Task-driven form features recognition: a tool for robotic activity management

One of the major research issues on theoretical problems in robotics is how to endow a robot with the capability to operate in an autonomous manner as is required, for example, to face an unstructured environment, handle error arising from abnormal situations and achieve goals expressed at the task level. This capability is based on the robot´s ability to characterize both the actual environment where it performs its activities and the objects on which it operates; this characterization is performed in terms of the features relevant to the task. These features are related to specific forms in the object or environment and are therefore named form features. The paper proposes to face these problems by adopting an approach, derived from the pattern recognition field, which allows the description of the same entity from several different points of view. In this approach the features of interest were defined using CAILs, a family of rewriting systems.<>