Eric Rombokas is interested in how brains control movement and sense their environment, and how these principles can be used to control robots and interface with next-generation prosthetic limbs. Today's robots operate in carefully controlled environments, and are specialists for specific tasks. Robots that operate in unstructured environments, interact with people, or perform manipulation of arbitrary objects, however, remain open topics of research. Prosthetic limbs exemplify the most challenging problem domain: they are expected to perform in close collaboration with the human user under unforgiving weight and power requirements in completely unstructured environments.

Dr Rombokas has been developing new ways of approaching the problem, allowing unprecedented control of a one-of-a-kind tendon-driven robotic hand, the ACT Hand. Instead of motors at the joints, as one would find in a factory robot, this robot operates like the actual human hand, by pulling tendons. The tendon network is made to precisely mimic human biomechanics, sliding over mechanical bones made from laser scans of human bones. This “biological” robot system provides a testbed for studying how humans move their bodies, while the methods used to control it offer a different way forward for robotics and active prosthetic limbs.  By relying on data from actual robot operation instead of idealizations of the physics, he has shown that classical approaches to control theory can be merged with current neuroscience understanding to produce practical behavior for previously-intractable applications.  These computational methods have also found application in interface with assistive robots and sensory substitution for powered prosthetic limbs.

Fundamentally, we seek for prosthetic devices to achieve the same thing that natural limbs do: to operate in the real world with intelligence, performing movements in concert with the body of the user, while providing feedback that produces a subjective feeling of “self-ownership.”  Dr. Rombokas is taking advantage of the massive parallel computation afforded by computer graphics cards (GPUs) to generate intelligent cost-minimizing behavior in real time. 

Despite promising results for providing sensory feedback for upper-limb prosthetic devices, there has been little work in lower limb.  Of particular interest is stair descent because the forces relevant to placement of the foot are not transmitted through the pylon and socket, so users must use vision to make sure they place their foot correctly.  Haptic feedback of forces sensed by sensors along the toe-heel axis of the shoe has the potential to improve confidence on stairs and prevent falls. 

Similarly, the "targeted reinnervation" (TR) surgical procedure has shown great potential for upper-limb amputees, and is only beginning to be applied to lower-limb amputees.  It is a technique for redirecting amputated nerves to intact muscles and skin, creating new junctions for both motor output and sensory capabilities.  Dr Rombokas and collaborators are developing sensory feedback for stair descent on a cohort of TR recipients.  Haptic feedback to the TR site is expected to be more intuitive to interpret, potentially improving functional movement but also improved "sense of self" when using the prosthetic device. 

Eric has also been working on the sensing and control involved in diverse domains such as insect flight and human-computer gesture interaction.  He has created a novel flying platform inspired by the hawkmoth Manduca Sexta, which uses inertial maneuvering to fly as dextrously as conventional uav rotorcraft using fewer actuators.  He is also exploring the use of electromyography (sensing the electrical activity of muscles) to amplify the utility of gesture-sensing cameras.