Abstract: Autonomous robots must be designed to function in real-world domains. These domains are characterized by uncertainty, yet little work has been done to quantify the effects of uncertainty on goal satisfaction. Our research on simulated repair robots indicates that different treatments of uncertainty have fundamentally different relationships to plan success. We investigate the impacts of sensor inaccuracies, exogenous events, and failures in both motion and gripper operations. Extensive simulations provide the basis for mathematical models of the relationship between these types of uncertainties and plan success. In addition, we investigate the impact of current heuristics such as {\it retry on failure}, and {\it tell me three times}. We conclude that even low levels of uncertainty can significantly impact goal satisfaction in a simulated repair robot domain, and that repeated attempts provide significant improvement in plan success rates in the face of uncertainty.

Abstract: This paper presents a model of the automated planning problem that can extend the range of tractable problems, and decrease the solution time for already tractable ones. It uses the concept of midpoint navigation to reduce planning problems with solutions of length N, into sub-problems with reduced length. Due to the exponential elaboration time of planning problems, this provides significant benefit to the planner. However, a critical aspect of this approach is selecting good midpoints for evaluation. We argue that several established data mining techniques might be used to determine these waypoints, and outline methods for applying these techniques.

Abstract: This paper presents preliminary work on the design of control systems for pedestrian mobility aids for the elderly. The elderly are often restricted in their mobility and must rely on canes, walkers and wheelchairs for locomotion. Restrictions in mobility lead to a loss of independence and autonomy, as well as a decrease in muscular strength. This paper focuses on design of intelligent wheeled walkers. By allowing the user varying degrees of control, from complete to collaborative, these walkers afford the user with the feeling of control, while helping to increase the ease and safety of their daily travels. The control systems of these walkers differ from those of other mobility aids and mobile robots because they must both assist in mobility and provide balance and support. These functions must be performed in a tight loop with a human whose input may be difficult to predict.

Abstract: We present a transform that enables traditional Shortest-Feasible-Plan planners to reason about uncertain operators and produce plans which have higher probabilities of success. This transform converts a probability-aware domain description into a STRIPS-style description, where the probability of success is expressed by plan length. Using this transformed description a plan can be generated by a traditional planner. The transform is shown to be at worst linear in the size of the input, and allows the planning system to trade-off accuracy against runtime as an anytime computation.

Abstract: In this paper we present a model for an agent architecture that supports adaptive goal prioritization. Goal prioritization is the ability of the agent to adjust the relative priorities of a goal set in response to changes in a dynamic environment. This architecture allows an agent to act pro-actively to achieve goals, and to dynamically determine which goals should be pursued at any time. We discuss a dynamic domain in which the architecture will be situated, and outline the techniques needed to support goal re-prioritization.

Abstract: This paper describes a method for coupling computer simulation models with Geographic Information Systems (GIS) that permits the model to incorporate the features of the landscape and topography, resulting in higher fidelity predictions. Traditional nonspatial computer models rely on aggregate data and produce good general predictions. The coupling of a computer simulation with the spatial information handling capabilities of a GIS allows the model to incorporate effects due to non-homogeneous domains. This can result in models with greater accuracy. The example used in this paper is a model predicting the spread of invasive plants with nonhomogeneous soil types. A raster based GIS package was coupled with a model of plant spread, which allowed the model to permit differential growth based on slope and soil characteristics. The modeling process predicted the eventual spread of the invasive species and the results were compared with the accepted aggregate population formulas. The method has general applicability to any modeling scenario in which the heterogeneous effects of terrain, surface characteristics, or wind/airflow patterns cause non-uniform dispersion of the model subject.

Abstract: One of the Department of Energy's (DOE) ongoing tasks is the storage and inspection of a large number of waste barrels containi9ng a variety of hazardous substances. Martin Marietta is currently contracted to develop a robotic system - the Intelligent Mobile Sensor System (IMSS) - for the automatic monitoring and inspection of these barrels. The IMSS is a mobile robot with multiple sensors: video cameras, illuminators, laser ranging and barcode reader. We assisted Martin Marietta in this task, specifically in the development of image processing algorithms that recognise and classify barrel lables. Our subsystem uses video images to detect and locate the barcode, so that the barcode reader can be pointed at the barcode.

Abstract: Robust autonomy, in the sense of performing tasks in the face of dynamic changes to the environment, requires that an autonomous system be capable of responding appropriately to such changes. One such response is to effectively adapt the allocation of resources from planning to execution. By adapting the resource allocation between deliberation and execution, an autonomous system can produce shorter plans more frequently in environments with high levels of uncertainty, while producing longer, more complex plans when the environment offers the opportunity to successfully execute complex plans. In this paper we propose the idea of the "effective planning horizon" which adapts to environmental changes to bound the deliberation in an interleaved planning/execution system. The effective planning horizon is developed from an analysis of the advantages and disadvantages of three classic autonomous system architectures as feedback control systems. This leads to the development of an analytic model which suggests the use of maximizing the expected value of plans by adjusting the planning horizon.

Abstract: This paper addresses the issue of modifying planners to produce better plans in the face of uncertainty. The traditional planning paradigm has serious flaws in the face of a dynamic and uncertain world. As a result, integrated end-to-end mobile robotic systems must overcome numerous obstacles to integrate planners into a multi-tiered architecture. This paper explores the types of problems caused by uncertainty, and proposes several specific changes to be incorporated into planning systems. A proposal of future work is detailed.

Abstract: TBD

Abstract: Multi-representation models are an attractive design solution for layered autonomous agent architectures in terms of man aging complexity of internal representations and separation of design concerns. However, the representations maintained by the multiple layers can become inconsistent with one another due to the nature of the layers' concerns and capabilities. Consistency maintenance in multi-representation models is an emerging concern in many domains, such as military simulations. We present an approach to consistency maintenance in multi-tier agent architectures. We draw on experience and techniques from the multi-representation modeling community. The benefit of this approach for autonomous agent designers is a conceptual framework through which to organize systems that must deal with temporal and mapping inconsistencies.