Dimitar Baronov

Autonomous vehicle control for ascending/descending along a potential field with two applications

In this paper, we present a reactive control law that can navigate a single, sensor-enabled vehicle to ascend or descend a scalar potential field. The design builds on our previous work on developing an online following control. The model framework enables us to establish performance bounds for the developed ascending/descending control that are related to the geometrical parameters of the field. The efficacy of our approach is demonstrated through two possible applications - source-centric mapping of a potential field and tracking of a single target, emitting a distance-related potential field.

Reactive exploration through following isolines in a potential field

In this paper we propose a control law aimed at tracing level curves (isolines) in a scalar potential field. An exploration agent governed by such a law can map simply connected regions in space where the potential field exceeds a predefined threshold. The distinguishing feature of our control is that it does not rely on higher order characteristics of the field such as the gradient at a point or the curvature of the isolines, in contrast with these parameters appearing as inputs in other proposed control laws [1], [2]. Furthermore, we establish relationships between the performance of our control law and the geometry of the isoline and show results from implementing the algorithm in both a simulated environment and a testbed.

Search decisions for teams of automata

The dynamics of exploration vs exploitation decisions are explored in the context of robotic search problems. Building on prior work on robotic search together with our own work on reactive control laws for potential field mapping, we propose a new set of search protocols for teams of sensor-enabled mobile robots. The focus is on collaborative strategies for the search of potential fields that are possibly time varying. We pose the problem of quickly finding regions where the potential achieves or exceeds a certain threshold. The search protocol has two distinct components. In an ¿exploration phase¿, agents execute either a randomized or structured search, seeking places where the field achieves or exceeds the prescribed threshold. Once a threshold point is reached, the ¿exploitation¿ component is initialized and the agents deploy so as to rapidly map the evolving isoline associated with the given value of the field. Conservative strategies will emphasize refining the detailed knowledge of the field in a small neighborhood of the isoline, while aggressive strategies will emphasize wide-ranging exploration of neighboring territory. The main decision problem under study involves finding the optimally aggressive exploration strategy. Additionally, the problem of the allocation of the agents between ¿exploration¿ and ¿exploitation¿ is considered. A performance metric is developed to compare the proposed methods with standard approaches such as random search and distributed raster scans.

Controlling a Magnetic Force Microscope to Track a Magnetized Nanosize Particle

In this paper, we introduce a scheme for tracking a magnetic nanoparticle moving in three dimensions using a magnetic force microscope (MFM). The stray magnetic field of the magnetic particle induces a shift in the phase of oscillation of the tip of the MFM. We present a nonlinear feedback control law that translates the measurement of this phase shift into a trajectory for the tip of the MFM and prove that this trajectory converges to a neighborhood of the magnetic particle. The viability of the proposed tracking scheme is verified through numerical simulations of the tracking algorithm.

A Motion Description Language for Robotic Reconnaissance of Unknown Fields

In this paper, we present two motion primitives that allow a mobile sensor to explore the features of an unknown scalar field. The first motion primitive is designed to follow and to map level contours (contours with constant value of the field). The second one steers the sensor to ascend or alternatively descend the field gradient and, as a result, to localize its extremum points. Both of these primitives are defined in terms of the geometric characteristics of the potential function and their performance is analyzed for different ranges of the parameters describing the geometry. The two motion primitives constitute a suitable library for rapid information acquisition aimed at the mapping of unknown fields. The motion control primitives developed below will provide the foundation of a theory of control for information acquisition in the exploration of unknown fields by means of mobile point sensors. This theory is treated in a companion paper.

Decision Making for Rapid Information Acquisition in the Reconnaissance of Random Fields

In this paper, research into several aspects of robot-enabled reconnaissance of random fields is reported. The work has two major components: the underlying theory of information acquisition in the exploration of unknown fields and the results of experiments on how humans use sensor-equipped robots to perform a simulated reconnaissance exercise. The theoretical framework reported herein extends work on robotic exploration that has been reported by ourselves and others. Several new figures of merit for evaluating exploration strategies are proposed and compared. Using concepts from differential topology and information theory, we develop the theoretical foundation of search strategies aimed at rapid discovery of topological features (locations of critical points and critical level sets) of a priori unknown differentiable fields. The theory enables study of efficient reconnaissance strategies in which the tradeoff between speed and accuracy can be understood. The proposed approach to rapid discovery of topological features has led in a natural way to the creation of parsimonious reconnaissance routines that do not rely on any prior knowledge of the environment. The design of topology-guided search protocols uses a mathematical framework that quantifies the relationship between what is discovered and what remains to be discovered. The quantification rests on an information theory inspired model whose properties allow us to treat search as a problem in optimal information acquisition. A central theme in this approach is that “conservative” and “aggressive” search strategies can be precisely defined, and search decisions regarding “exploration” versus “exploitation” choices are informed by the rate at which the information metric is changing. The paper goes on to describe a computer game that has been designed to simulate reconnaissance of unknown fields. Players carry out reconnaissance missions by choosing sequences of motion primitives from two families of control laws that enable mobile robots to either ascend/descend in gradient directions of the field or to map contours of constant field value. The strategies that emerge from the choices of motion sequences are classified in terms of the speed with which information is acquired, the fidelity with which the acquired information represents the entire field, and the extent to which all critical level sets have been approximated. The game thus records each players performance in acquiring information about both the topology and geometry of the unknown fields that have been randomly generated.

Tracking a nanosize magnetic particle using a magnetic force microscope

A scheme for tracking nano-sized magnetic particles using a magnetic force microscope (MFM) is introduced. The stray magnetic field of the particle induces a shift in the phase of the oscillation of the MFM tip. The magnitude of this shift depends on the distance between the tip and the particle and can be expressed as a spatial field. We present a control law which steers the tip to a level set of this field. The approach is based on the previous work of two of the authors on a novel method for mapping unknown potential fields using sensor- enabled mobile robots. Because the method involves geometric properties of the field and its domain, it is not surprising that it can be applied to problems where the characteristic length scales are small. Additionally, we introduce to the original control law an adaptive term to compensate for uncertainties in the parameter values in the model of the magnetic force. The efficacy of this approach is illustrated through simulation. This approach to tracking will provide the capability to investigate the dynamics of single molecules with a higher resolution (in both space and time) than is currently possible.

Tracking a magnetic nanoparticle in 3-D with a magnetic force microscope

In this paper we introduce a scheme for tracking a magnetic nanoparticle in 3-D with the tip of a magnetic force microscope. The stray magnetic field of the magnetic particle induces a shift in the phase of oscillation of the tip of the MFM. We present a feedback control law which translates the measurement of this phase shift into actuator commands to the MFM and prove that the trajectory of the tip converges to a neighborhood of the magnetic particle. This geometric control law depends only on the derivative of the potential along the trajectory of the tip and in particular does not rely on any detailed prior knowledge about the nanoparticle. The results of simulation studies are shown to illustrate the algorithm.

Next generation patient monitor powered by in-silico physiology.

The goal of this paper is to introduce a next generation patient monitoring technology that relies on objective and continuous data analytics to alleviate the data overload in the critical care unit. The technology provides the foundation for increasing the consistency and efficacy of data use in clinical practice and improving outcomes. This paper presents results for applying the approach to the hemodynamic monitoring of infants immediately following cardiac surgery and demonstrates its efficacy of estimating the probability of inadequate systemic oxygen delivery, which is an essential risk attribute in the management of critically ill patients.

Novel risk-based monitoring solution to the data overload in intensive care medicine

Patients in the ICU generate more data than any other clinical environment. Data overload leads to preventable mortality and increased costs. Clinical decision support systems that assist clinicians with interpreting the vast amount of acquired physiologic signals have the potential to save lives and reduce cost. This paper presents a novel methodology which employs physiologic models to translate patient data into actionable risks that are relevant for informed treatment decisions. At the core of the reported technology is a particle-based inference scheme implemented using a Dynamic Bayesian Network that estimates the probabilities of specific pathologies and their causes. The methodology is demonstrated through a pilot study on a post-operative congenital single ventricle population.