ADSafetyPilot

安全三支柱Safety Trinity

🛡️

FuSa

ISO 26262

功能安全: HARA、ASIL分解、安全目标、安全机制设计

Functional Safety: HARA, ASIL decomposition, safety goals, safety mechanisms

🎯

SOTIF

ISO 21448

预期功能安全: 触发条件分类、未知不安全场景识别、场景驱动V&V

Safety of the Intended Functionality: triggering conditions, unknown unsafe scenarios, V&V

🔐

CyberSec

ISO 21434

网络安全: TARA威胁分析、安全启动、入侵检测

Cybersecurity Engineering: TARA, secure boot, intrusion detection

核心价值Why ADSafetyPilot?

🔗

三合一分析Integrated Analysis

FuSa + SOTIF + CyberSec 交叉映射的统一分析流程，不再是三个独立的孤岛。FuSa + SOTIF + CyberSec as one integrated workflow, not three separate silos.

🇨🇳

中国标准深度覆盖China Standards

L2组合驾驶辅助安全要求、GB 44495 (L3)、GB 44496 (泊车) 等强制标准完整覆盖。Full coverage of L2 ADAS, GB 44495 (L3), GB 44496 (Parking) mandatory standards.

📊

真实数据驱动Real Data Driven

基于驭研科技 DRIVEResearch 800h+ 航测数据，提供中国驾驶员行为参数基线。Grounded in DRIVEResearch 800h+ aerial data for Chinese driver behavior baselines.

🧾

场景证据包Scenario Evidence Pack

把 cut-in 真实数据转成参数分布、收敛检查、证据缺口和测试用例候选。Turn real cut-in data into parameter distributions, convergence checks, evidence gaps, and test-case candidates.

查看 cut-in v0.1 →View cut-in v0.1 →

系统架构Architecture

ADSafetyPilot
|
+-- skills/
|   +-- sotif-deep/           ISO 21448 全生命周期
|   +-- china-regulatory/      中国强制标准 (L2/L3/Parking)
|   +-- safety-trinity/        FuSa + SOTIF + CyberSec 三合一
|   +-- dfm-query/             驾驶员基础模型查询
|
+-- scenarios/
|   +-- jama-v4/               JAMA V4.0 (58 场景类型)
|   +-- china-l2-adas/         中国L2强标 (45 测试场景)
|   +-- china-l3-ads/          GB 44495
|   +-- china-parking/         GB 44496
|
+-- knowledge-base/
|   +-- standards/             标准法规索引
|   +-- projects/              PEGASUS, ENABLE-S3, SAKURA
|   +-- best-practices/        最佳实践
|
+-- data/
|   +-- dfm-baselines/         DFM 行为参数基线 (待注入)
|   +-- scenario-params/       场景参数 P5-P95
|
+-- LICENSE                    MIT

场景库Scenario Library

基于 JAMA V4.0 安全评价框架，对标中国 L2 组合驾驶辅助系统安全要求。45个场景中40个为中国独有。 Based on JAMA V4.0 framework, mapped to Chinese L2 ADAS mandatory standards. 40 of 45 scenarios are China-unique.

ID	场景Scenario	来源Source	安全域Safety
SC-01	高速跟车（稳态）Highway car-following	JAMA+CN	FuSa+SOTIF
SC-02	高速前车切入Highway cut-in	JAMA+CN	SOTIF
SC-03	前车急刹Lead vehicle braking	JAMA+CN	FuSa+SOTIF
SC-04	匝道合流Ramp merge	JAMA	SOTIF
SC-05	环岛通行Roundabout	China	Combined
SC-06	信号灯识别Traffic light	China	Combined
SC-07	交叉口冲突Intersection conflict	China	Combined
SC-08	施工区域Construction zone	China	SOTIF
SC-09	雨雾天ODC边界Rain/Fog boundary	China	SOTIF
SC-10	最小风险策略 (RMF)Risk Mitigation	China	FuSa

显示 10/45+ 个场景。Showing 10 of 45+ scenarios. 在 GitHub 查看完整目录 →View full catalog on GitHub →

路线图Roadmap

Phase 1 — 安全知识引擎 (SOTIF + 中国标准 + 三合一分析) Safety Knowledge Engine (SOTIF + China standards + Safety Trinity) In Progress
Phase 2 — 场景库 + DFM数据引擎 (JAMA V4.0 × 中国标准 + DRIVEResearch数据注入) Scenario Library + DFM Data Engine (JAMA V4.0 x China + DRIVEResearch data) In Progress
Phase 3 — OpenSCENARIO场景生成 + 仿真平台对接 OpenSCENARIO generation + simulation platform integration Planned

关于作者About

张玉新 / Yuxin Zhang

autozyx.com · LinkedIn · Google Scholar · 公众号: 张玉新-AutoZYX WeChat: 张玉新-AutoZYX