Skip to content

Repository files navigation

Value-Iteration ADP Presentation Toolkit

Source scripts and final artifacts for explaining value-iteration adaptive dynamic programming to a technical audience.

JavaScript Python Slides

This project turns a control-theory paper into a concise Chinese presentation, speaker script, equations, algorithm flow, and practical interpretation. The source material is Wei, Liu, and Lin, IEEE Transactions on Cybernetics (2016).

中文简介:将 value-iteration ADP 论文整理成 10 页中文汇报、配套讲稿和可重复 生成脚本,重点解释有限迭代后控制律“是否可用”的判定问题。

Selected pages from the ADP presentation

Scope

This is a paper-explanation and presentation-engineering project. It does not claim to be a full numerical reproduction of the paper's control experiments.

Deliverables

  • Ten-slide presentation
  • Chinese speaker script
  • build_adp_outline_10slides.mjs: final outline-oriented deck generator.
  • build_adp_speech_docx.py: matching Word script generator.
  • Earlier deck generators are retained to show the iteration process.

Narrative

Nonlinear optimal-control problem
              |
              v
   Value-iteration ADP update
              |
              v
 Convergence is not enough for deployment
              |
              v
  Admissibility-based stopping condition
              |
              v
 Critic/action network implementation

The presentation focuses on the practical gap between asymptotic convergence and a controller that can safely stop after a finite number of iterations.

Topics covered

  • Bellman and value-iteration updates for nonlinear discrete-time systems.
  • Convergence properties and stopping criteria.
  • Admissibility of the finite-iteration control policy.
  • Critic and action network roles.
  • Interpretation of the paper's simulation section.

Regenerating the speaker script

git clone https://github.com/sclm3010/adp-presentation-toolkit.git
cd adp-presentation-toolkit
python -m pip install -r requirements.txt
python build_adp_speech_docx.py

Regenerating the slides

The slide generators were built around the Codex desktop artifact runtime.

pnpm install
node build_adp_outline_10slides.mjs

They locate the bundled runtime under ~/.cache/codex-runtimes by default. If your runtime is stored elsewhere, set CODEX_RUNTIME_ROOT to its dependencies directory. The checked-in .pptx remains available even when that runtime is not installed.

Repository structure

adp-presentation-toolkit/
  build_adp_outline_10slides.mjs  # Final deck generator
  build_adp_report_deck*.mjs      # Earlier design iterations
  build_adp_speech_docx.py        # Speaker-script generator
  assets/adp-deck-preview.png     # README preview
  outputs/                        # Final PowerPoint and Word files

Design principles

  • One central idea per slide.
  • Equations paired with plain-language interpretation.
  • Clear distinction between convergence and finite-iteration usability.
  • Explicit scope boundaries between paper explanation and code reproduction.

About

Paper study and presentation toolkit for value-iteration adaptive dynamic programming.

Topics

Resources

Stars

0 stars

Watchers

0 watching

Forks

Releases

Packages

Contributors

Languages