Research map / Probabilistic and scientific modeling / Dynamical systems
Risk-Driven Design of Perception Systems
Anthony Corso, Sydney Michelle Katz, Craig A Innes, Xin Du, Subramanian Ramamoorthy, Mykel Kochenderfer
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 a risk-sensitive loss to perception training | Inverted-pendulum perception system · Inverted pendulum | mean time to failure: 234 ± 62 -> 492 ± 6 | Improved |
| Train on states sampled by risk-driven weights | Inverted-pendulum perception system · Inverted pendulum | mean time to failure: 152 ± 89 -> 500 | Improved |
| Add risk-driven data generation | YOLOv5s DAA perception system · Aircraft DAA encounters | NMACs: 33% | Improved |
| Combine risk-sensitive loss and risk-driven data generation | YOLOv5s DAA perception system · Aircraft DAA encounters | NMACs: 37% | Improved |
| Add risk-sensitive loss | YOLOv5s DAA perception system · Aircraft DAA encounters | NMACs | Improved |
| Set α=0.99 rather than α=0.00 for risk loss (ablation) | Inverted-pendulum perception system · Inverted pendulum | mean time to failure: 492 ± 6 -> 391 ± 59 | Hurt |
| Set α=0.99 rather than α=0.20 for risk-driven data (ablation) | Inverted-pendulum perception system · Inverted pendulum | mean time to failure: 500 ± 0 -> 273 ± 65 | Hurt |
| Apply risk-sensitive loss | YOLOv5s DAA perception system · DAA validation set | recall: 0.38 ± 0.03 -> 0.37 ± 0.02 | No clear effect |
| Apply risk-driven data generation | YOLOv5s DAA perception system · DAA validation set | precision: 0.78 ± 0.04 -> 0.77 ± 0.01 | No clear effect |
| Use risk-driven data generation and risk-sensitive loss | YOLOv5s DAA perception system · DAA validation set | recall: 0.38 ± 0.03 -> 0.39 ± 0.01 | No clear effect |
Closest papers by meaning
- Monitoring of Perception Systems: Deterministic, Probabilistic, and Learning-Based Fault Detection and Identification (Abstract Reprint) (AAAI 2024 · similarity 0.87)
- AGENTSAFE: Benchmarking the Safety of Embodied Agents on Hazardous Instructions (CVPR 2026 · similarity 0.84)
- SafeVLA: Towards Safety Alignment of Vision-Language-Action Model via Constrained Learning (NeurIPS 2025 · similarity 0.82)
- Towards Computational Foreseeability (AAAI 2025 · similarity 0.82)
- Defining Operational Conditions for Safety-Critical AI-Based Systems from Data (NeurIPS 2026 · similarity 0.82)
- A Physics-Informed Machine Learning Framework for Safe and Optimal Control of Autonomous Systems (ICML 2025 · similarity 0.81)
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.
