Production-grade engineering playbooks for high-definition mapping and autonomous-vehicle
spatial data pipelines — built with Python GIS tooling and validated for fleet-scale deployment.
This site exists to help AV engineers, mapping specialists, and Python GIS developers build
reproducible spatial-data pipelines: lane geometry extraction, road-network validation, sensor
fusion, simulation data generation, and quality-control automation. Every guide is grounded in
deterministic processing, strict coordinate governance, and automotive-grade safety practices.
The material focuses on the hard parts of real systems — CRS drift, alignment failures, memory
ceilings on edge compute, batch scaling, and format synchronization across OpenDRIVE and runtime
stacks. Each workflow pairs the architecture you need with the concrete Python patterns to
implement it, from streaming OpenDRIVE parsers to SLERP-based pose interpolation.
Coverage now runs the full length of the map's life: building it, certifying it against withheld
survey control, serving it to a fleet inside a latency budget, localizing against it, and
detecting the day it stops being true. That includes the parts with no ground truth to copy —
junction turn paths are authored rather than extracted — and the parts that only exist once
vehicles are driving, such as change detection from fleet telemetry and the narrowing-only rule
that decides what may be repaired automatically.
85 pages are organized into three connected domains of the
HD-mapping stack, each with its own topics and step-by-step guides. Follow a section from its
overview down to implementation detail, or jump straight to the topic you need below.
