Research map / Probabilistic and scientific modeling / Molecules and proteins
SAD-Flower: Flow Matching for Safe, Admissible, and Dynamically Consistent Planning
Tzu-Yuan Huang, Armin Lederer, Dai-Jie Wu, Xiaobing Dai, Sihua Zhang, Hsiu-Chin Lin, Shao-Hua Sun, Stefan Georg Sosnowski, Sandra Hirche
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 control-augmented flow (CBF/CLF controller) to flow matching | FM · Hopper (Med-Expert) | dynamic consistency violation: 0.23 -> 0.01 | Improved |
| add control-augmented flow to flow matching | FM · Walker2d (Med-Expert) | safety violation: 0.40 -> 0.00 | Improved |
| add control-augmented flow to flow matching | FM · Hopper (Med-Expert) | reward: 1.02 -> 0.93 | Mixed |
| add control-augmented flow to flow matching | FM · Maze2d (Large) | reward: 2.62 -> 2.70 | No clear effect |
| use QP-based control with delayed activation instead of DPCC non-convex optimization (ablation) | SAD-Flower · Hopper (Med-Expert) | computation time [s]: 4.24 -> 0.06 | Improved |
| use QP-based control instead of DPCC non-convex optimization (ablation) | SAD-Flower · Hopper (Med-Expert) | reward: 0.61 -> 0.93 | Improved |
| replace training-free constraint enforcement with reject-sampling on FM (ablation) | FM + reject-sample · Hopper (Med-Expert) | safety violation: 0.00 -> 0.14 | Hurt |
| replace training-free constraint enforcement with reject-sampling on FM (ablation) | FM + reject-sample · Maze2d (Large) | safety violation: 0.00 -> 0.16 | Hurt |
| add CLF dynamic-consistency term to CBF-only controller (ablation) | SAD-Flower · Maze2d | local trap / dynamic consistency | Improved |
| increase ODE integration steps (25 to about 100) (ablation) | SAD-Flower · Maze2d | dynamic consistency violation | Improved |
| delay controller activation time T0 to larger values (ablation) | SAD-Flower · Maze2d | reward | Improved |
| vary constraint strength c from 1.0 down to 0.4 (ablation) | SAD-Flower · Maze2d (Large) | reward: 1.65 -> 1.50 | No clear effect |
Closest papers by meaning
- SafeFlowMatcher: Safe and Fast Planning using Flow Matching with Control Barrier Functions (ICLR 2026 · similarity 0.88)
- On the Guidance of Flow Matching (ICML 2025 · similarity 0.85)
- Optimal Flow Matching: Learning Straight Trajectories in Just One Step (NeurIPS 2024 · similarity 0.85)
- Path-Guided Flow Matching for Dataset Distillation (NeurIPS 2026 · similarity 0.84)
- FlowPolicy: Enabling Fast and Robust 3D Flow-Based Policy via Consistency Flow Matching for Robot Manipulation (AAAI 2025 · similarity 0.84)
- Aligning Flow Map Policies with Optimal $Q$-Guidance (NeurIPS 2026 · 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.
