🐟 Project: Optimization for Soft Robotic Swimmers
UC San Diego · 2022–2024
Embedded Files
Challenge:
Existing robotic fish designs were mostly manually tuned, lacked performance optimization, and were slow to iterate due to physical prototyping constraints.
Designing robotic fish is a multi-disciplinary challenge.
What I Built:
Developed a gradient-based optimization framework to co-design the body geometry and swimming motion of soft robotic fish.
Integrated adjoint derivative computation into an MDO workflow for high-dimensional, scalable optimization.
Implemented and tested three types of optimization:
Kinematics-only
Co-design (kinematics + geometry)
Multipoint co-design (for multiple speeds + turning)
Impact:
Up to 41.7% reduction in cost of transport (CoT) compared to baseline (kinematics-only designs).
Demonstrated that co-design outperforms kinematics-only optimization across all tested speeds (0.3–0.9 m/s).
fish_2pt5_hrz.mp4fish_4_hrz.mp4
Preliminary results for a swimming fish using vortex panel method
Yan, J., …, Tolley, M.T., Hwang, J. T., robotic fish optimization paper (in preparation)
3d printing and testing wave propagation of the fish
VID_20230718_142020.mp4Report abuse