This paper presents a method for recognition of 3D objects with curved surfaces from linear tactile data. For every surface model in a given database, a lookup table is constructed to store principal curvatures precomputed at points of discretization on the surface. To recognize an object, a robot hand with touch sensing capability obtains data points on its surface along three concurrent curves. The two principal curvatures estimated at the curve intersection point are used to look up the table associated with each model to locate surface discretization points that have similar local geometries. Local searches are then performed starting at these points to register the tactile data onto the model. The model with the best registration result is recognized. The presented method can recognize closed-form surfaces as well as triangular meshes, as demonstrated through simulation and robot experiments. Potential applications include recognition of (manufactured) home items that are grasped daily, and dexterous manipulation during which recognition is simultaneously performed.

Model-based recognition of an object typically involves matching dense 3D range data. The computational cost is directly affected by the amount of data of which a transformation needs to be found before carrying out the match against a model. This paper investigates recognition using "one-dimensional" data, more specifically, points sampled along three concurrent curves on the surface of an object. The introduced method determines the quality of match against a model in two steps. First, the Gaussian and mean curvatures at the curve intersection point are estimated and used in a table lookup to find multiple candidate points on the model that have similar local geometry. Second, starting at each point, local optimization is conducted to search for a possible location of the curve intersection on the model as well as an orientation that leads to a good match of all data points. The best match between the model and the data curves is chosen over the results obtained from all candidate points. The quality of this match is used for comparison against other models. Simulation and experiment have been conducted to validate the recognition approach

This paper studies the recognition and localization of 2-D shapes bounded by low-degree polynomial curve segments based on minimal tactile data. We have derived differential invariants for quadratic curves and two special classes of cubic curves. Such an invariant, independent of translation and rotation, is computed from the local geometry at any two points on the curve. Recognition of a curve class becomes verifying the corresponding invariant with more than one pairs of data points. Next, the actual curve is determined in its canonical parametric form using the same tactile data. Finally, the contact locations on the curve are computed, thereby localizing the shape completely relative to the touching hand. Simulation results support the working of the method in the presence of small noise, although real experiments need to be carried out in the future to demonstrate its applicability. The presented work distinguishes from traditional model-based recognition in its ability to simultaneously recognize as well as localize a shape from one of several classes, each consisting of a continuum of shapes.