Dual arm peg-in-hole force/torque values during contact

In order to evaluate deep neural networks for contact state recognition during a bi-manual peg in hole assembly, a data set is gathered during an error recovery sequence when the peg and the hole are misaligned from the center. This experiment only covers 8 concentric error positioning class (y) and for every class the F/T vector of 12 features that correspond to the six component of sensor 1 and six component of sensor 2. Further explanation can be located on the research article. One file used for trainning has 2056 examples/rows and the second file used for validation has 159 examples/rows.