Abstract:
Fish demonstrate an efficient and agile swimming modality that allows them to navigate complex fluid environments with ease. This has been attributed to complex, intricate flow interactions with the surrounding fluid. However, underwater robots have yet to realize the same performance as their biological counterparts. This thesis addresses some of the challenges behind achieving bio-inspired robotic swimming, which span simulation, design optimization, and control. I will discuss prior work on multiphysics modeling and differentiable simulation of fluid-robot interaction. I will then discuss prior work on data-driven control for nonlinear systems and high-dimensional observations, which are often seen in fluid settings. Finally, I will propose future work that builds upon these simulation and control efforts to advance the design and policy of a bio-inspired robotic swimmer toward fluid-aware locomotion.

Thesis Committee Members:
Zac Manchester​ (Chair​)
Carmel Majidi
Keenan Crane
Robert Katzschmann (ETH Zurich)

More Information