Humans who operate in high altitudes for prolonged durations often suffer from hypoxia. The commencement of physiological and cognitive changes due to the onset of hypoxia may not be immediately apparent to the exposed individual. These changes can go unrecognized for minutes and even hours and may lead to serious performance degradation or complete incapacitation. A dynamic system capable of monitoring and detecting decreased physiologic states due to the onset of hypoxia has the potential to prevent adverse outcomes. In this study, we develop a real-time hypoxia monitoring system based on a parallel M -ary decision fusion architecture. Blood oxygen saturation levels and altitude readings are the inputs and estimates of the level of hypoxia are the outputs. We develop new temporal evolution models for blood oxygen saturation and functional impairment with respect to varying altitude. The proposed models enable accurate tracking of various hypoxia levels based on the duration of stay of the subject at an altitude. Using a Bayesian decision-making formulation, the system generates global estimates of the degree of hypoxia. The detection system is tested against synthetic and real datasets to demonstrate applicability and accuracy.

We present a fully automated multi-sperm tracking algorithm. It has the demonstrated capability to detect and track simultaneously hundreds of sperm cells in recorded videos while accurately measuring motility parameters over time and with minimal operator intervention. Algorithms of this kind may help in associating dynamic swimming parameters of human sperm cells with fertility and fertilization rates. Specifically, we offer an image processing method, based on radar tracking algorithms, that detects and tracks automatically the swimming paths of human sperm cells in timelapse microscopy image sequences of the kind that is analyzed by fertility clinics. Adapting the well-known joint probabilistic data association filter (JPDAF), we automatically tracked hundreds of human sperm simultaneously and measured their dynamic swimming parameters over time. Unlike existing CASA instruments, our algorithm has the capability to track sperm swimming in close proximity to each other and during apparent cell-to-cell collisions. Collecting continuously parameters for each sperm tracked without sample dilution (currently impossible using standard CASA systems) provides an opportunity to compare such data with standard fertility rates. The use of our algorithm thus has the potential to free the clinician from having to rely on elaborate motility measurements obtained manually by technicians, speed up semen processing, and provide medical practitioners and researchers with more useful data than are currently available.

With the advancement of genome-wide monitoring technologies, molecular expression data have become widely used for diagnosing cancer through tumor or blood samples. When mining molecular signature data, the process of comparing samples through an adaptive distance function is fundamental but difficult, as such datasets are normally heterogeneous and high dimensional. In this article, we present kernelized information-theoretic metric learning (KITML) algorithms that optimize a distance function to tackle the cancer diagnosis problem and scale to high dimensionality. By learning a nonlinear transformation in the input space implicitly through kernelization, KITML permits efficient optimization, low storage, and improved learning of distance metric. We propose two novel applications of KITML for diagnosing cancer using high-dimensional molecular profiling data: (1) for sample-level cancer diagnosis, the learned metric is used to improve the performance of k -nearest neighbor classification; and (2) for estimating the severity level or stage of a group of samples, we propose a novel set-based ranking approach to extend KITML. For the sample-level cancer classification task, we have evaluated on 14 cancer gene microarray datasets and compared with eight other state-of-the-art approaches. The results show that our approach achieves the best overall performance for the task of molecular-expression-driven cancer sample diagnosis. For the group-level cancer stage estimation, we test the proposed set-KITML approach using three multi-stage cancer microarray datasets, and correctly estimated the stages of sample groups for all three studies.

Reverberation chambers are closed reflective spaces that can emulate highly reverberant electromagnetic environments. The electromagnetic environment is primarily determined by the size of the cavity, effective conductivity, and leakage via apertures. In this effort, we investigate the performance of wireless OFDM communications in relation to the latter two by controlling the loading of a reverberation chamber and the effective aperture into a coupled cavity. A software defined radio measurement platform was used to assess the communication performance through a selection of link-level metrics including error vector magnitude, post processing signal-to-noise ratio, and throughput. The degradation of link quality is quantified for increasingly diffuse environments, as well as the improvement when leveraging a maximal ratio combining receiver diversity scheme. The link quality was found to improve in both the reverberation chamber and the coupled cavity for larger effective apertures. This result was analyzed using a time-dependent model for RF propagation in coupled cavities.

System call analysis is a behavioral malware detection technique that is popular due to its promising detection results and ease of implementation. This study describes a system that uses system call analysis to detect malware that evade traditional defenses. The system monitors executing processes to identify compromised hosts in production environments. Experimental results compare the effectiveness of multiple feature extraction strategies and detectors based on their detection accuracy at low false positive rates. Logistic regression and support vector machines consistently outperform log-likelihood ratio and signature detectors as processing and detection methods. A feature selection study indicates that a relatively small set of system call 3-grams provide detection accuracy comparable to that of more complex models. A case study indicates that the detection system performs well against a variety of malware samples, benign workloads, and host configurations.

Mixed Integer Nonlinear Programming (MINLP) techniques are increasingly used to address challenging problems in robotics, especially Multi-Vehicle Motion Planning (MVMP). A particular challenge in using this framework is encoding stochastic phenomena such as communication connectivity in the form of MINLP constraints. The main contribution of this paper is an analytical formulation of communication connectivity constraints using stochastic physical layer communication models. These constraints account for the log-normal channel shadowing in noisy communication environments and specify inter-vehicle connectivity in terms of the outage probability of communication. A method is developed to provably accord robustness to communication failure by specifying an upper bound on the outage probability in terms of the inter-vehicle communication range. Finally, we demonstrate the utility of this formulation in the context of a realistic decentralized Multi-Vehicle Path Coordination (MVPC) scenario in which multiple robotic vehicles travel along predetermined fixed paths and are required to maintain communication connectivity during their transit. Conditions that affect the feasibility of the MVPC problem are formalized. Examples that assist in visualizing these conditions are provided.

Real world Multi-Vehicle Motion Planning (MVMP) problems require the optimization of suitable performance measures under an array of complex and challenging constraints involving kinematics, dynamics, communication connectivity, target tracking, and collision avoidance. The general MVMP problem can thus be formulated as a mathematical program (MP). In this paper we present a mathematical programming (MP) framework that captures the salient features of the general MVMP problem. To demonstrate the use of this framework for the formulation and solution of MVMP problems, we examine in detail four representative works and summarize several other related works. As MP solution algorithms and associated numerical solvers continue to develop, we anticipate that MP solution techniques will be applied to an increasing number of MVMP problems and that the framework and formulations presented in this paper may serve as a guide for future MVMP research.

We present a software library that aids in the design of mobile ad hoc networks (MANET). The OMAN design engine works by taking a specification of network requirements and objectives, and allocates resources which satisfy the input constraints and maximize the communication performance objective. The tool is used to explore networking design options and challenges, including: power control, adaptive modulation, flow control, scheduling, mobility, uncertainty in channel models, and cross-layer design. The unaddressed niche which OMAN seeks to fill is the general framework for optimization of any network resource, under arbitrary constraints, and with any selection of multiple objectives. While simulation is an important part of measuring the effectiveness of implemented optimization techniques, the novelty and focus of OMAN is on proposing novel network design algorithms, aggregating existing approaches, and providing a general framework for a network designer to test out new proposed resource allocation methods. In this paper, we present a high-level view of the OMAN architecture, review specific mathematical models used in the network representation, and show how OMAN is used to evaluate tradeoffs in MANET design. Specifically, we cover three case studies of optimization. The first case is robust power control under uncertain channel information for a single physical layer snapshot. The second case is scheduling with the availability of directional radiation patterns. The third case is optimizing topology through movement planning of relay nodes.

The acoustic through-metal channel is characterized by strong multipath components caused by the echoing of acoustic energy within the channel. Transmission at high data rates is therefore difficult to achieve with traditional single-carrier systems. This paper applies an adaptive bit-loading technique to the transmission of digital signals through metal barriers using ultrasonic signaling. The multi-carrier approach discussed here allows us to mitigate severe frequency selectivity of the through-metal communication link and improve spectral efficiency by exploiting the stationary nature of the channel. Experimental performance of bit loading is examined in an ultrasonic through-metal channel. Our results indicate that non-power-scaled rate adaptive bit loading significantly outperforms non-adaptive modulation. Adaptive bit loading was shown to adhere to a strict BER constraint while increasing data rates by roughly 240% from values of 5 Mbps to approximately 12 Mbps when compared to narrowband modulation techniques.

We present a mathematical programming based decentralized framework to generate time optimal velocity profiles for a group of path constrained mobile vehicle robots subject to communication connectivity constraints. Each vehicle robot starts from a fixed start point and moves towards a goal point along a fixed path so as to avoid collisions with other robots, and remain in communication connectivity with other robots. The main contribution of this paper is the discrete time decentralized Receding Horizon Mixed Integer Nonlinear Programming (RH-MINLP) formulation of the multi-vehicle path coordination problem with constraints on kinematics, dynamics, collision avoidance, and communication connectivity, and the application of state-of-the-art MINLP solution techniques. We test scenarios involving up to ten (10) robots to demonstrate (i) the effect of communication connectivity requirements on robot velocity profiles; and (ii) the dependence of the solution computation time on communication connectivity requirements.

This paper presents a framework for generating time-optimal velocity profiles for a group of path-constrained vehicle robots that have fixed and known initial and goal locations and are required to maintain communication connectivity. Each robot must follow a fixed and known path, arrive at its goal as quickly as possible (or at least not increase the time for the last robot to arrive at its goal) and stay in communication with other robots in the arena throughout its journey. The main contribution of this paper is the formulation of the problem as a discrete time nonlinear programming problem (NLP) with constraints on robot kinematics, dynamics, collision avoidance, and communication connectivity. We develop partition elimination constraints that assist in ensuring that the communication network is fully connected (no network partitions). These constraints are enforced only when network partitions would otherwise occur, an approach which significantly reduces the problem size and the required computational effort. In addition, we introduce path-constrained jammer robots with known paths and velocity profiles into the scenario. These jammer robots have an effective jamming range and disrupt all communications within this range. Except for the jammers, all robots must remain outside this jamming range at all times. We investigate the scalability of the proposed approach by testing scenarios involving up to fifty (50) robots. Solutions demonstrate (i) the trade off between the arrival time and the communication connectivity requirements in scenarios with and without jamming; and (ii) the dependence of computation time on the number of robots.

The genetics of prostate cancer are highly variable and not well understood. No consistent pattern of mutations across prostate cancer samples has yet been found. Due to this inherent heterogeneity it is natural to wonder whether or not using paired controls in gene expression studies might be useful. Although other studies have previously analyzed the use of paired controls for the expression of a handful of genes at a time no study has yet been performed to assess the benefits of using paired controls rather than independent controls on a large scale. By using a prostate cancer microarray data set that consisted of 58 pairs of paired cancer and control samples as well as 18 independent controls we found that searches for differentially expressed genes and for upregulated genes were significantly enhanced by using paired controls instead of independent controls.

Path planning and network design are often treated by architects of mobile communication networks as separate problems. In fact, most mobile ad hoc network (MANET) designs do not consider the path that the network nodes would take as part of the objective set, but incorporate them in an abstract form as general constraints on mobility (limit on the initial configuration and the node velocity). It appears that significant performance improvement may be possible if multiobjective optimization is attempted on performance indices such as connectivity of the network and arrival time of all nodes at a specified terminal set. In this study we provide an efficient decentralized method for computing the Pareto optimal set of plans, and show the benefit of path planning under six different metrics of MANET performance.

We propose two new filter-based feature selection criteria for identifying differentially expressed genes, namely the average difference score (ADS) and the mean difference score (MDS). These criteria replace the serial noise estimator used in existing criteria by a parallel noise estimator. The result is better detection of changes in the variance of expression levels, which t-statistic type criteria tend to under-emphasize. We compare the performance of the new criteria to that of several commonly used feature selection criteria, including the Welch t-statistic, the Fisher correlation score, the Wilcoxon rank sum, and the independently consistent expression discriminator, on synthetic data and real biological data obtained from acute lymphoblastic leukemia and acute myeloid leukemia patients. We find that ADS and MDS outperform the other criteria by exhibiting higher sensitivity and comparable specificity. ADS is also able to flag several biologically important genes that are missed by the Welch t-statistic.

High resolution direction-of-arrival (DOA) estimation is one of the most challenging problems in array signal processing. In this paper, a variation of the improved polynomial rooting (IPR) method is proposed for DOA estimation of multiple targets by a sensor array. The variation, unitary IPR (UIPR), transforms the complex-valued covariance matrix of the sensor signals to a real-valued matrix using unitary transformations. Then the IPR method is applied to determine the DOA of the targets. Simulation results indicate the potential improvement provided by our approach compared with MUSIC, Root-MUSIC, ESPRIT, and IPR.

A secure email protocol (SEP) is introduced to provide a decentralized means of revoking trust access and preventing false repudiation. Verification of the exchange of a message between sender and recipient is accomplished through an identification cookie (IC) and an encryption key (EK) handled by Verification and Authentication Agents (VAAs). The communicated narrative is encrypted by the EK. The augmented message is then encrypted again using one of the widely available encryption algorithms, such as the Pretty Good Privacy (PGP) scheme, and sent to the addressee. The extracted IC serves as a label which the addressee uses to obtain the EK from the sender's VAA. Once the addressee possesses the message's EK, s/he is able to decrypt the narrative, verify that the IC is correct, and the process is complete. Quick revocation of access privileges and cancellation of old messages is achieved by deleting ICs not requested by the addressee. The structure of the message body is written in the eXtensive Markup Language (XML) to take advantage of the language's extendibility and compatibility features.

A binary distributed detection system comprises a bank of local decision makers (LDMs) and a central information processor or data fusion center (DFC). All LDMs survey a common volume for a binary {H/sub 0/, H/sub 1/} phenomenon. Each LDM forms a binary decision: it either accepts H/sub 1/ (target-present) or H/sub 0/ (target-absent). The LDM is fully characterized by its performance probabilities. The decisions are transmitted to the DFC through noiseless communication channels. The DFC then optimally combines the local decisions to obtain a global decision which minimizes a Bayesian objective function. The DFC remembers and uses its most recent decision in synthesizing each new decision. When operating in a stationary environment, our architecture converges to a steady-state decision LDM in finite time with probability one, and its detection performance during convergence and in steady state is strictly determined. Once convergence is proven, we apply the results to the detection of signals with random phase and amplitude. We further provide a geometric interpretation for the behaviour of the system.

Since the introduction of fuzzy clustering by Ruspini (1970), fuzzy logic has provided a family of interesting clustering algorithms which expanded the abilities of 'crisp' techniques. The most popular among these algorithms is the fuzzy c means algorithm (FCM). However, FCM and most of its variants are sensitive to the presence of outliers in the data set. Past attempts to reduce this sensitivity included the addition of a "noise cluster" and the introduction of measures that assess the typicality of a vector to a cluster. In this study we provide additional means for outlier rejection through the introduction of a new variable, the credibility of a vector. Credibility measures the typicality of the vector to the entire data set (not to specific subsets as some previous techniques have done). An outlier is expected to have a low value of credibility compared to a non-outlier. The use of the new variable leads to the credibilistic fuzzy c means algorithm.

Inspired by an idea of Rimon and Koditschek (1992), we develop a motion planning algorithm for a point robot traveling among moving obstacles in an N-dimensional space. The navigating point must meet a goal point at a fixed time T, while avoiding several translating, nonrotating, nonintersecting obstacles on its way. All obstacles, the goal point, and the navigating point are confined to the interior of a star-shaped set in R/sup N/ over the time interval [0, T]. Full a priori knowledge of the goal's location and of the obstacle's trajectories is assumed. We observe that the topology of the obstacle-free space is invariant in the time interval [0, T] as long as the obstacles are nonintersecting and as long as they do not cover the goal point at any time during [0, T]. Using this fact we reduce the problem, for any fixed time t/sub 0//spl isin/[0, T], to a stationary-obstacle problem, which is then solved using the method of Rimon and Koditschek. The fact that the obstacle-free space is topologically invariant allows a solution to the moving-obstacle problem over [0, T] through a continuous deformation of the stationary-obstacle solution obtained at time t/sub 0/. We construct a vector field whose flow is in fact one such deformation. We believe that ours is the first global solution to the moving-obstacle path-planning problem which uses vector fields.

We review techniques for sensor fusion in robot navigation, emphasizing algorithms for self-location. These find use when the sensor suite of a mobile robot comprises several different sensors, some complementary and some redundant. Integrating the sensor readings, the robot seeks to accomplish tasks such as constructing a map of its environment, locating itself in that map, and recognizing objects that should be avoided or sought. The review describes integration techniques in two categories: low-level fusion is used for direct integration of sensory data, resulting in parameter and state estimates; high-level fusion is used for indirect integration of sensory data in hierarchical architectures, through command arbitration and integration of control signals suggested by different modules. The review provides an arsenal of tools for addressing this (rather ill-posed) problem in machine intelligence, including Kalman filtering, rule-based techniques, behavior based algorithms, and approaches that borrow from information theory, Dempster-Shafer reasoning, fuzzy logic and neural networks.

The planar stationary-obstacle path-planning problem for polygonal obstacles has been correctly and completely solved by T. Lozano-Perez and M. Wesley (1979), i.e., a global, optimal algorithm was provided which requires O(mu(2)logmu) computation time, where mu is the number of obstacle-faces in the scene. That algorithm is known as the VGRAPH algorithm. Two variants of VGRAPH have been developed to solve the same problem in O(mu(2)) computation time. Our paper discusses a recent algorithm proposed by C. Alexopoulos and P.M. Griffin (1992), called V*GRAPH, which also claims to provide an optimal solution. We demonstrate by counter-example that V*GRAPH is neither global nor optimal.

We analyze a motion planning algorithm proposed by Shih, Lee, and Gruver (1990). In particular, we discuss the case of motion planning among moving obstacles, for which an algorithm is presented by Shih, Lee, and Gruver (1990). The algorithm is claimed to always yield a collision-free path when one exists. We provide a specific counter-example to this claim, examine the possibility of repairing the algorithm, and discuss the computational complexity of a corrected algorithm.

The problem is to find the largest observation in a random sample of size n by asking binary-type questions of people (or items) in the sample. At each stage of the search, a threshold is calculated and a binary-type question is posed to each member of the sample. The threshold is determined from answers given to the previous questions, and no exact data is ever collected, i.e., no member is asked to explicitly provide his observation. Arrow, Pesotchinsky, and Sobel (APS) calculated the optimal threshold sequence for two different objectives: (i) minimize the average number of questions required for a solution, and (ii) maximize the probability of solving the problem in, at most, r questions. APS have assumed that the number of respondents in the affirmative at each stage of the search is exactly known. There exist applications where the number of affirmative answers is only known to be one member of the set (0, 1, more than 1). For these applications, we calculate exactly the optimal thresholds, in the sense of maximizing the probability of getting precisely one affirmative answer to the next binary question. An application of the threshold-calculation procedure is demonstrated in resolution of packet collisions over multiuser communication channel.>

The authors have devised a simple suboptimal decentralized binary detection system where each local detector independently, and the data function center, are minimizing Bayesian risks (not necessarily identical). They compare the hardware volume of a system with identical local risks to that of an optimal centralized system which minimizes the same global risk for several mathematically tractable cases. The results can be used by designers of distributed detection systems (even less restricted than the ones that the authors consider in this work) to obtain upper bounds on the number of (decentralized) detectors which are required in order to guarantee prespecified performance.>

Results are presented on two neural network strategies for the control of dynamic locomotive systems, in particular a one-legged hopping robot. The control task is to make corrections to the motion of the robot that serve to maintain a fixed level of energy (and minimize energy losses), which yields a stable periodic limit cycle in the system's state space. Control of the robot is achieved by the use of artificial neural networks (ANNs) with a continuous learning memory. Through continuous reinforcement for past successes and failures, the control system develops a stable strategy for accomplishing the desired control objectives. The results are presented in the form of computer simulation that demonstrate the ability of two different ANNs to devise proper control signals that will develop a stable hopping strategy, and hence a stable limit cycle in the robot's state space, using imprecise knowledge of both the current state and the mathematical model of the robot leg.>

An important class of neural models is described as a set of coupled nonlinear differential equations with state variables corresponding to the axon hillock potential of neurons. Through a nonlinear transformation, these models can be converted to an equivalent system of differential equations whose state variables correspond to firing rates. The firing rate formulation has certain computational advantages over the potential formulation of the model. The computational and storage burdens per cycle in simulations are reduced, and the resulting equations become quasilinear in a large significant subset of the state space. Moreover, the dynamic range of the state space is bounded, alleviating the numerical stability problems in network simulation. These advantages are demonstrated through an example, using the authors' model for the so-called neural solution to the traveling salesman problem proposed by J.J. Hopfield and D.W. Tank (1985).>

An architecture for an adaptive neuromorphic system designed to control a robot is suggested. The proposed architecture utilizes two important features of neural networks: the abundance of local minima in the network's state space and the uniformity of convergence of these minima in the face of growing dimensionality. The proposed approach is expected to yield controllers which are both faster and simpler than controllers which are designed by the methods of model reference adaptive control and self-tuning regulator. The controller's complexity is expected not to grow exponentially with the number of unknown parameters, and to allow adaptation in both continuous and discrete parameter domains. The possible benefits of the architecture are demonstrated on a single-degree-of-freedom manipulator, whose controller is assisted by a neural estimator.>

A sinusoid accompanied by stationary, additive, nonzero-mean, band-limited, white Gaussian noise is passed through a Gaussian nonlinear characteristic, and the output power spectral density is evaluated by using known results due to Rice and Atherton. Several characteristics of the Gaussian nonlinearity are revealed in the process, and input tuning is shown to contribute to output noise reduction. The results are applicable in the analysis of tracking systems employing sinusoidal dithers.