ad-safety-research-skills

Daily news reporting for autonomous driving industry. Tracks OEMs, AD companies, regulations, conferences, expos, and technical breakthroughs across global and Chinese media. Designed for high-signal daily briefings.

Behavior modeling for autonomous driving: trajectory prediction (Trajectron++, HiVT, QCNet, MTR), interaction modeling, game-theoretic planning, driver behavior analysis, social force models, and graph-based multi-agent modeling. Use when researching motion forecasting, behavior prediction, interaction-aware planning, or driver behavior analysis for AD safety.

Analyze crash data, naturalistic driving data, and compute safety metrics for autonomous driving research. Covers FARS, CRSS, GIDAS, SHRP2, and fleet telemetry analysis. Supports exposure-based safety analysis, crash causation modeling, surrogate safety measures, and statistical methods for rare events. Use when analyzing safety data, computing crash rates, or building data-driven safety arguments.

Design and analyze experiments for autonomous driving safety research: simulation studies, closed-course tests, naturalistic driving analysis, and benchmark evaluations. Covers CARLA, SUMO, CommonRoad, nuScenes, Waymo Open, and standard AD evaluation protocols. Use when planning experiments, choosing benchmarks, or designing test matrices for AD safety validation.

Foundation models for autonomous driving: Vision-Language-Action models (RT-2, OpenVLA, DriveVLM), end-to-end driving (UniAD, VAD, SparseDrive), LLM/VLM-based planning and reasoning, and safety implications of deploying foundation models in safety-critical AD systems. Use when researching VLAs, E2E driving, or LLM/VLM applications in autonomous driving.

Generative models for autonomous driving: world models (GAIA-1, DriveDreamer, GenAD, UniSim), diffusion models (CTG++, DiffScene, MagicDrive), flow matching, and 3D Gaussian Splatting (StreetGaussians, DrivingGaussian) for scene generation, sensor data synthesis, scenario augmentation, and closed-loop simulation. Use when researching or implementing generative approaches for AD data, simulation, or planning.

Specialized literature review for autonomous driving safety research. Searches across AD-specific venues (IV, ITSC, T-ITS, T-IV, CVPR WAD, NeurIPS ML4AD), standards bodies (ISO, SAE, UNECE), and crash databases. Use when searching for AD safety papers, building related work sections, or surveying the state of the art in autonomous driving.

Paper writing specialized for autonomous driving safety conferences and journals. Knows AD-specific terminology, metrics, experimental conventions, and formatting for venues like IV, ITSC, T-ITS, T-IV, CVPR WAD, NeurIPS ML4AD. Use when writing or revising AD safety papers, rebuttals, or supplementary materials.

Orchestrator for autonomous driving safety research. Routes to specialized sub-skills for literature review, paper writing, scenario analysis, standards navigation, experiment design, and dataset analysis. Use when conducting any AD safety research task spanning automotive, AI, or transportation domains.

Analyze autonomous driving safety scenarios, define Operational Design Domains (ODD), identify edge cases, perform hazard analysis (HARA, STPA, FMEA, FTA), and assess scenario criticality. Use when analyzing driving scenarios, safety cases, triggering conditions, or functional insufficiencies for automated driving systems.

Navigate automotive safety standards and regulations for autonomous driving: ISO 26262 (functional safety), ISO 21448 (SOTIF), UL 4600, ISO/PAS 8800, ISO 34502/34503, SAE J3016, UNECE R157, and emerging regulatory frameworks. Use when researching standards requirements, compliance gaps, or regulatory landscape for automated driving systems.