We propose a language, called {\sc Charon}, for modular specification of interacting hybrid systems. For hierarchical description of the system architecture, {\sc Charon} supports building complex agents via the operations of instantiation, hiding, and parallel composition. For hierarchical description of the behavior of atomic components, {\sc Charon} supports building complex modes via the operations of instantiation, scoping, and encapsulation. Features such as weak preemption, history retention, and externally defined Java functions, facilitate the description of complex discrete behavior. Continuous behavior can be specified using differential as well as algebraic constraints, and invariants restricting the flow spaces, all of which can be declared at various levels of the hierarchy. The modular structure of the language is not merely syntactic, but can be exploited during analysis. We illustrate this aspect by presenting a scheme for modular simulation in which each mode can be compiled solely based on the locally declared information to execute its discrete and continuous updates, and furthermore, submodes can integrate at a finer time scale than the enclosing modes.
This paper deals with the problem of recognizing objects seen from arbitrary viewpoints based on information obtained from a small number of reference images. We proceed by constructing a database that consists of a small number (ten is typical) of images of each object. Corner features are detected in each of the images in the database and local descriptors for these features are formed from the grayscale values in the surrounding regions. When a test image is obtained, its interest points and local descriptors are compared to those in the database. The recognition procedure also enforces an appropriate version of the epipolar constraint for affine cameras. The space requirements for this scheme increase linearly with the number of objects, and the time complexity scales logarithmically.
The method combines the advantages of feature based and image based recognition algorithms, while avoiding many of the drawbacks. While the method does identify and match interest points from images, we avoid the combinatorial issues inherent in the correspondence schemes employed by many feature based methods. In contrast to image based recognition techniques, the number of stored images is relatively small, yet the algorithm is capable of recognizing objects from viewpoints not represented in the database.
This paper deals with the problem of recovering the dimensions of an object and its pose from a single image acquired with a camera of unknown focal length. It is assumed that the object in question can be modeled as a polyhedron where the coordinates of the vertices can be expressed as a linear function of a dimension vector, $\lambda$. The reconstruction program takes as input a set of correspondences between features in the model and features in the image. From this information the program determines an appropriate projection model for the camera (scaled orthographic or perspective), the dimensions of the object, its pose relative to the camera and, in the case of perspective projection, the focal length of the camera. We demonstrate that this reconstruction task can be framed as an unconstrained optimization problem involving a small number of variables, no more than four, regardless of the number of parameters in the dimension vector.
Vertically-structured operating systems are a recent proposal for providing application-level resource management. While this has been accomplished, the issue of how the structure of the operating system affects application behaviour remains largely unexplored.
We introduce the concept of OS intrusion as a measure of the overheads that the operating system's policies and mechanisms impose upon applications. Vertical operating systems address policy intrusion; we analyse the costs of two forms of mechanism intrusion---interrupt handling and system call overhead---in the Linux kernel, and demonstrate their significance.
Piglet is a new architecture intended to provide the resource-management capabilities of a vertical OS, while reducing mechanism intrusion. Measurements of a prototype implementation show that Piglet achieves these goals; for example, round-trip times measured by the ping application are up to 20% lower than for Linux.
Almost all imaging systems require some form of registration. A few examples are aligning medical images for diagnosis, matching stereo images to recover shape, and comparing facial images in a database to recognize people. Recently, a new type of solution to the registration problem has emerged, based on information theory. In particular, the mutual information similarity metric has been used to register multi-modal medical images. Mutual information compares the statistical dependence between the two images. Unlike many other registration techniques, mutual information makes few a priori assumptions about the surface properties of the object or the imaging process, making it adaptible to changes in lighting and changes between sensors. The method can be applied to larger dimensional registration and many other imaging situations. In this report, we compare two approaches taken towards the implementation of rigid 2D mutual information image registration. We look further at algorithm speedup and noise reduction efforts. A full background is provided.
This paper describes an approach to capturing the appearance and structure of immersive environments based on the video imagery obtained with an omnidirectional camera system. The scheme proceeds by recovering the 3D positions of a set of point and line features in the world from image correspondences in a small set of key frames in the image sequence. Once the locations of these features have been recovered the position of the camera during every frame in the sequence can be determined by using these recovered features as fiducials and estimating camera pose based on the location of corresponding image features in each frame. The end result of the procedure is an omnidirectional video sequence where every frame is augmented with its pose with respect to an absolute reference frame and a 3D model of the environment composed of point and line features in the scene.
By augmenting the video clip with pose information we provide the viewer with the ability to navigate the image sequence in new and interesting ways. More specifically the user can use the pose information to travel through the video sequence with a trajectory different from the one taken by the original camera operator. This freedom presents the end user with an opportunity to immerse themselves within a remote environment and to control whay they see.
We present the load-calculus, used to model dynamic loading, and prove it sound. The calculus extends the polymorphic lambda-calculus with a load primitive that dynamically loads terms that are closed, with respect to values. The calculus is meant to approximate the process of dynamic loading in TAL/Load [4], an version of Typed Assembly Language [7] extending with dynamic linking. To model the key aspects of TAL, the calculus contains references and facilities for named types. Loadable programs may refer to named types defined by the running program, and may export new types to code loaded later. Our approach follows the framework initially outlined by Glew et. al [3]. This calculus has been implemented in the TALx86 [6] version of Typed Assembly Language, and is used to implement a full-featured dynamic linking library, DLpop [4].
The use of Coulomb's friction law with the principles of classical rigid body dynamics introduces mathematical inconsistencies. Specifically, the forward dynamics problem can have no solutions or multiple solutions. In these situations, compliant contact models, while increasing the dimensionality of the state vector, can resolve these problems. The simplicity and efficiency of rigid body models, however, provide strong motivation for their use during those portions of a simulation when the rigid body solution is unique and stable.
In this paper, we use singular perturbation analysis in conjunction with linear complementarity theory to establish conditions under which the solution predicted by the rigid body dynamic model is stable. We employ a general model of contact compliance to derive stability criteria for planar mechanical systems. In particular, we show that for cases with one sliding contact, there is always at most one stable solution. Our approach is not directly applicable to transitions between rolling and sliding where the Coulomb friction law is discontinuous. To overcome this difficulty, we introduce a smooth nonlinear friction law, which approximates Coulomb friction. Such a friction model can also increase the efficiency of both rigid body and compliant contact simulation. Numerical simulations for the different models and comparison with experimental results are also presented.
The ubiquity and complexity of modern networks require automated management and control. With increases in scale, automated solutions based on simple data access models such as SNMP will give way to more distributed and algorithmic techniques. This article outlines present and near-term solutions based on the ideas of active networks and mobile agents, which permit sophisticated programmable control and management of ultra large scale networks.
Multiple Attribute Decision Making (MADM) involves "making preference decisions (such as evaluation, prioritization, selection) over the available alternatives that are characterized by multiple, usually conflicting, attributes". The problems of MADM are diverse, and can be found in virtually any topic. In this paper, we use three different scoring methods for evaluating the performance of different imaging techniques used to detect cancers in the female breast. The need for such a decision support system arises from the fact that each of the several techniques which helps diagnose breast cancer today, has its own specific characteristics, advantages and drawbacks. These characteristics or attributes are generally conflicting. The goal is to detect as many malignant lesions in the breast as is possible, while identifying the maximum number of benign lesions. The four imaging techniques that are compared here are Magnetic Resonance Imaging (MRI), Mammography, Ultrasonography, and Nuclear Medicine. The three different multiattribute scoring methods are the Simple Additive Weighting method (SAW), the Weighted Product Method (WPM), and the Technique for Order Preference by detail, and then used to rank the four imaging techniques. The results are analyzed and the validity and robustness of the methods are tested using post-evaluation analysis.
Breast cancer is the most commonly diagnosed cancer and the second leading cause of cancer death among women in America. A few years ago the odds of developing breast cancer were reported as 1 in 13. Now the chance is 1 in 9. The only way today to find out for sure if a breast lump or abnormal tissue is cancer, is by having a biopsy: A suspicious tissue is removed by a surgical excision or needle biopsy and is examined under a microscope by a pathologist who makes the diagnosis. Imaging techniques of the breast are therefore vital since they will allow early detection of cancer, and localization of the suspicious lesion in the breast for a biopsy procedure.
Recent research in active networking has motivated the use of programmable, or active, packets in which a traditional packet header is replaced by a program that controls a packet's actions in the network. Despite the popularity of this idea in the active networking community, it has not taken hold in general. We claim that this is because the current active packet systems are not sufficiently practical.
Our goal is to consider why current systems are not practical and to propose directions that will facilitate movement towards practical programmable packets. To this end, we first establish a framework that defines what it means to be "practical". Next, we look at the existing "first-generation" active packet systems, paying particular regard to the key aspects of our framework. This survey is one of the major contributions of this paper. Finally, we discuss the overall lessons we draw from the first generation systems and present a view of what "second-generation" systems should consist of. We illustrate our views on second-generation systems by using SNAP (Safe Networking with Active Packets), a second-generation system we are currently exploring.
We present the design and implementation of a framework for flexible and safe dynamic linking of native code. Our approach extends Typed Assembly Language with a primitive for loading and typechecking code, which is flexible enough to support a variety of linking strategies, but simple enough that it does not significantly expand the trusted computing base. Using this primitive, along with the ability to compute with types, we show that we can program many existing dynamic linking approaches. As a concrete demonstration, we have used our framework to implement dynamic linking for a type-safe dialect of C, closely modeled after the standard linking facility for Unix C programs. Aside from the unavoidable cost of verification, our implementation performs comparably with the standard, untyped approach.
Tele-immersion is a new medium that enables a user to share a virtual space with remote participants. The user is immersed in a rendered 3D-world that is transmitted from a remote site. To acquire this 3D description we apply bi-and trinocular stereo techniques. The challenge is to compute dense stereo range data at high frame rates, since participants cannot easily communicate if the processing cycle or network latencies are long. Moreover, new views of the received 3D-world must be as accurate as possible. We address both issues of speed and accuracy and we propose a method for combining motion and stereo in order to increase speed and robustness.