 Hello, my name is Patrick Dills and I'm from the Reach Lab at the University of Wisconsin-Madison. I'm here to talk to you about some recent work focusing on the stability properties of a parallel hybrid actuator we've been working on in the lab. Traditional active-only haptic devices have a limited rendering range. A new class of hybrid devices composed of active and passive actuators arranged in parallel has been experimentally shown to increase the maximum virtual stiffness a haptic device can achieve. Understanding the mechanism and limitations by which the rendering range is increased could inform the design and control future devices. Our hybrid actuator is composed of an active actuator and a passive actuator, in this case a DC motor and a particle brake arranged in parallel, and we developed a model of our hybrid actuator which includes a virtual stiffness, a time delay, the linear dynamics of our actuator, a filter on the measured passive actuator feedback, and a nonlinear doll model representing the nonlinear brake dynamics. These elements are arranged in parallel such that rendering accuracy of the hybrid actuator is maintained. Simplifying our nonlinear brake model using equivalent stiffness and damping leads to large and small amplitude linearized models of our hybrid actuator, as you can see here. Stability analysis of the small amplitude model assuming unfiltered passive feedback results in an inequality expressing the device's passive rendering range in terms of virtual stiffness, damping, time delay, and brake stiffness. And this result is supported by additional analysis using pseudo-delay methods, which shows the system is asymptotically stable independent of delay. Stability results show that the hybrid actuator outperforms an active actuator especially at larger time delays, and reforming the passivity expression reveals the ratio of damping to brake stiffness as an important factor determining the range of delays, where our hybrid actuator is most effective at increasing the device's maximum stable stiffness. However, unfiltered passive actuator feedback is not always possible or even desirable. Filtering passive actuator feedback can increase the range of uncoupled asymptotically stable virtual stiffness beyond the unfiltered case previously examined. Increasing the stable virtual stiffness in this way must be done carefully, though, because low bandwidth filtered feedback and large time delays can distort the system's output impedance. So we tested our hybrid actuator by perturbing the system and incrementally increasing the virtual stiffness until the system becomes unstable, and results under increasing time delays and while varying filtered feedback bandwidth agree well with theoretical results. We were limited to lower filter bandwidths in this configuration, though, due to noise in the passive actuator feedback signal. So we also tested a compliant version of our hybrid approach using encoder feedback to estimate the brake torque and to test a wider range of filter bandwidths and time delays. Again, experimental stability results agree well with theoretical results. So in summary, our hybrid actuator demonstrates a maximum virtual stiffness that's strongly linked to the brake stiffness. Maximum virtual stiffness can be increased again by implementing filtered brake feedback, although low filter bandwidths and large time delays present a limitation on accurately rendering the desired output impedance. And we feel our results demonstrate the rendering advantages of our parallel hybrid actuator, and ultimately we hope that our results will influence future kinesthetic optic devices and control systems.