Abstract
dc:description.abstractMachine knitting is an emerging form of additive manufacturing capable of producing complex, functional textile structures with applications across apparel, soft robotics, architecture, and healthcare. However, designing machine-knittable patterns for arbitrary 3D surfaces remains a difficult and expertise-intensive task: stitch layouts must satisfy strict geometric and topological constraints, and current pipelines rely heavily on manual editing or fragile heuristics. This thesis develops a sequence of geometric and topological tools that collectively transform knit-graph construction into a principled, globally informed process. Building on the popular knit graph abstraction, we interpret stitch planning through the lens of discrete 1-forms and vector field analysis, enabling robust, customizable, and manufacturable knitting structures. Our first contribution introduces a stripe-generation framework formulated directly in the space of discrete 1-forms. This representation allows direct control over stripe singularities, alignment of course and wale directions with geometric features, and enforcement of manufacturability through simple linear constraints. We then extend this framework by interpreting stripe patterns as oriented foliations of an undervilying vector field, yielding a topological understanding of stripe behavior and enabling the use of a single type of linear constraint to enforces all manufacturing guarantees. This foliations-based formulation introduces automatic singularity matching, an effective stripe interpolant for robust tracing, and theoretical guarantees on the global structure of resulting knit graphs. To complete an end-to-end pipeline for knit graph generation, our next contribution presents a fully automatic method for placing knit singularities through curl quantization of normalized vector fields derived from a knitting time function. By iteratively inserting course and wale singularities in regions of high curl, this approach eliminates the need for mixed-integer optimizations, manual singularity placement, and post-hoc matching, while producing smooth, helix-free knit graphs with high geometric fidelity. This curl-based perspective also enables intuitive user control through direct curl signal editing. Building on this formulation, our final contribution leverages optimal transport theory to place knit singularities globally rather than greedily, improving both efficiency and scalability for generating knittable models of larger sizes and complex geometries. Beyond knitting, the methods developed in this thesis have broader implications for fabrication-aware geometry processing. The proposed stripe-based frameworks can be modified to naturally produce globally coherent and self-supporting spacef illing curves that can be used as toolpaths for 3D printing, ceramic extrusion, and other fabrication processes. Together, these contributions form a unified geometric pipeline for customizable, manufacturable knit-graph generation, while providing new structural insights with potential impact across a wide range of computational fabrication domains.
Degree
thesis:*- Name thesis:degree_name
- Doctor of Philosophy
- Level thesis:degree_level
- doctoral
- Discipline thesis:degree_discipline
- Computer Science
- Grantor
- Boston University
- Year dc:date.issued
- 2026
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Mitra, Rahul
- Advisor dc:contributor.advisor
-
- Chien, Edward
- ORCID dc:identifier.orcid
Subjects
dc:subject × 4Rights
- Language dc:language.iso
- en_US
Identifiers
dc:identifier.*- Handle dc:identifier.uri
- https://hdl.handle.net/2144/53720