Slam2Lanelet2 — Automated Lanelet2 Map Generation (Beyond bag2lanelet & Manual VMB Workflows)

[Abinesh0610]

Hi Autoware Team,
My name is Abinesh K., and I am from the BITS-Pilani WILP Division, Hyderabad Campus, India.
I built a ROS 2 package called Slam2Lanelet2 that directly addresses a major bottleneck in the traditional AV mapping workflow.

The Old Way (The Pain):
In the conventional pipeline, after recording a SLAM run, the team has to:

1. Manually pick the initial pose in the PCD and match it to a real-world reference
2. Manually pick the end pose and match it again
3. Open Vector Map Builder and hand-draw every single lane, boundary, and lanelet relation — one by one

For even a modest route, this process easily eats up several hours of painstaking manual work, and any misalignment at step 1 or 2 cascades into errors across the entire map.

Autoware's own official tool — bag2lanelet — only partially solves this. It requires the user to manually type in the MGRS grid coordinate, applies no live GNSS drift correction, and by its own documentation, produces jagged lane outputs that still need to be re-imported and manually corrected in Vector Map Builder.

The New Way (What My Package Does):
Slaml2Lanelet2 eliminates all of that. It runs live alongside your SLAM stack, fuses the local trajectory with real-time GNSS data, computes the exact heading drift, and corrects it mathematically. No manual pose picking. No Vector Map Builder lane drawing. but for your location you have give the MGRS coordinates in the code itself. iam from India so i hard coded as 44QKE that's it.

The output is auto-generated and deployment-ready:
→ lanelet2_map.osm — Lanelets, boundaries, and relations drawn automatically along the driven route
→ PCD transform parameters — exact tx, ty, rz values so your PointCloud overlays the map perfectly
→ Projection config (MGRS/UTM) — plug directly into Autoware

What used to take a full day of manual work in Vector Map Builder now happens automatically the moment you hit Ctrl+C after your mapping run.

It's configurable for different road widths, environm
ents, and sensor setups — built for both research teams and field deployment at scale.

If you are intrested i will share you the documentation of my work you will understand completely. I don't know how to present my work so i posted in the discussion panel

Output Vide and Image Google Drive Link :

Mail Contact : k.abhinesh@pilani.bits-pilani.ac.in

[mitsudome-r]

@Abinesh0610 Thanks for sharing your great work!
Is it possible to share your code and document by pushing it to your GitHub?
If that tool is useful, we could consider adding it our documentation and Autoware Index so people can find your package.

[Abinesh0610]

@mitsudome-r
Thank you for your interest in the package and for considering it for the Autoware documentation and Autoware Index.

I have pushed the source code and complete documentation to GitHub. The repository includes the README, installation instructions, workflow details, implementation approach, and example outputs.

Here is the repository link:
[GitHub Repository Link]

I would greatly appreciate it if you could review the package and share your feedback. I am actively working on improving it, so any suggestions regarding functionality, performance, code structure, documentation, testing, or integration with the Autoware ecosystem would be very valuable.

My goal is to make this package useful and easy to adopt for the wider autonomous driving community. I look forward to your recommendations and any guidance on the steps required for inclusion in the Autoware documentation and Index.

Thank you for your time and support.
Best regards,
Abinesh K
Mail: k.abhinesh@pilani.bits-pilani.ac.in
BITS Pilani, Hyderabad Campus