Optimal tactile sensor placement

By intelligently locating a tactile sensor with respect to a sensed object it is possible to minimize the number of sensed points required to identify or localize the object. The author applies principles of statistical decision theory to determine the optimal sensing location to constrain an object model, maximally based on any prior object information, including models or previously sensed points. He shows how information about an object´s shape can be combined with sensory data to produce a probablistic membership function on the workspace. Utility functions on the workspace are derived which qualitatively describe the constraining value of obtaining sensory data at each location in the environment. Also described are techniques of sequential analysis which are used to process sensory information as it is acquired. An implementation of these principles is presented, namely, a two-dimensional sensing problem, using a camera with a restricted field of view to acquire sparse sensory data, which are used to discriminate the identity of a shape from among a given set