The U.S. Defense Advanced Research Projects Agency's (DARPA) Neovision2 program aims to develop artificial vision systems based on the design principles employed by mammalian vision systems. Three such algorithms are briefly described in this paper. These neuromorphic-vision systems' performance in detecting objects in video was measured using a set of annotated clips. This paper describes the results of these evaluations including the data domains, metrics, methodologies, performance over a range of operating points and a comparison with computer vision based baseline algorithms.

The process of diagnosis involves learning about the state of a system from various observations of symptoms or findings about the system. Sophisticated Bayesian (and other) algorithms have been developed to revise and maintain beliefs about the system as observations are made. Nonetheless, diagnostic models have tended to ignore some common sense reasoning exploited by human diagnosticians. In particular, one can learn from which observations have not been made, in the spirit of conversational implicature. In order to extract information from the observations not made, we propose the following two concepts. First, some symptoms, if present, are more likely to be reported before others. Second. most human diagnosticians and expert systems are economical in their data-gathering, searching first where they are more likely to find symptoms present. Thus, there is a desirable bias toward reporting symptoms that are present. We develop a simple model for these concepts that can significantly improve diagnostic inference.

A number of algorithms have been developed to solve probabilistic inference problems on belief networks. These algorithms can be divided into two main groups: exact techniques which exploit the conditional independence revealed when the graph structure is relatively sparse, and probabilistic sampling techniques which exploit the "conductance" of an embedded Markov chain when the conditional probabilities have non-extreme values. In this paper, we investigate a family of "forward" Monte Carlo sampling techniques similar to Logic Sampling [Henrion, 1988] which appear to perform well even in some multiply connected networks with extreme conditional probabilities, and thus would be generally applicable. We consider several enhancements which reduce the posterior variance using this approach and propose a framework and criteria for choosing when to use those enhancements.

The mammalian visual system is still the gold standard for recognition accuracy, flexibility, efficiency, and speed. Ongoing advances in our understanding of function and mechanisms in the visual system can now be leveraged to pursue the design of computer vision architectures that will revolutionize the state of the art in computer vision.

This paper describes a novel approach for the detection and classification of man-made objects using discriminating features derived from higher-order spectra (HOS), defined in terms of higher-order moments of hyperspectral-signals. Many existing hyperspectral analysis techniques are based on linearity assumptions. However, recent research suggests that significant nonlinearity arises due to multipath scatter, as well as spatially varying atmospheric water vapor concentrations. Higher-order spectra characterize subtle complex nonlinear dependencies in spectral phenomenology of objects in hyperspectral data and are insensitive to additive Gaussian noise. By exploiting these HOS properties, we have devised a robust method for classifying man-made objects from hyerspectral signatures despite the presence of strong background noise, confusers with spectrally similar signatures and variable signal-to-noise ratios. We tested classification performance hyperspectral imagery collected from several different sensor platforms and compared our algorithm with conventional classifiers based on linear models. Our experimental results demonstrate that our HOS algorithm produces significant reductions in false alarms. Furthermore, when HOS-based features were combined with standard features derived from spectral properties, the overall classification accuracy is substantially improved.

Unmanned Air Vehicles (UAVs) are expected to dramatically alter the way future battles are fought. Autonomous collaborative operation of teams of UAVs is a key enabler for efficient and effective deployment of large numbers of UAVs under the U. S. Army's vision for Force Transformation. Autonomous Collaborative Mission Systems (ACMS) is an extensible architecture and collaborative behavior planning approach to achieve multi-UAV autonomous collaborative capability. Under this architecture, a rich set of autonomous collaborative behaviors can be developed to accomplish a wide range of missions. In this article, we present our simulation results in applying various autonomous collaborative behaviors developed in the ACMS to an integrated convoy protection scenario using a heterogeneous team of UAVs.

We view planning as a search through a space of plan models. A plan model consists of a partial description of a course of action (a plan) and a set of decision models that support analysis of the plan. In this framework, model building is focussed on the development of techniques to support the incremental evaluation and construction of plans. There are special problems associated with planning that make model construction in this context much more difficult than it might be for other applications. Since many alternative structures must be generated and evaluated while planning, exhaustive search techniques for model construction are inappropriate. We seek to develop techniques for building sparse decision structures (models) that contain enough information to allow us to make search choices without swamping the evaluator with a lot of inessential trivia.

UAVs are critical to the U. S. Army's Force Transformation. Large numbers of UAVs will be employed per Future Combat System (FCS) Unit of Action (UoA). To relieve the burden of controlling and coordinating multiple UAVs in a given UoA, UAVs must operate autonomously and collaboratively while engaging in RSTA and other missions. Rockwell Scientific is developing Autonomous Collaborative Mission Systems (ACMS), an extensible architecture and behavior planning/collaboration approach, to enable groups of UAVs to operate autonomously in a collaborative environment. The architecture is modular, and the modules may be run in different locations/platforms to accommodate the constraints of available hardware, processing resources and mission needs. The modules and uniform interfaces provide a consistent and platform-independent baseline mission collaboration mechanism and signaling protocol across different platforms. Further, the modular design allows for the flexible and convenient extension to new autonomous collaborative behaviors to the ACMS. In this article, we first discuss our observations in implementing autonomous collaborative behaviors in general and under ACMS. Second, we present the results of our implementation of two such behaviors in the ACMS as examples.

UAVs are a key element of the U.S. Army's vision for force transformation. UAVs are employed in large numbers per future combat system (FCS) unit of action (UoA). This necessitates a multi-UAV level of autonomous collaboration behavior capability that meets RSTA and other mission needs of the FCS UoAs. The autonomous collaborative mission systems (ACMS) are an extensible architecture and behavioral planning/collaborative approach to achieve this level of capability. We present a market-based approach that we developed as the main mechanism for autonomous collaboration in the ACMS. To enable flexible collaboration among a variety of heterogeneous unmanned vehicles for a broad range of missions, this market-based collaboration approach adopts a two-stage task specification and negotiation process that can accommodate different mission planning and task allocation strategies. We describe our market-based approach, its main features, and the collaboration protocol in this article

We illustrate an approach for planning UAV sensing actions in urban or constrained domains. We plan and optimize a collection strategy for a target of interest using Design Sheet, a numeric/symbolic algebraic constraint propagation package. Once a set of sensing plans have been developed, we use a probabilistic roadmap planning algorithm to plan a route for a fixed wing UAV through urban terrain to collect that information. This planner has several novel features to improve performance for urban domains.

UAVs are a key element of the U. S. Army's vision for Force Transformation, and are expected to be employed in large numbers per FCS Unit of Action (UoA). This necessitates a multi-UAV level of autonomous collaboration behavior capability that meets RSTA and other mission needs of FCS UoAs. Autonomous Collaborative Mission Systems (ACMS) is an extensible architecture and behavior planning / collaborative approach to achieve this level of capability. The architecture is modular and the modules may be run in different locations/platforms to accommodate the constraints of available hardware, processing resources and mission needs. The modules and uniform interfaces provide a consistent and platform-independent baseline mission collaboration mechanism and signaling protocol across different platforms. Further, the modular design allows flexible and convenient extension to new autonomous collaborative behaviors to the ACMS through: adding new behavioral templates in the Mission Planner component; adding new components in appropriate ACMS modules to provide new mission specific functionality; adding or modifying constraints or parameters to the existing components, or any combination of these. We describe the ACMS architecture, its main features on extensibility, and updates on current spiral development status and future plans for simulations in this report.

UAVs are a key element of the Army’s vision for Force Transformation, and are expected to be employed in large numbers per FCS Unit of Action (UoA). This necessitates a multi-UAV level of autonomous collaboration behavior capability that meets RSTA and other mission needs of FCS UoAs. Autonomous Collaborative Mission Systems (ACMS) is a scalable architecture and behavior planning / collaborative approach to achieve this level of capability. The architecture is modular and the modules may be run in different locations/platforms to accommodate the constraints of available hardware, processing resources and mission needs. The Mission Management Module determines the role of member autonomous entities by employing collaboration mechanisms (e.g., market-based, etc.), the individual Entity Management Modules work with the Mission Manager in determining the role and task of the entity, the individual Entity Execution Modules monitor task execution and platform navigation and sensor control, and the World Model Module hosts local and global versions of the environment and the Common Operating Picture (COP). The modules and uniform interfaces provide a consistent and platform-independent baseline mission collaboration mechanism and signaling protocol across different platforms. Further, the modular design allows flexible and convenient addition of new autonomous collaborative behaviors to the ACMS through: adding new behavioral templates in the Mission Planner component, adding new components in appropriate ACMS modules to provide new mission specific functionality, adding or modifying constraints or parameters to the existing components, or any combination of these. We describe the ACMS architecture, its main features, current development status and future plans for simulations in this report.

: This effort focused on representation and evaluation of plans accounting for uncertainty in the parameters upon which the plan is based and also uncertainty in the outcomes that will result from potential actions of the plan. Methods used to accomplish these results included the use of Action Networks, and development of a suite of analysis tools in support of the AFRL Campaign Assessment Tool (CAT - AKA Causal Analysis Tool). Action networks is a language for representing actions and their effects based on Bayesian networks. The analysis tools were intended to provide supporting analysis of air campaign plans including analysis of the value of observing new information and the value of controlling key uncertainties.

We present a strategy for suspending recursive open conditions during planning. We also show conditions under which plans with suspended open conditions can be pruned. To make this suspension and pruning strategy efficient, we use an operator graph to analyze potential recursion before the planning process begins. This approach covers a broader range of recursive problems than the approaches of Morris and Kambhampati, and is much more tractable than Kambhampati's approach. We give experimental results that indicate 1) significant improvement on recursive problems and 2) negligible overhead when applied to recursive and non-recursive problems alike.