Research & Tech Development
Oscillators are physical systems that are capable of complex dynamics. Surprisingly, they can act as non-traditional computers through the use of machine learning, which can convert them to reservoir computers. We built a Hopf oscillator-based reservoir computer that is capable of reconfigurable computing tasks. For a stream of True and False statements shown in (a), the OR, AND, and NOT gates are computed with the reservoir computer. The results of these gates are shown in (b), (c), and (d), respectively.
Adaptive oscillators are a subset of nonlinear oscillators that can intrinsically learn and store information. This information is directly stored in an analog form in a plastically-deformable, dynamic state. For instance, a pendulum (shown on the left) can store frequency content as the length of the pendulum's rod. Adaptive oscillators have unique applications as resonance-tracking broadband energy harvesters, smart robotic gait controllers, and fully analog frequency analyzers.
Biological systems have profound abilities, which rely on the complex relationships between functional morphology & computing. We seek to tap into these abilities through several directions. We want to create enhanced shape memory alloy actuators that are capable of self-sensing, quick responses, large deformations, and optimized trajectories, while still having a small form factor and inexpensive fabrication process. We also seek to unlock the computational abilities of biological tensegrities, such as the one depicted on the right. This simulation highlights our custom discrete element method joint simulator, which directly uses CT scanned images.