Haptic Interface Design and Control
One of the main competencies in my lab at ETH Zürich is the design and control of haptic interfaces. Primarily, my colleagues are interested in developing robots that can be used to investigate the changes that occur in the brain following a stroke, as well as to provide devices that can be used as part of a rehabilitative training and assessment paradigm. In addition, we have developed a course geared toward upperclassmen and graduate students on physical human-robot interaction via the “ETHZ Haptic Paddle” (Gassert et al, IEEE Trans. Education 2013).
ETHZ Haptic Paddle
First Version
The ETHZ Haptic Paddle is a one degree-of-freedom torque-controlled interface that was developed in our lab for teaching the practical component of our course on physical human-robot interaction (Figure 17). In the initial version shown, no direct force measurements were provided at the output. This meant that the interaction force had to be estimated from the current supplied to the motor, which resulted in low performance.
Second Version
For the second version, I was tasked with finding a low-cost way to instrument the paddles for direct output force measurement. Using an approach similar to the flywheel test bench above, finite element analysis was used to find the location and magnitude of maximum strain under full load (Figure 18). Single grid linear strain gages were bonded to the paddle on opposite sides at this location and wired to form half a Wheatstone bridge. The magnitude of strain aided the selection of an appropriate gain for the instrumentation amplifier, which was later integrated with the current amplifier PCB designed by my lab colleague. We have now instrumented all of our paddles in this manner, as well as the ones we delivered to our collaborators teaching a similar course at the other Swiss Federal Institute of Technology, EPF Lausanne.
Hybrid Ultrasonic Motor—Clutch Actuator (HUCA)
Electromechanical actuators and human joints achieve peak energetic efficiencies at very different operating points. Electric motors prefer high rotational speeds at low torque, while human joints prefer low speeds at high torque. As a result, a prosthesis or exoskeleton powered by an electric motor will require a high speed reduction transmission in order to achieve the desired magnitude of torque. However, the presence of the transmission reduces the ability of the motor to render a desired torque trajectory, and can also result in tremendous shock loading during an unexpected event, such as tripping or stumbling.
To investigate a possible workaround, I borrowed a haptic interface developed by one of my colleagues dubbed the Hybrid Ultrasonic Motor-Clutch Actuator (Figure 19). The input to the system is an ultrasonic motor, which, by the nature of its operation, can only be controlled in velocity (i.e. it is not possible to control its torque). In order to control the torque at the output knob, a differential gear is placed between the motor and the knob. A magnetic particle brake is connected to the third port of the differential. The brake acts to dissipate input power from the motor and, by controlling the torque applied by the brake, it is possible to control the torque at the output knob. This topology may provide a solution to the torque control problem in prostheses and exoskeletons mentioned above (Tucker et al., Eng. Med. Biol. Conf, 2012).