In this work, we develop a general solution to a broad class of grasping and manipulation problems that we term as C-SLAM for contact simultaneous localization and modeling, where the robots need to accurately track the motions of the contacted bodies and the locations of contacts, while simultaneously estimating important system parameters, such as body dimensions, masses and friction coefficients between contacting surfaces. Our solution framework is based on a dynamic Bayesian inference framework, and hence, we refer to it as Dynamic Bayesian C-SLAM (DBC-SLAM). DBC-SLAM combines an NCP-based dynamic model with the dynamic Bayesian network, and incorporates model parameter estimation as an intrinsic part of the overall inference procedure. We show two preliminary “proof-of-concept” examples that demonstrate the use of DBC-SLAM in robotic contact tasks.

Advanced grasp control algorithms could benefit greatly from accurate tracking of the object as well as an accurate all-around knowledge of the system when the robot attempts a grasp. This motivates our study of the G-SL(AM)2problem, in which two goals are simultaneously pursued: object tracking relative to the hand and estimation of parameters of the dynamic model. We view the G-SL(AM)2problem as a filtering problem. Because of stick-slip friction and collisions between the object and hand, suitable dynamic models exhibit strong nonlinearities and jump discontinuities. This fact makes Kalman filters (which assume linearity) and extended Kalman filters (which assume differentiability) inapplicable, and leads us to develop a particle filter. An important practical problem that arises during grasping is occlusion of the view of the object by the robot's hand. To combat the resulting loss of visual tracking fidelity, we designed a particle filter that incorporates tactile sensor data. The filter is evaluated off-line with data gathered in advance from grasp acquisition experiments conducted with a planar test rig. The results show that our particle filter performs quite well, especially during periods of visual occlusion, in which it is much better than the same filter without tactile data.