Abstract The authors report on a series of experiments to automate the assessment of document qualities such as depth and objectivity. The primary purpose is to develop a quality‐sensitive functionality, orthogonal to relevance, to select documents for an interactive question‐answering system. The study consisted of two stages. In the classifier construction stage, nine document qualities deemed important by information professionals were identified and classifiers were developed to predict their values. In the confirmative evaluation stage, the performance of the developed methods was checked using a different document collection. The quality prediction methods worked well in the second stage. The results strongly suggest that the best way to predict document qualities automatically is to construct classifiers on a person‐by‐person basis.

We introduce computational support tools for the analysis and design of systems with multiple frictional contacts, with a focus on applications to part feeding and assembly processes. The tools rely on dynamic models of the processes. We describe two approaches to modeling, the Stewart-Trinkle model (1996) and the Song-Pang-Kumar model (2003), that allow the designer to experiment with different geometric, material and dynamic properties and optimize the design for performance. In order to accommodate contact transitions, we introduce a smooth cone model for friction. We illustrate the models and the design process by describing the design optimization of a part feeder.

We describe a framework for controlling and coordinating a group of robots for cooperative manipulation tasks. The framework enables a decentralized approach to planning and control. It allows the robots to approach the object, organize themselves into a formation that will trap the object, and then transport the object to the desired destination. Our controllers and planners are derived from simple potential fields and the hierarchical composition of potential fields. We show how these potential field based controllers and planners benefit complex group interactions, specifically for manipulating and transporting objects in the plane. Theoretically, we show how we can derive results on formation stability with potential field based controllers in many cases. Simulation results demonstrate successful application to a wide range of examples without showing sensitivity to parameters. Because the framework is decentralized at both trajectory generation level and the estimation and control agent level, our framework can potentially scale to groups of tens and hundreds of robots.

In this paper we consider the dynamics and control of whole arm grasping systems. We develop a control scheme that employs a minimal set of inputs to control the trajectory of the system while using the surplus inputs to control the interaction forces in order to maintain the unilateral constraints at both rolling and sliding contacts. Since the number of surplus inputs is less than the number of output force variables, we propose a controller that controls the critical contact force components. We emphasize the dynamic models and algorithms for computing contact forces, which are crucial to the development of the control algorithms. Finally, we show how compliant contact models and an integrated simulation approach are used to overcome the difficulties with uniqueness and existence of solutions. A planar whole arm manipulation system is used as an example to illustrate the basic ideas.

We propose a novel approach to dynamic simulation of grasps that overcomes difficulties due to inconsistencies in the forward dynamics problem. The key idea in the paper is a minimalist model of the contact compliance and the use of this model in situations when the performance of the rigid body model is not satisfactory. Our general framework allows for online diagnostics that enable the automatic switching between models to maximize efficiency while avoiding ambiguous situations. We illustrate the basic ideas by simulating the dynamics of several whole arm grasps.