GSoC 2026 - Project #7 SLAM and Localization Package (Simulation-Based)

[akky20]

Hello mentor,
I have been going through this project and found it overlapping with my previous work on ROS2-based AMRs with SLAM, sensor fusion, and real-world deployment, I'm new to DORA so I wanted to understand a few things

• Are we expected to strongly focus on LiDAR + IMU sensor fusion for localization?
• Since DORA uses a dataflow model compared to ROS style pub-sub systems, how we are expecting to handle high frequency IMU data with slower LiDAR scans data. ( I have already seen this issue in my project so want to clarify and want to know your views).
• For the 'loop closure' requirement, would a simple pose graph approach be acceptable? I want to scope this correctly before the proposal

Also I read the discussion #1513 and wanted to ask a doubt
Since the map is static and shared in a one-to-many manner, is the recommended approach for each operator (like localization) to cache the map after the initial subscription, or should it rely on continuous access through the topic ?
Just trying to understand the expected design pattern in DORA for handling such large but static data efficiently.

[RuPingCen]

We also have an important suggestion for your project scope: we strongly recommend porting mature open-source mapping and localization nodes into the Dora framework, rather than developing them entirely from scratch. For example, you can integrate high-performance open-source solutions like lightning-lm

https://github.com/gaoxiang12/lightning-lm