Soft Robotic Mannequin with Programmable Shape Deformation

Abstract

A mannequin is an articulated model commonly used in the garment industry for displaying clothing or assessing garment fit. Recently, there has been growing concern about the significant material waste—over half of fast fashion produced is discarded within a year [1], [2]—and the substantial environmental pollution associated with the traditional size-to-shape fitting system and unsustainable garment production methods [1]. This has led to increased attention on sustainable garment fabrication, as exemplified by innovations such as Dutch designer Aniela Hoitink’s novel garment solution NEFFA [3] and Manel Torres’s spray-on dress [4] showcased at Paris Fashion Week. These methods involve spraying biodegradable materials, like mycelium, onto a fixed mannequin to create perfectly fitting garments. In this thesis, a soft mannequin with programmable deformation is proposed as a deformable mold to enhance sustainable material deposition on free-form surfaces. This custom-designed deformable robotic mannequin utilizes soft robotics to achieve substantial localized deformation and adaptability to various shapes. Unlike conventional robots with rigid linkages, the deformable surface in this system is made of soft materials, which introduces distinct challenges in design and fabrication. Moreover, accurately modeling surface deformation with infinite DoFs remains a significant challenge. In contrast to traditional soft robots, such as bending manipulators, where reduced analytical models are often sufficient to describe task space variables (e.g., end-effector motion [5], [6]). However, modeling the whole configuration space of the soft robot with highly deformable free-form surfaces introduces substantial complexity, which will be addressed systematically in this thesis. Overall, this thesis first presents the design, fabrication, and model-free control of the soft robotic mannequin. The free-form surface is divided into multiple pneumatic actuators, known as chambers, each composed of a solid lower layer and a flexible upper silicone layer, connected through microstructural bonding between the soft and rigid materials. A carefully designed mold is used to inject silicone liquid, forming the upper layer. This design and fabrication process enables the mannequin to deform into various shapes. The deformation control method is introduced by iteratively adjusting the actuation parameters, with the Jacobian calculated on the physical hardware in a model-free approach. Additionally, to model the large deformations in soft robot systems in the virtual environment, including the soft mannequin, a geometric simulator with collision-handling capabilities is proposed. This simulator transforms pneumatic actuation and collision 17 interactions into geometric variables, which are then resolved through an optimization pipeline. Serving as a rapid design tool, this collision-aware simulator also enables virtual verification of the IK algorithm. However, simulation results naturally face discrepancies with real-world performance, particularly for free-form surface morphing with infinite DoFs. To address this challenge, a function-based Sim-to-Real pipeline is proposed, which learns the function space and translates surface deformations into real-world cases using limited data, even in the presence of marker-missing issues. This approach improves both the efficiency—thanks to fast analytical gradient evaluation through backpropagation in the Sim-to-Real network—and the accuracy of the IK. For the actuation system, OpenPneu is developed to ensure stable pressure regulation across various pneumatically actuated soft robots, including soft mannequins. It features a modular design concept that allows for scalability, making it adaptable to soft robots with multiple chambers. To demonstrate the practical potential of this soft robotic mannequin, sustainable garments are fabricated using various deformations of the mannequin, highlighting its ve

Date of Award	6 Jan 2025
Original language	English
Awarding Institution	The University of Manchester
Supervisor	Charlie C.L. Wang (Supervisor) & Andrew Weightman (Supervisor)