Two types of ambiguity are recognizable in Dempster and Shafer's theory of evidence. One is the ambiguity of the system itself and the other is the ambiguity of human knowledge or information about the system. In this paper, from this point of view, the property for measuring the ambiguity of the system and the property for measuring the ambiguity of human knowledge or information are discussed. A global measure of information which satisfies both of these properties is proposed. The unicity of the, proposed measure is proved when it must be an extension of Shannon entropy. The soundness of this measure of information is confirmed by a simple experiment.
This paper is concerned with an extended photometric stereo method using polarimetry. The photometric stereo can be utilized for surfaces with non-uniform in its reflectance properties in case of the surfaces are Lambertian. Reflectance properties can be represented with two components : specular components and diffuse components. When the specular components of reflection are suppressed by polarimetry and only diffuse components exist, the surface of objects can be approximated to Lambertian. This paper presents experimental results of reflectance properties under polarized light, and the extended photometric stereo for Lambertian surfaces. Surface orientation of objects with unknown and uneven reflectance properties can be extracted by this extended photometric stereo.
We propose a concurrent layered rule-based system for an advanced computerized control. The layered structure is organized as a group of rulesets that work concurrently with various cycles and intercommunicate with each other via variables in working memory. A ruleset having a short execution cycle performs a real-time task such as calculation of a control algorithm and executes a complex computation in the background, so that the system can perform complex tasks in process operation without violating real-time constraints. We also show this rulebase can be implemented with a simple shell that contains a scheduler in an inference, engine and can be executed on a common process computer.
The quadratic stability theory is known as a powerful method for robust stabilization of uncertain systems. This paper applies it to the tracking problem from the viewpoint of the inverse regulator problem and proposes a method of designing robust tracking controllers for linear systems with uncertain parameters. First, we solve the inverse problem of a quadratically stabilization problem. We then apply this result to Inverse LQ design theory and then give a criterion of choosing design parameters assuring quadratic stability of the closed system. Finally we show an example to illustrate the validity of the design method.
Large-scale nonlinear programming problems usually contain a relatively small number of nonlinear variables. For such a problem, it is often effective to treat the linear part as a linear programming problem by temporarily fixing the nonlinear variables. As a result, we obtain a nonsmooth optimization problem containing the nonlinear variables only. Based on this idea, we propose a successive quadratic programming algorithm for large-scale nonlinear programming problems. We show convergence of the algorithm and report some computational experience.