This SGER project is a feasibility study of algorithms for manipulation and recognition using haptic information. The Manipulation-and-Perceiving-Simultaneously (MAPS) paradigm borrows key ideas from recent advances in simultaneous localization and mapping (SLAM) and visual object recognition with modifications necessary to adapt them to haptic exploration.
In MAPS, exploration is driven by the objective of locating "interest areas" that carry high information content about the object. The haptic properties of these interest points will be encoded, together with location information, in a coordinate-system independent fashion. During a learning phase, a large set of such points will be acquired and stored in a database. During recognition, newly acquired points will be compared with those in the database to generate plausible model hypotheses. The final recognition decision will also incorporate the geometric constraints implicit in the location and orientation of the surface structure.
The key contributions of this project will be: - The development of active exploration strategies to extract, localize and map distinctive surface features; - The development and evaluation of haptics-based object recognition algorithms using both spatial and haptic "appearance" information; - The development of an experimental environment to implement and evaluate MAPS algorithms.
