Research map / Learning theory and algorithms / Prediction and decision losses
Blending Neural Control Density Functions for Stabilization and Safety
Sahil Chaudhary, Chaitanya Murti, Chiranjib Bhattacharyya
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add the exponential parametrization of density and control (Eq. 18) instead of a/b and c/a ratio parametrization (ablation) | NCDF on system (33) · Prajna et al. system (33) control synthesis | stabilization | Improved |
| blend learned NCDF controllers with the SoS controller of Prajna et al. | NCDF controllers · Prajna et al. systems (33)-(35) | region of attraction | Improved |
| use neural density function instead of neural Lyapunov function | stability certification · system (24) with four equilibria | stability certification | Improved |
| use NCDF instead of neural Lyapunov function | stability certification · system (32) with two equilibria | stability certification | Improved |
| use NCDF instead of Yang et al. 2024 for certified RoA | NCDF · Inverted pendulum | certified region of attraction | Improved |
| use NCDF instead of Yang et al. 2024 for certified RoA | NCDF · Path tracking | certified region of attraction | No clear effect |
| call the falsifier every 500 iterations instead of every 100 (ablation) | NCDF on inverted pendulum · Inverted Pendulum falsification frequency | total learning time: 3m 23s -> 23s | Improved |
| call the falsifier every 250 iterations instead of every 100 (ablation) | NCDF on inverted pendulum · Inverted Pendulum falsification frequency | total learning time: 3m 23s -> 3m 53s | Hurt |
| call the falsifier every 750 iterations instead of every 500 (ablation) | NCDF on inverted pendulum · Inverted Pendulum falsification frequency | total learning time: 23s -> 22.4s | No clear effect |
Closest papers by meaning
- Lyapunov-stable Neural Control for State and Output Feedback: A Novel Formulation (ICML 2024 · similarity 0.88)
- Two‑Stage Learning of Stabilizing Neural Controllers via Zubov Sampling and Iterative Domain Expansion (NeurIPS 2025 · similarity 0.87)
- Stability Analysis of Switched Linear Systems with Neural Lyapunov Functions (AAAI 2024 · similarity 0.85)
- Safe and Stable Control via Lyapunov-Guided Diffusion Models (NeurIPS 2025 · similarity 0.85)
- FESSNC: Fast Exponentially Stable and Safe Neural Controller (ICML 2024 · similarity 0.84)
- Certifying Stability of Reinforcement Learning Policies using Generalized Lyapunov Functions (NeurIPS 2025 · similarity 0.84)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
