Design and experimental learning of a modular robot platform for fish-inspired swimming

Restricted (Penn State Only)
- Author:
- Deng, Hankun
- Graduate Program:
- Mechanical Engineering
- Degree:
- Doctor of Philosophy
- Document Type:
- Dissertation
- Date of Defense:
- April 19, 2024
- Committee Members:
- Robert Kunz, Professor in Charge/Director of Graduate Studies
Minghui Zhu, Outside Unit & Field Member
Guhaprasanna Manogharan, Major Field Member
Bo Cheng, Chair & Dissertation Advisor
Azar Panah, Special Member
Chris Rahn, Major Field Member - Keywords:
- bio-inspired robotics
fish swimming
motor learning
form and function
Strouhal number - Abstract:
- Evolution has successfully explored various forms of fluid-structure interaction (FSI) for underwater locomotion, from which a great diversity of fish swimming has emerged. Yet, it remains unclear how fish or fish-inspired swimming emerges from the complex physics of FSI which includes interactions among active structures (e.g., fins and elongated bodies), neuromuscular or motor control, and the physics of fluids. It is well known that biological organisms and biologically-inspired robots can leverage embodied physical intelligence to simplify control and improve performance (e.g., robustness). However, no study to date systematically studies the embodied traits in fish or fish-inspired swimming. Understandings of these traits and their implications in motor learning, neuromuscular control, gait generation, and swimming performance have critical importance for advancing the knowledge of the evolution and biomechanics of fish swimming, and will provide informed, principled mimicry of fish swimming in robotics. This dissertation aimed to advance the knowledge in how undulatory swimming emerges from the complex motor-controlled FSI using a fish-inspired robot platform. To achieve the goal, a robophysical model of undulatory swimming (a magnetic, modular, undulatory robot platform, named μBot) was developed. Compared with existing swimming robots, a distinctive aspect of μBot is the modular design which enabled rapid prototyping with different design configurations (such as body length, shape, and stiffness). The robot also has a compact size, allowing the experiments to be conducted in more controlled lab settings. Another feature is that the swimming gaits are not prescribed; instead, they emerge from FSI via motor control (i.e., motor-controlled FSI). Then, motor control programs for various locomotion tasks (e.g., forward/backward swimming, turning maneuver) were developed and experimental setup for motor learning was built. Reinforcement learning (RL) algorithm was applied to optimize the motor control for maximizing the selected locomotion tasks. With the robot platform and experimental motor learning, the effects of the robot design (e.g., body length, caudal fin stiffness) on the emergent swimming gaits and performance were explored. The results showed that: (1) the forward swimming speed increased with the number of actuators or Degrees-of-Freedoms but the normalized speed in body length per second (BL/s) decreased; (2) Optimized forward swimming and turning maneuver in μBots shared an identical caudal fin stiffness; (3) the number of actuators helped the backward swimming speed. In the meanwhile, the embodied properties in the motor-controlled FSI of μBot forward swimming were also studied. The results showed that: (1) the swimming gaits and speed were highly sensitive to the motor control frequency while stayed robust against disruptions applied to the intersegmental phases and intensities of motor control; (2) for all robots and frequencies tested, swimming speed was proportional to the mean undulation velocity of body and caudal-fin combined, yielding an invariant, undulation-based Strouhal number. The Strouhal number also revealed two fundamental classes of undulatory swimming in both biological and robotic fishes; (3) the identical robot actuators were demonstrated to function as motors, virtual springs, and virtual masses. Next, I used time-resolved Particle Image Velocimetry (PIV) techniques to study the frequency modulation of wake structures and thrust generation mechanisms of μBot forward swimming. The results suggested that at low-to-medium frequencies, μBot shed two pairs of vortices per undulation cycle with added-mass force as the main thrust source. At high frequencies, the robot had a Karman vortex wake structure with pressure force as the main thrust source. In addition, μBot was upgraded to a fully autonomous swimming robot. Equipped with inertial measurement units (IMU) sensors for feedback control, the robot had the capability to correct its heading and follow desired paths, suggesting its potential for functioning in various aquatic applications. These results provide novel insights in understanding fish-inspired locomotion and guiding principles for novel bio-inspired robot design.