Sahika Genc

Predictability of event occurrences in partially-observed discrete-event systems

This paper studies the problem of predicting occurrences of a significant event in a partially-observed discrete-event system. The predictability of occurrences of an event in a system is defined in the context of formal languages. The predictability of a language is a stronger condition than the diagnosability of the language. Two necessary and sufficient conditions for predictability of occurrences of an event in systems modeled by regular languages are presented. Both conditions can be algorithmically tested. The first condition employs diagnosers. The second condition employs verifiers and results in a polynomial-time (in the number of states) complexity test for verification of predictability. When predictability holds, diagnosers can be used online to predict the significant event.

Predictability of sequence patterns in discrete event systems

Abstract The problem of predicting the occurrences of a pattern in a partially-observed discrete-event system is studied. The system is modeled by a labeled transition system. The pattern is a set of event sequences modeled by a finite-state automaton. The occurrences of the pattern are predictable if it is possible to infer about any occurrence of the pattern before the pattern is completely executed by the system. A novel off-line algorithm to verify the property of predictability is presented. The verification is polynomial in the number of states of the system. An on-line algorithm to track the execution of the pattern during the operation of the system is also presented. This algorithm is based on the use of a diagnoser automaton.

Gait characterization via pulse-Doppler radar

Falls are a major cause of injury in the elderly with almost 1/3rd of people aged 65 and more falling each year [1]. This work aims to use gait measurements from everyday living environments to estimate risk of falling and enable improved interventions. For this purpose, we consider the use of low-cost pulse-Doppler range control radar. These radars can continuously acquire data during normal activity of a person in night and day conditions and even in the presence of obstructing furniture. A short-time Fourier transform of the radar data reveals unique Doppler signatures from the torso motion and the leg swings. Two algorithms that can extract these features from the radar spectrogram are proposed in this study for estimating gait velocity and stride durations. The performance of the proposed radar system is evaluated with experimental data, which consists of 9 different walk types and a total of 27 separate tests. A high accuracy motion-capture camera system has also been used to acquire data simultaneously with the radar and provides the ground truth reference. Results indicate that the proposed radar system is a viable candidate for gait characterization and can be used to accurately track mean gait velocity, mean stride duration and stride duration variability. The gait velocity variability can also be estimated but with relatively larger error levels.

A coordinated optimization approach to Volt/VAr control for large power distribution networks

Electric distribution networks are operated under a number of constraints in order to deliver power at a certain quality and reliability level. A distributed management system (DMS) is a supervisory control layer in the distribution system used by the utilities for managing distribution assets in a coordinated fashion. For large distribution systems (those consisting of thousands of nodes and multiple tens of capacitor banks and voltage regulators), an integrated Volt/VAr Control (IVVC), which maximizes asset lifetime, is non-trivial due to the size of the search space for determining the optimal settings of these devices. This paper presents coordinated optimization approach to IVVC for large power distribution networks that will enable a more optimal operation of the distribution network while maximizing distribution control asset lifetime through the minimization of unnecessary device switching.

An early respiratory distress detection method with Markov models.

A method for early detection of respiratory distress in hospitalized patients which is based on a multi-parametric analysis of respiration rate (RR) and pulse oximetry (SpO2) data trends to ascertain patterns of patient instability pertaining to respiratory distress is described. Current practices of triggering caregiver alerts are based on simple numeric threshold breaches of SpO2. The pathophysiological patterns of respiratory distress leading to in-hospital deaths are much more complex to be detected by numeric thresholds. Our pattern detection algorithm is based on a Markov model framework based on multi-parameter pathophysiological patterns of respiratory distress, and triggers in a timely manner and prior to the violation of SpO2 85-90% threshold, providing additional lead time to attempt to reverse the deteriorating state of the patient. We present the performance of the algorithm on MIMIC II dataset resulting in true positive rate of 92% and false positive rate of 6%.

Model predictive building thermostatic controls of small-to-medium commercial buildings for optimal peak load reduction incorporating dynamic human comfort models: Algorithm and implementation

The peak kW of a typical New York State office building is thought to primarily be a function of the HVAC system, often the buildings largest load, but may also be influenced by occupancy and other loads. First, a simple lumped parameter model with a minimum amount of buildings physical input data, and trained with actual thermal and electrical data, is considered to approximate the thermal/electric consumption performance of the building and HVAC system on a zonal basis. Then, the lumped parameter model integrated with a dynamic human comfort model is used to develop optimized zonal thermostat setpoint schedules to minimize the cooling systems contribution to the buildings peak power load while maintaining human comfort at a desired level. A 24-hour weather and occupancy forecasts are also incorporated into the optimization algorithm. The key difference of our approach compared to previous approaches that utilize model-predictive control is that a minimal set of measurement profiles are utilized to reduce the installation cost resulting in a cost effective advanced controls solution for a large number of small and medium size office buildings. The model predictive optimization approach is implemented at multiple demonstration sites. The hardware architecture and software platform installed at one of the demonstration buildings are discussed. Finally, it is demonstrated that the proposed controller can effectively minimize peak cooling load on the HVAC equipment while achieving a satisfactory thermal comfort inside the building.

Prediction of mortality from respiratory distress among long-term mechanically ventilated patients.

With the advent of inexpensive storage, pervasive networking, and wireless devices, it is now possible to store a large proportion of the medical data that is collected in the intensive care unit (ICU). These data sets can be used as valuable resources for developing and validating predictive analytics. In this report, we focus on the problem of prediction of mortality from respiratory distress among long-term mechanically ventilated patients using data from the publicly-available MIMIC-II database. Rather than only reporting p-values for univariate or multivariate regression, as in previous work, we seek to generate sparsest possible model that will predict mortality. We find that the presence of severe sepsis is highly associated with mortality. We also find that variables related to respiration rate have more predictive accuracy than variables related to oxygenation status. Ultimately, we have developed a model which predicts mortality from respiratory distress in the ICU with a cross-validated area-under-the-curve (AUC) of approximately 0.74. Four methodologies are utilized for model dimensionality-reduction: univariate logistic regression, multivariate logistic regression, decision trees, and penalized logistic regression.

Hemodynamic-impact-based prioritization of ventricular tachycardia alarms.

Ventricular tachycardia (V-tach) is a very serious condition that occurs when the ventricles are driven at high rates. The abnormal excitation pathways make ventricular contraction less synchronous resulting in less effective filling and emptying of the left ventricles. However, almost half of the V-tach alarms declared through processing of patterns observed in electrocardiography are not clinically actionable. The focus of this study is to provide guidance on determining whether a technically-correct V-tach alarm is clinically-actionable by determining its “hemodynamic impact”. A supervisory learning approach based on conditional inference trees to determine the hemodynamic impact of a V-tach alarm based on extracted features is described. According to preliminary results on a subset of Multiparameter intelligent monitoring in intensive care II (MIMIC-II) database, true positive rate of more than 90% can be achieved.

Distributed estimation of lumped parameters of multi-zone small-middle size commercial buildings with minimal observations & implementation

Installation of an energy management system (EMS) that could minimize the buildings peak load can be impractical, having a long and uncertain return on investment. Identification of building thermal lumped parameters 1) using only zone and supply air temperature measurements for constant- and variable-air volume HVAC equipment, and 2) a distributed algorithm for buildings with large number of zones is discussed for inexpensive thermostatic control retrofits as alternative to traditional supervisory Heat, Ventilation, and Air-Conditioning (HVAC) controller. The challenges in implementation of the proposed approach at two New York State buildings for demonstration and analysis are described: One building which we retrofitted with low-cost off-the-shelf wireless sensors and another building which was recently renovated and fitted with a new HVAC system with the state-of-art supervisory controllers. The differences in simulation-based (e.g., TrnSysTM) analysis and on-site implementation are highlighted. Finally, the results and performance of the proposed distributed estimation algorithm on the actual data from demonstration buildings are provided.

Prediction of mean arterial blood pressure with linear stochastic models

A model-based approach that integrates known portion of the cardiovascular system and unknown portion through a parameter estimation to predict evolution of the mean arterial pressure is considered. The unknown portion corresponds to the neural portion that acts like a controller that takes corrective actions to regulate the arterial blood pressure at a constant level. The input to the neural part is the arterial pressure and output is the sympathetic nerve activity. In this model, heart rate is considered a proxy for sympathetic nerve activity. The neural portion is modeled as a linear discrete-time system with random coefficients. The performance of the model is tested on a case study of acute hypotensive episodes (AHEs) on PhysioNet data. TPRs and FPRs improve as more data becomes available during estimation period.