ADAS & Autonomous Driving

Proving safety when you cannot drive 275 million miles: the ISO 26262 ASIL table, the four SOTIF areas, scenario testing, and Euro NCAP 2026's four pillars.

The four decision layers of a self-driving stack: Lanelet2 routing, behavior FSMs, Frenet motion planning, and MPC — plus why rule-based PDM won nuPlan 2023.

Where exactly is the car? GNSS/IMU fusion, NDT LiDAR localization at 1.6 cm lateral error, and map-free lane perception with MapTR.

From fused sensor data to scene understanding: CenterPoint vs DETR3D, Occ3D occupancy prediction, and ByteTrack/OC-SORT tracking on nuScenes.

Series finale: Waymo's 220M miles, an IIHS report showing 68% fewer crashes, Apollo Go's 22M trips — and why methodology beats the headline number.

A practical guide to Levels 2, 2+, 3, and 4: responsibility, ODD, driver monitoring, redundancy, and minimal-risk fallback.

Choose early, late, or deep/BEV fusion for ADAS through latency, sensor-failure robustness, and compute trade-offs.

The 2026 shift: world models are judged not by FVD but by whether generated scenes measure policy quality. GAIA-4, Orbis 2, WorldLens.

NVIDIA Alpamayo-R1 proves a self-driving car can explain why it braked — with structured causal chains you can audit, not decorative text.

Three types of benchmark scores that cannot substitute each other. Read this before trusting any autonomous driving number in 2026.

Hands-on tutorial covering the architecture, installation, and inference of OpenDriveVLA — the AAAI 2026 driving VLA from TU Munich — on the nuScenes dataset.

A four-axis vocabulary — input representation, planning output, supervision signal, evaluation protocol — for reading any end-to-end driving system.