TO-Agents: AI Framework for Preference-Guided Topology Optimization
TO-Agents is an AI framework utilizing multiple agents that links natural-language design intentions with iterative topology optimization. It transforms problem descriptions given by humans into validated inputs for solvers, executes a topology optimization solver, generates 3D topologies, and employs vision-language reasoning alongside an independent judge agent to assess outcomes and adjust parameters. The framework underwent testing on two design challenges: a cantilever beam benchmark and a phone stand product design, where designers indicated aesthetic preferences for hierarchically branched structures modeled after natural tree forms.
Key facts
- TO-Agents is a multi-agent AI framework for topology optimization.
- It connects natural-language design intent with iterative topology optimization.
- The framework converts human-provided problem descriptions into validated solver inputs.
- It runs a topology optimization solver and renders 3D topologies.
- It uses vision-language reasoning with an independent judge agent to critique results.
- The framework was evaluated on a cantilever beam benchmark and a phone-stand product design.
- Designers specified aesthetic preferences for hierarchically branched structures inspired by natural tree morphologies.
- The paper is available on arXiv with ID 2605.21622.
Entities
Institutions
- arXiv