Abstract
Robotic motion planning algorithms for manipulation of deformable objects, such as in medical robotics applications, rely on accurate estimations of object deformations that occur during manipulation. An estimation of the tissue response (for off-line planning or real-time on-line re-planning), in turn, requires knowledge of both object constitutive parameters and boundary constraints. In this paper, a novel algorithm for estimating boundary constraints of deformable objects from robotic manipulation data is presented. The proposed algorithm uses tissue deformation data collected with a vision system, and employs a multi-stage hill climbing procedure to estimate the boundary constraints of the object. An active exploration technique, which uses an information maximization approach, is also proposed to extend the identification algorithm. The effects of uncertainties on the proposed methods are analyzed in simulation. The results of experimental evaluation of the methods are also presented.