XML has emerged as the primary standard of data representation and data exchange [13]. Although many software tools exist to assist the XML implementation process, data must be manually entered into the XML documents. Current form filling technologies are mostly for simple data entry and do not provide support for the complexity and nested structures of XML grammars. This paper presents SmartXAutofill, an intelligent data entry assistant for predicting and automating inputs for XML documents based on the contents of historical document collections in the same XML domain. SmartXAutofill incorporates an ensemble classifier, which integrates multiple internal classification algorithms into a single architecture. Each internal classifier uses approximate techniques to propose a value for an empty XML field, and, through voting, the ensemble classifier determines which value to accept. As the system operates it learns which internal classification algorithms work better for a specific XML document domain and modifies its weights (confidence) in their predictive ability. As a result, the ensemble classifier adapts itself to the specific XML domain, without the need to develop special learners for the infinite number of domains that XML users have created. We evaluated our system performance using data from eleven different XML domains. The results show that the ensemble classifier adapted itself to different XML document domains, and most of the time (for 9 out of 11 domains) produced predictive accuracies as good as or better than the best individual classifier for a domain.

The PRISM (Polar Radar for Ice Sheet Measurements) project at the University of Kansas is developing intelligent radar sensors for the measurement and study of the mass balance of the polar ice sheets. An important component of PRISM is intelligent, autonomous synthetic aperture radar that can reason about its operating mode (monostatic vs. bistatic) and frequency, based on a variety of environmental and sensor-related factors. The PRISM sensors are placed on autonomous robotic vehicles ("rovers") that use the sensor and environmental information to decide about what paths to traverse, how to traverse them, and at what speeds. In our work, we have implemented the reasoning component of the autonomous radar and the rovers, using intelligent agents and Bayesian networks. This implementation is the first ever of a dynamically modifying adaptive radar and mobile data collection system based on autonomous rovers for accurate polar ice sheet measurements.

The PRISM (Polar Radar for Ice Sheet Measurement) project is developing mobile, autonomous sensors for the measurement and study of the mass balance of the polar ice sheets. These sensors consist of intelligent radars integrated into robotic vehicles. They autonomously decide where and how to measure by examining a variety of information including onboard sensor data and collections of a priori knowledge. These data include the health and status of the rover, health and status of the sensors themselves, the state of the environment as measured by the sensors, satellite measurements of the area indicating expected ice sheet motion, and so on. All of this information is used to direct the data collection process by allowing for the dynamic configuration of sensors and the motion of the rovers that carry them. The PRISM intelligent sensor and rover control system is built upon a multiagent collaborative architecture that involves a number of distinct data collection and data dissemination agents functioning continuously and autonomously in a distributed computing framework. A critical component of this system is an agent service called the Matchmaker. The Matchmaker coordinates requests for information and services within the agent community and allows decision-making agents to locate and communicate with the data source-agents that can fulfill these requests.

We present an intelligent system for satellite sea ice image analysis named Advanced Reasoning using Knowledge for Typing Of Sea ice (ARKTOS). ARKTOS performs fully automated analysis of synthetic aperture radar (SAR) sea ice images by mimicking the reasoning process of sea ice experts. ARKTOS automatically segments a SAR image of sea ice, generates descriptors for the segments of the image, and then uses expert system rules to classify these sea ice features. ARKTOS also utilizes multisource data fusion to improve classification and performs belief handling using Dempster-Shafer. As a software package, ARKTOS comprises components in image processing, rule-based classification, multisource data fusion, and graphical user interface-based knowledge engineering and modification. As a research project over the past ten years, ARKTOS has undergone phases such as knowledge acquisition, prototyping, refinement, evaluation, deployment, and operationalization at the U.S. National Ice Center. In this paper, we focus on the methodology, evaluations, and classification results of ARKTOS.

In this paper we propose a coalition formation model for a cooperative multiagent system in which an agent forms sub-optimal coalitions in view of incomplete information about its noisy, dynamic, and uncertain world, and its need to respond to events within time constraints. Our model has two stages: (1) when an agent detects an event in the world, it first compiles a list of coalition candidates that it thinks would be useful (coalition initialization), and (2) then negotiates with the candidates (coalition finalization). A negotiation is an exchange of information and knowledge for constraint satisfaction until both parties agree on a deal or one opts out. Each successful negotiation adds a new member to the agent's final coalition. This paper talks about the steps we have designed to enhance the finalization stage.

This paper defines the restricted growing concept (RGC) fur object separation and provides an algorithmic analysis of its implementations. Our concept decomposes the problem of object separation into two stages. First, separation is achieved by shrinking the objects to their cores while keeping track of their originals as masks. Then the core is grown within the masks obeying the guidelines of a restricted growing algorithm. In this paper, we apply RGC to the remote sensing domain, particularly the synthetic aperture radar (SAR) sea ice images.

The authors define the concept of information fusion and show how they used it to estimate summer sea ice concentration in the marginal ice zone (MIZ) from single-channel SAR satellite imagery. They used data about melt stage, wind speed, and surface temperature to generate temporally-accumulated information, and fused this information with the SAR image, resulting in an interpretation of summer MIZ imagery. They also used the results of previous classifications of the same area to guide and correct future interpretations, thus fusing historical information with imagery and nonimagery data. They chose to study the summer MIZ since summer melt conditions cause classification based upon backscatter intensity to fail, as the backscatter of open water, thin ice, first-year ice, and multiyear ice overlap to a large degree. This makes it necessary to fuse various information and data to achieve proper segmentation and automated classification of the image. Their results were evaluated qualitatively and showed that their approach produces very good ice concentration estimates in the summer MIZ.

This paper presents a preliminary study for mapping sea ice patterns (texture) with 100-m ERS-1 synthetic aperture radar (SAR) imagery. The authors used gray-level co-occurrence matrices (GLCM) to quantitatively evaluate textural parameters and representations and to determine which parameter values and representations are best for mapping sea ice texture. They conducted experiments on the quantization levels of the image and the displacement and orientation values of the GLCM by examining the effects textural descriptors such as entropy have in the representation of different sea ice textures. They showed that a complete gray-level representation of the image is not necessary for texture mapping, an eight-level quantization representation is undesirable for textural representation, and the displacement factor in texture measurements is more important than orientation. In addition, they developed three GLCM implementations and evaluated them by a supervised Bayesian classifier on sea ice textural contexts. This experiment concludes that the best GLCM implementation in representing sea ice texture is one that utilizes a range of displacement values such that both microtextures and macrotextures of sea ice can be adequately captured. These findings define the quantization, displacement, and orientation values that are the best for SAR sea ice texture analysis using GLCM.

The authors describe a segmentation technique that integrates traditional image processing algorithms with techniques adapted from knowledge discovery in databases (KDD) and data mining to analyze and segment unstructured satellite images of natural scenes. They have divided their segmentation task into three major steps. First, an initial segmentation is achieved using dynamic local thresholding, producing a set of regions. Then, spectral, spatial, and textural features for each region are generated from the thresholded image. Finally, given these features as attributes, an unsupervised machine learning methodology called conceptual clustering is used to cluster the regions found in the image into N classes-thus, determining the number of classes in the image automatically. They have applied the technique successfully to ERS-1 synthetic aperture radar (SAR). Landsat thematic mapper (TM), and NOAA advanced very high resolution radiometer (AVHRR) data of natural scenes.

Leads are relatively linear features in the sea ice cover, which are composed of open water or new, thin ice. Because of their composition, leads impact the ocean/air heat exchange. Automated analysis of leads from sea ice imagery may provide a means of gathering important information about the sea ice cover and its climatic influence. This paper describes: (1) a method for extracting and analyzing leads from ERS-1 synthetic aperture radar (SAR) images classified by ice type and (2) the results of using this method on images of the Beaufort Sea. The methodology consists of identifying potential lead features in the image and measuring their characteristics both before and after using a thinning or skeletonization technique on the features. The measurements obtained using this method include lead area, average width, number of leads in an area, amount of branching, and linearity of the lead. These measurements were analyzed with respect to the time of year and the latitude of the images. Results indicate that the measurements produced by the methodology are consistent with lead measurement distributions that others have found. The results of the study suggest that the methodology is appropriate to study lead characteristics on a large scale.

Documents an approach to sea ice classification through a combination of methods, both algorithmic and heuristic. The resulting system is a comprehensive technique, which uses dynamic local thresholding as a classification basis and then supplements that initial classification using heuristic geophysical knowledge organized in expert systems. The dynamic local thresholding method allows separation of the ice into thickness classes based on local intensity distributions. Because it utilizes the data within each image, it can adapt to varying ice thickness intensities to regional and seasonal changes and is not subject to limitations caused by using predefined parameters.>

The research presented concentrates on bringing together case-based reasoning and expert knowledge-based reasoning in a planner called TOLTEC. The expert's domain-specific knowledge is modelled as dynamic memory structures and this representation is used to help the planner reason and control its planning process. TOLTEC uses a complex indexing of its cases, so as to allow incremental retrieval. The TOLTEC planner is applied to a highly constrained domain, and it is shown how the final plan is created by adding memory- and situation-selected chunks of subtask expansion to each subtask, until the problem is reduced to primitive (nonexpandable) tasks. It is also shown how the use of dynamic memory structures and dynamic, user-directed backtracking allows the planner to predict and discover failures, recover from them, and modify its knowledge according to them. Finally, it is shown how in a domain of multiple possible solutions for each goal the methodology developed allows the planner to slowly model itself to the preferences of the user. The paper also discusses some of the application domains where TOLTEC has been used, including process planning of cylindrical and prismatic parts, design checking, material selection, and design of communications systems.>

Design has been one of the most difficult problems that Artificial Intelligence has faced until now. Design knowledge is almost impossible to gather and put into rules or other structured reasoning tool, and studies of human expert designers have shown that often the design process seems ad hoc and based on “intuition”. This paper presents a novel approach to automating design by using case-based reasoning and learning. It discusses how cases of previous designs are stored, how they are retrieved, how they are judged based on their adherence to design specifications, how the designs are created, and how the memory of design cases is expanded by learning. Finally, this paper presents the current state of an automated design tool based on these principles, which will be used in the intelligent design of network topologies.

The authors present an intelligent manufacturing system that combines design with process planning. For the design stage they have implemented an experimental package that combines boundary representation, CSG, and models of manufacturing features in a simple graphics environment. This package, named XTWIN, was combined with an intelligent process planning system TOLTEC which uses cases of previous process plans stored as dynamic memory structures. This knowledge organization allows TOLTEC to rearrange its memory, thus learning from experience, and to also apply its experience to new cases.>