A reservoir-driven non-stationary hidden Markov model

In this work, we propose a novel approach towards sequential data modeling that leverages the strengths of hidden Markov models and echo-state networks (ESNs) in the context of non-parametric Bayesian inference approaches. We introduce a non-stationary hidden Markov model, the time-dependent state transition probabilities of which are driven by a high-dimensional signal that encodes the whole history of the modeled observations, namely the state vector of a postulated observations-driven ESN reservoir. We derive an efficient inference algorithm for our model under the variational Bayesian paradigm, and we examine the efficacy of our approach considering a number of sequential data modeling applications.     

A nonparametric Bayesian approach toward robot learning by demonstration

In the past years, many authors have considered application of machine learning methodologies to effect robot learning by demonstration. Gaussian mixture regression (GMR) is one of the most successful methodologies used for this purpose. A major limitation of GMR models concerns automatic selection of the proper number of model states, i.e., the number of model component densities. Existing methods, including likelihood- or entropy-based criteria, usually tend to yield noisy model size estimates while imposing heavy computational requirements. Recently, Dirichlet process (infinite) mixture models have emerged in the cornerstone of nonparametric Bayesian statistics as promising candidates for clustering applications where the number of clusters is unknown a priori. Under this motivation, to resolve the aforementioned issues of GMR-based methods for robot learning by demonstration, in this paper we introduce a nonparametric Bayesian formulation for the GMR model, the Dirichlet process GMR model. We derive an efficient variational Bayesian inference algorithm for the proposed model, and we experimentally investigate its efficacy as a robot learning by demonstration methodology, considering a number of demanding robot learning by demonstration scenarios.     

Dual arm manipulation—A survey

Recent advances in both anthropomorphic robots and bimanual industrial manipulators had led to an increased interest in the specific problems pertaining to dual arm manipulation. For the future, we foresee robots performing human-like tasks in both domestic and industrial settings. It is therefore natural to study specifics of dual arm manipulation in humans and methods for using the resulting knowledge in robot control. The related scientific problems range from low-level control to high level task planning and execution. This review aims to summarize the current state of the art from the heterogenous range of fields that study the different aspects of these problems specifically in dual arm manipulation.     

A spatially-constrained normalized Gamma process prior

In this work, we propose a novel nonparametric Bayesian method for clustering of data with spatial interdependencies. Specifically, we devise a novel normalized Gamma process, regulated by a simplified (pointwise) Markov random field (Gibbsian) distribution with a countably infinite number of states. As a result of its construction, the proposed model allows for introducing spatial dependencies in the clustering mechanics of the normalized Gamma process, thus yielding a novel nonparametric Bayesian method for spatial data clustering. We derive an efficient truncated variational Bayesian algorithm for model inference. We examine the efficacy of our approach by considering an image segmentation application using a real-world dataset. We show that our approach outperforms related methods from the field of Bayesian nonparametrics, including the infinite hidden Markov random field model, and the Dirichlet process prior.     

Diffuse Surface Water Pollution: Driving Factors for Different Geoclimatic Regions

The quantification of point and non-point losses of Nitrogen (N) and Phosphorus (P) to surface waters is currently a major issue for the implementation of Environmental Directives, such as the Water Framework Directive in Europe. However, the drivers behind nutrient pollution are location specific and are affected by regional hydroclimatic and geomorphological characteristics. In this study the river basin model SWAT was used in order to improve the process-based understanding of mechanisms behind nutrient transport from land to water recipients in two European catchments with significantly different meteorological conditions: the Greek catchment of Ali Efenti, representative of the Mediterranean climate, and the Norwegian catchment of Vansjø-Hobølv, representative of the cold climate typical in Scandinavia. The models were firstly calibrated according to measured river flows and nutrient loads, and then four Best Management Practices (BMPs), related to nutrient application and soil management were applied in order to examine their effectiveness under the different geoclimatic conditions of the two catchments. The results of the baseline indicated that diffuse agricultural sources were the largest contributor of N losses in both catchments and of P losses in the Greek catchment, while point sources were significant contributors to P levels in Norwegian rivers. Nutrient losses to surface waters in Ali Efenti exhibited high seasonal variation, attributed to the extremeness of precipitation events that is typical in the Mediterranean, as well as to the temporal distribution of sediment losses to waters. On the other hand, in Scandinavia, the losses of N and P occurred with less deviation throughout the year and independently of the freezing of soils. The values of the calibrated parameters that mainly governed the hydrological and erosion processes in the catchments demonstrated the natural driving forces of nutrient losses to waters and their temporal distribution indicating that these forces are also crucial in determining the appropriate implementation of agricultural management practices in various geoclimatic regions.     

Can simulation technology enable a paradigm shift in process control?: Modeling for the rest of us

Commercial process simulation software makes it easy for experts to develop very complex models with thousands of equations. But how well are these models used? Remember the admonition of Box [Box, G. E. P. (1979). Robustness in scientific model building. In R. L. Launer & G. N. Wilkinson (Eds.), Robustness in statistics (pp. 201–236). New York: Academic Press], “All models are wrong, but some are useful”. Are case runs just captured in a report and then filed away? Is the expert the only one who can run additional cases? We believe that process dynamics simulation should be ubiquitous in chemical engineering practice and education. Undergraduate engineers should experience unit operations through a virtual process simulator. In industry, engineers must be able to quickly build dynamic models to study operability and design control strategies. We feel that DuPont has undergone a paradigm shift where engineers are much more likely to use dynamic simulation as part of their day-to-day work. This paper illustrates some of the features that process dynamic simulators need to enable this paradigm shift.     

Partial oxidation of methane to synthesis gas via the direct reaction scheme over Ru/TiO2 catalyst

The partial oxidation of methane to synthesis gas has been investigated over various supported metal catalysts. The effects of operational variables on mass and heat transport resistances were investigated for defining the kinetic regime. It is observed that, in the absence of significant mass and heat transfer resistances, high selectivity (up to 65%) to synthesis gas is obtained over Ru/TiO₂ catalysts in the low methane conversion range ( ) whereas only negligibly small selectivity to synthesis gas is observed over all other catalysts investigated under similar conditions. This indicates that the Ru/TiO₂ catalyst possesses unique properties, offering high selectivity to synthesis gas formation via the direct reaction scheme, whereas the other catalysts promote the sequence of total oxidation of methane to CO₂ and H₂O, followed by reforming reactions to synthesis gas. An increase of selectivity to synthesis gas, in the presence of oxygen, is achieved over the Ru/TiO₂ catalyst by multi-feeding oxygen, which is attributed to suppression of deep oxidation of H₂ and CO.     

Simulating the seismic response of embankments via artificial neural networks

Geotechnical earthquake engineering may generally be considered as an “imprecise” scientific area due to the unavoidable uncertainties and the simplifications adopted during the design process of geostructures. Therefore, relatively accurate predictions using advanced soft computing (SC) techniques can be tolerated rather than solving a problem conventionally. Artificial neural networks (ANNs), being one of the most popular SC techniques, have been used in many fields of science and technology, as well as, into an increasing number of earthquake engineering applications on structures and infrastructures. In this work the implementation of ANNs is focused on the simulation of the seismic response of a typical embankment. The dynamic response of the embankment is evaluated utilizing the finite-element method, where the nonlinear behavior of the geo-materials can be taken into account by an equivalent-linear procedure. In the present study, this extremely time-consuming process is replaced by properly trained ANNs.     

Spatially distributed area coverage optimisation in mobile robotic networks with arbitrary convex anisotropic patterns

Coordination of the mobile nodes for area coverage applications in a distributed wireless sensor network is examined in this article, where the nodes’ sensing domains are arbitrary strictly convex compact sets, though identical among each other as far as shape and orientation are concerned. Unlike works that utilise standard Voronoi tessellation for coordination of homogeneous mobile nodes with circular sensing patterns, in this article the planar space is partitioned via an innovative way into subsets based only on the nodes’ coverage footprints, rather than the nodes’ coordinates. The proposed control scheme is spatially distributed, leading to a locally area-optimal network configuration, while connectivity issues are analysed via worst case scenario topology. Efficiency of the proposed scheme is further confirmed via comparative simulation studies.