Search

In this paper we describe a coalition formation model for a cooperative multiagent system in which each agent has incomplete information about its dynamic and uncertain world and must respond to sensed events within time constraints. With incomplete information and uncertain world parameters while lacking time, an agent cannot afford organizing a rationally optimal coalition formation. Instead, our agents use a two-stage methodology. When an agent detects an event in the world, it first compiles a list of coalition candidates that it thinks would be useful, and then negotiates with the candidates. 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. The agent that initiates the coalition needs to determine the task distribution among the members of the coalition and designs its coalition strategy to increase the chance of successfully forming a working coalition. Since the environment is dynamic, noisy, and the agents are resource-constrained, agents must form the working coalition to react to events as soon as possible and with whatever partial information they currently hold., Sponsored in part by the Defense Advanced Research Projects Agency (DARPA).

In this paper, we present a set of algorithms for describing sea ice features in SAR images. These algorithms we have implemented and incorporated in the ARKTOS software package, an intelligent sea ice classification. These algorithms have unique characteristics. Some are extensions or adaptations of existing image processing techniques to the specific problem domain of satellite sea ice classification, while some are innovative designs, inspired by the aforementioned problem domain. These feature descriptors may be generalized to other remote sensing applications.

ARKTOS (Advanced Reasoning Using Knowledge for Typing of Sea Ice) is a fully automated intelligent sea ice classification system. ARKTOS is in use at the U.S. National Ice Center (NIC) for daily operations related to the NIC’S task of mapping the ice covered oceans. ARKTOS incorporates image processing, input from ancillary data, and artificial intelligence (AI) to analyze and classify RADARSAT Synthetic Aperture Radar (SAR) imagery. The NIC and Naval Research Laboratory (NRL/ERIM) have been testing and evaluating ARKTOS through the freeze-up, winter, melt-out and summer seasons of the Beaufort Sea. In this paper we outline the development and evolution of ARKTOS, evaluate current output, and discuss future implementation.

ARKTOS is a fully automated intelligent system that classifies sea ice and that is now being used by the U.S. National Ice Center (NIC) for daily operations related to the NIC's task of mapping the ice covered oceans. In this paper we describe the process of taking a research project and transitioning it to an operational environment. We discuss the theoretical methodologies implemented in ARKTOS, and how ARKTOS was developed, tested, and finally moved to operations.

The ARKTOS (Advanced Reasoning using Knowledge for Typing of Sea Ice) system is a sea ice classification system that incorporates image processing and Artificial Intelligence (Al) to analyze RADARSAT Synthetic Aperture Radar (SAR) images. As shown in Figure 1, ARKTOS first segments the input image using Watershed region growing technique based on image gradients, and subsequently merges regions based on area, average intensity, and strength of common boundaries. ARKTOS then computes attributes for each contiguous region (henceforth, feature) such as area, perimeter, average intensity, and circularity. Given these measurements, facts regarding each feature are formed by quantizing the values into symbols. For example, if the feature's average intensity is less than 50.0, then intensity (feature) = black. The decision points (or thresholds) were determined and refined via visual inspection of features., Prepared in cooperation with National Ice Center, Washington, and Canadian Ice Service, Ottawa, Ontario.

ARKTOS is a fully automated intelligent system that classifies sea ice and that is now being used by the U.S. National Ice Center (NIC) for daily operations related to the NIC’s task of mapping the ice covered oceans. In this paper we describe the process of taking a research project and transitioning it to an operational environment. We discuss the theoretical methodologies implemented in ARKTOS, and how ARKTOS was developed, tested, and finally moved to operations.