Modeling and algorithms for sensorimotor control of a soft continuum arm

The octopus's arm is a marvel of natural engineering, combining hyper-flexibility, infinite degrees of freedom, and sophisticated sensorimotor control. Its ability to perform complex tasks—such as reaching, grasping, and manipulating objects—with minimal central brain involvement has inspired roboticists to explore bio-inspired designs for soft continuum manipulators. This thesis addresses the challenge of modeling and controlling such arms, drawing inspiration from the octopus to develop algorithms for sensorimotor control in soft robotics. Bend propagation is one of the stereotypical octopus arm movements, where a localized bend travels from the base to tip during reaching motions. This maneuver is energy-efficient and robust, yet its underlying mechanics and control principles remain incompletely understood. Part I of this thesis investigates such phenomenon through a control-oriented model based on Cosserat rod theory. By reducing the arm's infinite-dimensional dynamics to a low-dimensional system, we identify key parameters for bend propagation and validate the model against experimental data. We then took the first step to propose a novel sensory feedback control law inspired by pursuit strategies in nature which reproduces life-like bend propagation motions. The octopus arm's versatility stems from its distributed neuromuscular system, which integrates sensing, actuation, and local computation. Part II develops a comprehensive model of this soft continuum arm system, incorporating the peripheral nervous system (PNS) and muscle dynamics. We integrated the sensory feedback control law at a neural level with a consensus algorithm for sensing that enables the arm to reach stationary targets, leveraging sensory inputs such as chemosensing and proprioception. Analytical and numerical results demonstrate the stability and effectiveness of these sensing and control strategies. Translating biological principles into robotic applications requires efficient algorithms for real-time control and sensing. Part III focuses on posture reconstruction for soft continuum arms, a critical task for closed-loop control. We introduce a physics-informed deep neural network framework that reconstructs the smooth shape and strain of the arm from sparse marker data, enabling real-time performance without costly labeled training. The method is validated on both simulated and physical soft robotic arms, showcasing its practical utility. The interdisciplinary study in this thesis bridges biology and engineering, offering new insights into bio-inspired systems and advancing the capabilities of soft robots across various applications.

Name thesis:degree_name

Ph.D.

Level thesis:degree_level

Dissertation

Discipline thesis:degree_discipline

Mechanical Engineering

Grantor

University of Illinois Urbana-Champaign

Year dc:date

2025

Author dc:creator

• Wang, Tixian

Contributors dc:contributor

• Mehta, Prashant G.
• Yim, Justin K.
• Gillette, Rhanor
• Halder, Udit

• Cosserat Rod
• Octopus
• Soft Robotics
• Distributed Sensing
• Sensorimotor Control
• Shape Reconstruction

Statement dc:rights

• Copyright 2025 Tixian Wang

Language dc:language

en, eng

Handle dc:identifier

https://hdl.handle.net/2142/130011

OAI record oai:identifier

https://www.ideals.illinois.edu/oai-pmh?verb=GetRecord&metadataPrefix=etdms&identifier=oai%3Awww.ideals.illinois.edu%3A2142%2F130011