2-D contact detection and localization using proprioceptive information

Knowledge of contact properties is important to interact with an unknown object in the environment. A method is presented to provide this information for a planar manipulator using general motions and the reading of joint angle and torque sensors. No complex hardware, such as a tactile sensor array, is required. In a first stage, the algorithm combines geometric evidence of the instantaneous center and a term for the intersection with the previous link postures in a filter to obtain a hypothesis of the contact location on the link. Torque domain evidence is used to verify this position estimate. Using this scheme, the algorithm can detect the contact location, surface normal, and can estimate the contact force for each link