Ivanov, Radoslav
Email Address
ORCID
Disciplines
Search Results
Now showing 1 - 10 of 16
Publication Estimation of Blood Oxygen Content Using Context-Aware Filtering(2016-04-01) Ivanov, Radoslav; Atanasov, Nikolay; Weimer, James; Simpao, Allan F; Rehman, Mohamed A; Pappas, George; Lee, Insup; Pajic, MiroslavIn this paper we address the problem of estimating the blood oxygen concentration in children during surgery.Currently, the oxygen content can only be measured through invasive means such as drawing blood from the patient. In this work, we attempt to perform estimation by only using other non-invasive measurements (e.g., fraction of oxygen in inspired air, volume of inspired air) collected during surgery. Although models mapping these measurements to blood oxygen content contain multiple parameters that vary widely across patients, the non-invasive measurements can be used to provide binary information about whether the oxygen concentration is rising or dropping. This information can then be incorporated in a context-aware filter that is used to combine regular continuous measurements with discrete detection events in order to improve estimation. We evaluate the filter using real-patient data collected over the last decade at the Children’s Hospital of Philadelphia and show that it is a promising approach for the estimation of unobservable physiological variables.Publication Robust Estimation Using Context-Aware Filtering(2015-09-01) Ivanov, Radoslav; Atanasov, Nikolay; Pappas, George; Lee, Insup; Pajic, MiroslavThis paper presents the context-aware filter, an estimation technique that incorporates context measurements, in addition to the regular continuous measurements. Context measurements provide binary information about the system’s context which is not directly encoded in the state; examples include a robot detecting a nearby building using image processing or a medical device alarming that a vital sign has exceeded a predefined threshold. These measurements can only be received from certain states and can therefore be modeled as a function of the system’s current state. We focus on two classes of functions describing the probability of context detection given the current state; these functions capture a wide variety of detections that may occur in practice. We derive the corresponding context-aware filters, a Gaussian Mixture filter and another closed-form filter with a posterior distribution whose moments are derived in the paper. Finally, we evaluate the performance of both classes of functions through simulation of an unmanned ground vehicle.Publication ModelGuard: Runtime Validation of Lipschitz-continuous Models(2021-07-01) Carpenter, Taylor J.; Ivanov, Radoslav; Lee, Insup; Weimer, JamesThis paper presents ModelGuard, a sampling-based approach to runtime model validation for Lipschitz-continuous models. Although techniques exist for the validation of many classes of models, the majority of these methods cannot be applied to the whole of Lipschitz-continuous models, which includes neural network models. Additionally, existing techniques generally consider only white-box models. By taking a sampling-based approach, we can address black-box models, represented only by an input-output relationship and a Lipschitz constant. We show that by randomly sampling from a parameter space and evaluating the model, it is possible to guarantee the correctness of traces labeled consistent and provide a confidence on the correctness of traces labeled inconsistent. We evaluate the applicability and scalability of ModelGuard in three case studies, including a physical platform.Publication Attack-Resilient Sensor Fusion(2014-02-01) Ivanov, Radoslav; Pajic, Miroslav; Lee, InsupThis work considers the problem of attack-resilient sensor fusion in an autonomous system where multiple sensors measure the same physical variable. A malicious attacker may corrupt a subset of these sensors and send wrong measurements to the controller on their behalf, potentially compromising the safety of the system. We formalize the goals and constraints of such an attacker who also wants to avoid detection by the system. We argue that the attacker’s capabilities depend on the amount of information she has about the correct sensors’ measurements. In the presence of a shared bus where messages are broadcast to all components connected to the network, the attacker may consider all other measurements before sending her own in order to achieve maximal impact. Consequently, we investigate effects of communication schedules on sensor fusion performance. We provide worst- and average-case results in support of the Ascending schedule, where sensors send their measurements in a fixed succession based on their precision, starting from the most precise sensors. Finally, we provide a case study to illustrate the use of this approach.Publication Robust Localization Using Context-Aware Filtering(2015-07-01) Ivanov, Radoslav; Atanasov, Nikolay; Pajic, Miroslav; Lee, Insup; Pappas, GeorgeIn this paper we develop a robot localization technique that incorporates discrete context measurements, in addition to standard continuous state measurements. Context measurements provide binary information about detected events in the robot’s environment, e.g., a building is recognized using image processing or a known radio station is received. Such measurements can only be detected from certain positions and can, therefore, be correlated with the robot’s state. We investigate two specific examples where context measurements are especially useful – an urban localization scenario where GPS measurements are not reliable as well as the capture of the RQ-170 Sentinel drone in Iran, where GPS measurements were spoofed. By focusing on a specific class of probability of context detection functions, we derive a closed-form Gaussian mixture filter that is precise, captures context, and has the theoretical properties of the Kalman filter. Finally, we provide simulations of the urban localization scenario with an unmanned ground vehicle and show that the proposed context-aware filter is more robust and more accurate than the conventional extended Kalman filter, which only uses continuous measurements.Publication Verisig 2.0: Verification of Neural Network Controllers Using Taylor Model Preconditioning(2021-07-01) Ivanov, Radoslav; Carpenter, Taylor; Weimer, James; Alur, Rajeev; Pappas, George; Lee, InsupThis paper presents Verisig 2.0, a verification tool for closed-loop systems with neural network (NN) controllers. We focus on NNs with tanh/sigmoid activations and develop a Taylor-model-based reachability algorithm through Taylor model preconditioning and shrink wrapping. Furthermore, we provide a parallelized implementation that allows Verisig 2.0 to efficiently handle larger NNs than existing tools can. We provide an extensive evaluation over 10 benchmarks and compare Verisig 2.0 against three state-of-the-art verification tools. We show that Verisig 2.0 is both more accurate and faster, achieving speed-ups of up to 21x and 268x against different tools, respectively.Publication Adaptive Transient Fault Model for Sensor Attack Detection(2016-10-01) Park, Junkil; Ivanov, Radoslav; Jo, Minsu; Weimer, James; Baek, Youngmi; Lee, Insup; Son, Sang HyukThis paper considers the problem of sensor attack detection for multiple operating mode systems, building upon an existing attack detection method that uses a transient fault model with fixed parameters. For a multiple operating mode system, the existing method would have to use the most conservative model parameters to preserve the soundness in attack detection, thus not being effective in attack detection for some operating modes. To address this problem, we propose an adaptive transient fault model to use the appropriate parameter values in accordance with the change of the operating mode of the system. The benefit of our proposed system is demonstrated using real measurement data obtained from an unmanned ground vehicle.Publication Security of Cyber-Physical Systems in the Presence of Transient Sensor Faults(2017-07-01) Park, Junkil; Ivanov, Radoslav; Weimer, James; Lee, Insup; Pajic, Miroslav; Son, Sang HyukThis paper is concerned with the security of modern Cyber-Physical Systems in the presence of transient sensor faults. We consider a system with multiple sensors measuring the same physical variable, where each sensor provides an interval with all possible values of the true state. We note that some sensors might output faulty readings and others may be controlled by a malicious attacker. Different from previous works, in this paper we aim to distinguish between faults and attacks and develop an attack detection algorithm for the latter only. To do this, we note that there are two kinds of faults – transient and permanent; the former are benign and short-lived whereas the latter may have dangerous consequences on system performance.We argue that sensors have an underlying transient fault model that quantifies the amount of time in which transient faults can occur. In addition, we provide a framework for developing such a model if it is not provided by manufacturers. Attacks can manifest as either transient or permanent faults depending on the attacker’s goal. We provide different techniques for handling each kind. For the former, we analyze the worst-case performance of sensor fusion over time given each sensor’s transient fault model and develop a filtered fusion interval that is guaranteed to contain the true value and is bounded in size. To deal with attacks that do not comply with sensors’ transient fault models, we propose a sound attack detection algorithm based on pairwise inconsistencies between sensor measurements. Finally, we provide a real-data case study on an unmanned ground vehicle to evaluate the various aspects of this paper.Publication Early Detection of Critical Pulmonary Shunts in Infants(2015-04-14) Ivanov, Radoslav; Weimer, James; Simpao, Allan F; Rehman, Mohamed A; Lee, InsupThis paper aims to improve the design of modern Medical Cyber Physical Systems through the addition of supplemental noninvasive monitors. Specifically, we focus on monitoring the arterial blood oxygen content (CaO2), one of the most closely observed vital signs in operating rooms, currently measured by a proxy - peripheral hemoglobin oxygen saturation (SpO2). While SpO2 is a good estimate of O2 content in the finger where it is measured, it is a delayed measure of its content in the arteries. In addition, it does not incorporate system dynamics and is a poor predictor of future CaO2 values. Therefore, as a first step towards supplementing the usage of SpO2, this work introduces a predictive monitor designed to provide early detection of critical drops in CaO2 caused by a pulmonary shunt in infants. To this end, we develop a formal model of the circulation of oxygen and carbon dioxide in the body, characterized by unknown patient-unique parameters. Employing the model, we design a matched subspace detector to provide a near constant false alarm rate invariant to these parameters and modeling uncertainties. Finally, we validate our approach on real-patient data from lung lobectomy surgeries performed at the Children's Hospital of Philadelphia. Given 198 infants, the detector predicted 81% of the critical drops in CaO2 at an average of about 65 seconds earlier than the SpO2-based monitor, while achieving a 0:9% false alarm rate (representing about 2 false alarms per hour).Publication Parameter-Invariant Design of Medical Alarms(2015-10-01) Weimer, James; Ivanov, Radoslav; Roederer, Alexander; Chen, Sanjian; Lee, InsupThe recent explosion of low-power low-cost communication, sensing, and actuation technologies has ignited the automation of medical diagnostics and care in the form of medical cyber physical systems (MCPS). MCPS are poised to revolutionize patient care by providing smarter alarm systems, clinical decision support, advanced diagnostics, minimally invasive surgical care, improved patient drug delivery, and safety and performance guarantees. With the MCPS revolution emerges a new era in medical alarm systems, where measurements gathered via multiple devices are fused to provide early detection of critical conditions. The alarms generated by these next generation monitors can be exploited by MCPS to further improve performance, reliability, and safety. Currently, there exist several approaches to designing medical monitors ranging from simple sensor thresholding techniques to more complex machine learning approaches. While all the current design approaches have different strengths and weaknesses, their performance degrades when underlying models contain unknown parameters and training data is scarce. Under this scenario, an alternative approach that performs well is the parameter-invariant detector, which utilizes sufficient statistics that are invariant to unknown parameters to achieve a constant false alarm rate across different systems. Parameter-invariant detectors have been successfully applied in other cyber physical systems (CPS) applications with structured dynamics and unknown parameters such as networked systems, smart buildings, and smart grids; most recently, the parameter-invariant approach has been recently extended to medical alarms in the form of a critical shunt detector for infants undergoing a lung lobectomy. The clinical success of this case study application of the parameter-invariant approach is paving the way for a range of other medical monitors. In this tutorial, we present a design methodology for medical parameter-invariant monitors. We begin by providing a motivational review of currently employed medical alarm techniques, followed by the introduction of the parameter-invariant design approach. Finally, we present a case study example to demonstrate the design of a parameter-invariant alarm for critical shunt detection in infants during surgical procedures.