Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

HeaRT: A Hierarchical Circuit Reasoning Tree-Based Agentic Framework for AMS Design Optimization

Souradip Poddar, Chia-Tung Ho, Ziming Wei, Weidong Cao, Haoxing Ren, David Z. Pan

TL;DR

HeaRT introduces an analytically guided, hierarchical circuit reasoning tree to enable agentic, context-aware AMS design optimization. By offline constructing a DC-current–compliant knowledge graph and online retrieval anchored to a topology-ranked knowledge base, HeaRT achieves high reasoning accuracy and efficient token usage while preserving prior design intent. The framework delivers up to $>3\\x$ faster convergence in sizing and topology adaptation across multiple optimization baselines and demonstrates notable FoM and PCKRI improvements in two design-scenario experiments. This approach paves the way for practical, explainable, and adaptive AMS design automation that leverages human-inspired abstraction and LLM reasoning with strong architectural locality.

Abstract

Conventional AI-driven AMS design automation algorithms remain constrained by their reliance on high-quality datasets to capture underlying circuit behavior, coupled with poor transferability across architectures, and a lack of adaptive mechanisms. This work proposes HeaRT, a foundational reasoning engine for automation loops and a first step toward intelligent, adaptive, human-style design optimization. HeaRT consistently demonstrates reasoning accuracy >97% and Pass@1 performance >98% across our 40-circuit benchmark repository, even as circuit complexity increases, while operating at <0.5x real-time token budget of SOTA baselines. Our experiments show that HeaRT yields >3x faster convergence in both sizing and topology design adaptation tasks across diverse optimization approaches, while preserving prior design intent.

HeaRT: A Hierarchical Circuit Reasoning Tree-Based Agentic Framework for AMS Design Optimization

TL;DR

HeaRT introduces an analytically guided, hierarchical circuit reasoning tree to enable agentic, context-aware AMS design optimization. By offline constructing a DC-current–compliant knowledge graph and online retrieval anchored to a topology-ranked knowledge base, HeaRT achieves high reasoning accuracy and efficient token usage while preserving prior design intent. The framework delivers up to faster convergence in sizing and topology adaptation across multiple optimization baselines and demonstrates notable FoM and PCKRI improvements in two design-scenario experiments. This approach paves the way for practical, explainable, and adaptive AMS design automation that leverages human-inspired abstraction and LLM reasoning with strong architectural locality.

Abstract

Conventional AI-driven AMS design automation algorithms remain constrained by their reliance on high-quality datasets to capture underlying circuit behavior, coupled with poor transferability across architectures, and a lack of adaptive mechanisms. This work proposes HeaRT, a foundational reasoning engine for automation loops and a first step toward intelligent, adaptive, human-style design optimization. HeaRT consistently demonstrates reasoning accuracy >97% and Pass@1 performance >98% across our 40-circuit benchmark repository, even as circuit complexity increases, while operating at <0.5x real-time token budget of SOTA baselines. Our experiments show that HeaRT yields >3x faster convergence in both sizing and topology design adaptation tasks across diverse optimization approaches, while preserving prior design intent.

Paper Structure

This paper contains 19 sections, 5 equations, 7 figures, 3 tables, 1 algorithm.

Table of Contents

  1. Introduction
  2. Preliminaries
  3. LLM-Aided AMS Design Automation
  4. Knowledge Retention and Design Locality
  5. The HeaRT Framework
  6. Analytical-Guided Hierarchical Circuit Reasoning Tree Construction
  7. Graph-Guided Reasoning for Tree Construction
  8. Bottom-Up Lightweight Analysis
  9. Agentic Retrieval
  10. Topology Knowledge Database Management
  11. Reasoning-Guided and Query-Conditioned Traversal
  12. Objective-Driven Retrieval for Topology Optimization
  13. Experimental Results
  14. Experimental Setup
  15. HeaRT's Circuit Reasoning Performance
  16. ...and 4 more sections

Figures (7)

  • Figure 1: An illustration of comparisons of traditional black-box optimization algorithm, vanilla LLM-based approach, and the proposed HeaRT framework for the AMS design task.
  • Figure 2: (a) Overall workflow of the HeaRT framework. (b) Illustrative 8-bit SAR ADC example. (c) Hierarchical circuit reasoning tree generated by HeaRT for the SAR ADC.
  • Figure 3: An illustrative example of HeaRT's analytical-guided circuit decomposition.
  • Figure 4: Statistical summary of our curated dataset repository.
  • Figure 5: Scenario 1: FoM and PCKRI vs. simulation count.
  • ...and 2 more figures