that the addition of targeting moiety adenovirus 2 fiber protein (Ad2F) may enhance such uptake during vaccination. RESULTS: The Hcbetatre serotype B immunogen was genetically fused to Ad2F (Hcbetatre/B-Ad2F), and its immunogenicity was tested in mice. In combination with the mucosal adjuvant, cholera toxin (CT), enhanced mucosal IgA and serum IgG Ab titers were induced by nasal Hcbetatre-Ad2F relative to Hcbetatre alone; however, similar Ab titers were obtained upon intramuscular immunization. These BoNT/B-specific Abs induced by nasal immunization were generally supported in large part by Th2 cells, as opposed to Hcbetatre-immunized mice that showed more mixed Th1 and Th2 cells. Using a mouse neutralization assay, sera from animals immunized with Hcbetatre and Hcbetatre-Ad2F protected mice against 2.0 LD50. CONCLUSION: These results demonstrate that Hcbetatre-based immunogens are highly immunogenic, especially when genetically fused to Ad2F, and Ad2F can be exploited as a vaccine delivery platform to the mucosa.

Characterization of autonomous thyroid adenoma: metabolism, gene expression, and pathology.

Fifty-one in vivo characterized autonomous single adenomas have been studied for functional parameters in vitro, for gene and protein expression and for pathology, and have been systematically compared to the corresponding extratumoral quiescent tissue. The adenomas were characterized by a high level of iodide trapping that corresponds to a high level of Na+ /iodide symporter gene expression, a high thyroperoxidase mRNA and protein content, and a low H2O2 generation. This explains the iodide metabolism characteristics demonstrated before, ie, the main cause of the "hot" character of the adenomas is their increased iodide transport. The adenomas spontaneously secreted higher amounts of thyroid hormone than the quiescent tissue and in agreement with previous in vivo data, this secretion could be further enhanced by thyrotropin (TSH). Inositol uptake was also increased but there was no spontaneous increase of the generation of inositol phosphates and this metabolism could be further activated by TSH. These positive responses to TSH are in agreement with the properties of TSH-stimulated thyroid cells in vitro and in vivo. They are compatible with the characteristics of mutated TSH receptors whose constitutive activation accounts for the majority of autonomous thyroid adenomas in Europe. The number of cycling cells, as evaluated by MIB-1 immunolabeling was low but increased in comparison with the corresponding quiescent tissue or normal tissue. The cycling cells are observed mainly at the periphery; there was very little apoptosis. Both findings account for the slow growth of these established adenomas. On the other hand, by thyroperoxidase immunohistochemistry, the whole lesion appeared hyperfunctional, which demonstrates a dissociation of mitogenic and functional stimulations. Thyroglobulin, TSH receptor, and E-cadherin mRNA accumulations were not modified in a consistent way, which confirms the near-constitutive expression of the corresponding genes in normal differentiated tissue. On the contrary, early immediate genes expressions (c-myc, NGF1B, egr 1, genes of the fos and jun families) were decreased. This may be explained by the proliferative heterogeneity of the lesion and the previously described short, biphasic expression of these genes when induced by mitogenic agents. All the characteristics of the autonomous adenomas can therefore be explained by the effect of the known activating mutations of genes coding for proteins of the TSH cyclic adenosine monophosphate (cAMP) cascade, all cells being functionally activated while only those at the periphery multiply. The reason of this heterogeneity is unknown.

Multilevel refinement for the vehicle routing problem

Multilevel approaches to computational problems are pervasive across many areas of applied mathematics and scientific computing. The multilevel paradigm uses recursive coarsening to create a hierarchy of approximations to the original problem, then an initial solution is found for the coarsest problem and iteratively refined and improved at each level, coarsest to finest. The solution process is aided by the global perspective (or `global view') imparted to the optimisation by the coarsening. This paper looks at their application to the Vehicle Routing Problem.

A study of modelling approaches for rail vehicle collision behaviour

A rigid wall model has been used widely in the numerical simulation of rail vehicle impacts. Finite element impact modelling of rail vehicles is generally based on a half-width and full-length or half-length structure, depending on the symmetry. The structure and components of rail vehicles are normally designed to cope with proof loading to ensure adequate ride performance. In this paper, the authors present a study of a rail vehicle with driving cab focused on improving the modelling approach and exploring the intrinsic structural weaknesses to enhance its crashworthiness. The underpinning research used finite element analysis and compared the behaviour of the rail vehicle in different impact scenarios. It was found that the simulation of a rigid wall impact can mask structural weaknesses; that even a completely symmetrical impact may lead to an asymmetrical result; that downward bending is an intrinsic weakness of conventional rail vehicles and that a rigid part of the vehicle structure, such as the body bolster, may cause uncoordinated deformation and shear fracture between the vehicle sections. These findings have significance for impact simulation, the full-scale testing of rail vehicles and rail vehicle design in general.

Multilevel refinement strategies for the capacity vehicle routing problem

We discuss the application of the multilevel (ML) refinement technique to the Vehicle Routing Problem (VRP), and compare it to its single-level (SL) counterpart. Multilevel refinement recursively coarsens to create a hierarchy of approximations to the problem and refines at each level. A SL algorithm, which uses a combination of standard VRP heuristics, is developed first to solve instances of the VRP. A ML version, which extends the global view of these heuristics, is then created, using variants of the construction and improvement heuristics at each level. Finally some multilevel enhancements are developed. Experimentation is used to find suitable parameter settings and the final version is tested on two well-known VRP benchmark suites. Results comparing both SL and ML algorithms are presented.

The application of multilevel refinement to the vehicle routing problem

We discuss the application of the multilevel (ML) refinement technique to the Vehicle Routing Problem (VRP), and compare it to its single-level (SL) counterpart. Multilevel refinement recursively coarsens to create a hierarchy of approximations to the problem and refines at each level. A SL heuristic, termed the combined node-exchange composite heuristic (CNCH), is developed first to solve instances of the VRP. A ML version (the ML-CNCH) is then created, using the construction and improvement heuristics of the CNCH at each level. Experimentation is used to find a suitable combination, which extends the global view of these heuristics. Results comparing both SL and ML are presented.

Analysis of the structural characteristics of an intermediate rail vehicle and their effect on vehicle crash performance

In the current paper, the authors present an analysis of the structural characteristics of an intermediate rail vehicle and their effects on crash performance of the vehicle. Theirs is a simulation based analysis involving four stages. First, the crashworthiness of the vehicle is assessed by simulating an impact of the vehicle with a rigid wall. Second, the structural characteristics of the vehicle are analysed based on the structural behaviour during this impact and then the structure is modified. Third, the modified vehicle is tested again in the same impact scenario with a rigid wall. Finally, the modified vehicle is subjected to a modelled head-on impact which mirrors the real-life impact interface between two intermediate vehicles in a train impact. The emphasis of the current study is on the structural characteristics of the intermediate vehicle and the differences compared to an impact of a leading vehicle. The study shows that, similar to a leading vehicle, bending, or jackknifing is a main form of failure in this conventionally designed intermediate vehicle. It has also been found that the location of the door openings creates a major difference in the behaviour of an intermediate vehicle. It causes instability of the vehicle in the door area and leads to high stresses at the joint of the end beam with the solebar and shear stresses at the joint of the inner pillar with the cantrail. Apart from this, the shapes of the vehicle ends and impact interfaces are also different and have an effect on the crash performance of the vehicles. The simulation results allow the identification of the structural characteristics and show the effectiveness of relevant modifications. The conclusions have general relevance for the crashworthiness of rail vehicle design

'Re-using an existing wheel' developing data architecture for cooperating autonomous and semi-autonomous, agent-based web services

This short position paper considers issues in developing Data Architecture for the Internet of Things (IoT) through the medium of an exemplar project, Domain Expertise Capture in Authoring and Development ­Environments (DECADE). A brief discussion sets the background for IoT, and the development of the ­distinction between things and computers. The paper makes a strong argument to avoid reinvention of the wheel, and to reuse approaches to distributed heterogeneous data architectures and the lessons learned from that work, and apply them to this situation. DECADE requires an autonomous recording system, ­local data storage, semi-autonomous verification model, sign-off mechanism, qualitative and ­quantitative ­analysis ­carried out when and where required through web-service architecture, based on ontology and analytic agents, with a self-maintaining ontology model. To develop this, we describe a web-service ­architecture, ­combining a distributed data warehouse, web services for analysis agents, ontology agents and a ­verification engine, with a centrally verified outcome database maintained by certifying body for qualification/­professional status.