e8da56eb93676e8f60ed2b696e44e7dc-Supplemental-Conference.pdf

The goal location is small region around (20,20). In each task, S0 was a set of arm con gurations establishing contact with the 539 end-effector, the 6-DoF change in stiffness, and 1-DoF gripper state. The fraction of start states in S0 that lead to success 557 IVF, classi er). The result of that execution is recorded as 552 Algorithm 1 is the pseudocode used for the experiments described in Section 4.1. Episodes last a maximum of 1000 steps.