• Design and control of dextrous robot hands (@ Stanford)
  [Photos]

  When I studied the issues related to designing viable robot hands, what became immediately apparent was the fact that robot hands had a hard time manipulating objects. Examination suggested two reasons: the hardware used to build them were quite different than those humans possess ("hard" vs "soft") and the models used in the underlying controls assumed point contacts instead of the more realistic area contact. So, my thesis was focused on building softer hands and on better modeling the hands. One of the approaches we took -- that of using plasticity models from soil mechanics -- was novel and won us an IEEE Philips Best Paper Award.

• Controlling the transition of robots from/to contact with the environment (@ MEL, Japan and at the Univeristy of Tokyo, Japan)
  

  As robot fingers open and close there grasp on objects, transitions occur in the dynamic and control models. So, my time in at MEL was spent working with colleagues from the University of Tokyo on controlling these transitions from and to contact. We identified intrisincally stable controllers so discontinuities would not destabilize the system.

• Modeling macro-micro manipulation (@ MEL, Japan)
  [Photos]

  While in Japan, I noticed a trend towards MEMS devices and a desire to use them for extremely delicate tasks (e.g., surgery). So, a question a graduate student of mine posed was "What will it take for the surgeon to "feel" (in grams of force) something that is taking place at a nano scale?" When we started with a modeling simpler problems like manipulating silicon parts in-situ, it turned out that electro-statics has a fundamentally different behavior to Newtonian mechanics. So, we built ourselves simple manipulators on the human side and simulated the electro-static behavior on a computer and identified some simple control models.

• Design, commercialization and establishment of safety standards of robotic devices that augment the capabilities of the human driving the devices. (@ General Motors, with Northwestern University and the University of California, Berkeley) [Photos]
  

  I've always been fascinated by cars; so when GM made me the offer to join them to solve a crucial problem, I jumped at it. Turned out that workers on the line assemble cars in fundamentally the same manner in which Henry Ford did -- pick them up, carry them over, twist and turn while they installed them. However, unlike ford's times when the parts were small, manufacturing economics had moved with times towards delivering sub-assemblies (e.g., an engine, the transmission, the fully built instrument panel, etc) which often weighed as much as 150lbs. So the workers were messing up their backs and the automobile industry was in OSHA's cross hairs. So, I led a team that landed up creating a new class of collaborative robotic devices called "Intelligent Assist Devices" and "Cobots." When I left GM, there were many suppliers providing devices that are being used in automotive assembly plants at GM, Ford and Toyota and in material handling situations like FedEx transfer centers in Memphis, TN. This work won us GM's highest technical award, the "Boss" Kettering Award.

Experimental structural mechanics has traditionally used photoelasticity and Moire techniques to indirectly measure the displacement of beams and plates under load. Many of these techniques need the use of expensive and sensitive holographic plates -- that arent always available in the third world (remember, the India of the early 80's was very different than India today). So, what were looking to do was find alternate ways to record (e.g., on 200ASA photographic paper vs 0.02ASA holographic ) measure with similar accuracies. I had a blast working in the lab for about 8 months trying to perfect the methodology.