Intrinsic Tactile Sensing and System Inversion II

Siddhartha Srinivasa

Abstract

This is the second of two talks by Matt Mason and Sidd Srinivasa on intrinsic tactile sensing for mobile manipulation. "Intrinsic tactile sensing" refers to a well-known technique to infer tactile information from sensed deformations of a compliant structure. Mike Erdmann, Matt Mason and I are working on applying this idea to mobile manipulation. At present we are studying an approach published by Bicchi, Caiti, and Prattichizzo. They developed a dynamical system model of the structural deformation, with the applied forces as inputs and the sensor readings as outputs. Intrinsic tactile sensing then becomes a system inversion problem--to map from outputs (sensor readings) back to inputs (applied forces).
I will quickly review the basic concepts of system inversion for linear systems and focus on a numerical technique for system inversion by framing the process as an initial value problem for a Differential Algebraic Equation (DAE). I will present a technique for solving DAEs using orthogonal basis functions and discuss the results of applying it to the problem of system inversion for the deformation of compliant structures.

Presentation Slides

Pdf (1.5Mb)

 

Speaker Appointments

For appointments, please contact Siddhartha Srinivasa (siddh@cs.cmu.edu).