Memory-Augmented Reinforcement Learning Agent for CAD Generation

2026-05-19

AI Summary

This work addresses the challenges faced by current large language models in generating complex CAD models, where long operation sequences, diverse action types, and stringent geometric constraints often lead to broken reasoning chains and an inability to self-correct. To overcome these limitations, the authors propose a memory-augmented reinforcement learning framework that encapsulates the geometric kernel into a structured toolchain, establishing a closed-loop mechanism encompassing design intent understanding, global planning, execution, and multi-dimensional validation. A key innovation is the introduction of a dual-track memory module—comprising a case repository and a skill buffer—coupled with a dynamic utility-based retrieval algorithm, enabling online self-correction and continual learning. This approach effectively avoids outputs that are semantically plausible yet geometrically infeasible, achieving significantly higher success rates and geometric consistency in complex CAD generation tasks without requiring additional large-scale annotated data.

Abstract

Automatic generation of computer-aided design (CAD) models is a core technology for enabling intelligence in advanced manufacturing. Existing generation methods based on large language models (LLMs) often fall short when handling complex CAD models characterized by long operation sequences, diverse operation types, and strong geometric constraints, primarily because reasoning chains break and effective error-correction mechanisms are lacking. To address this problem, this paper proposes a memory-augmented reinforcement learning framework for CAD generation agents. The framework encapsulates the underlying geometric kernel into a structured toolchain callable by the agent and builds a closed-loop mechanism of design intent understanding, global planning, execution, and multi-dimensional verification. It also designs a dual-track memory module consisting of a case library and a skill library, and proposes a dynamic utility retrieval algorithm. By introducing reinforcement learning into retrieval and policy optimization, the agent can effectively avoid retrieval traps in which examples are semantically similar but geometrically infeasible, enabling online self-correction and continual evolution without additional large-scale annotated data. Experiments show that the proposed method significantly improves both the success rate and geometric consistency on complex CAD model generation tasks.

Problem

• CAD generation
• complex geometric constraints
• reasoning chain breakdown
• error-correction mechanism
• long operation sequences

Innovation

• memory-augmented reinforcement learning
• CAD generation
• geometric constraints
• dual-track memory
• dynamic utility retrieval