NotSundial
High-precision pose estimation for indoor robotics
NotSundial is an open-hardware 6-DOF positioning system providing high-precision pose estimation for indoor robotics. Right now, the robotics community relies on inexpensive 2D planar markers (like ArUco) that suffer from notable angular jitter at range. The only alternatives are high-end tracking systems that remain cost-prohibitive and closed-source. NotSundial bypasses this trade-off by shifting the measurement complexity from software inference directly into physical geometry. Building on the core mechanics of the open-source "Mojoptix Digital Sundial", it relies on a 3D-printed Volumetric Angular Mask (VAM) that filters light rays spatially and angularly. Instead of inferring angles from the minimal perspective distortion of a flat pattern, a backlit VAM displays angularly dependent patterns that fully encode the 6-DOF pose into a stable photometric signal. Through this project, we will deliver a completely sovereign tracking stack: parametric scripts to 3D-print custom VAM targets, a native ROS 2 vision engine, and a public benchmarking report.
- The project's own website: https://gitlab.com/lineact-valo/notsundial
Run by CESI Engineering School
This project was funded through the NGI0 Commons Fund, a fund established by NLnet with financial support from the European Commission's Next Generation Internet programme, under the aegis of DG Communications Networks, Content and Technology under grant agreement No 101135429. Additional funding is made available by the Swiss State Secretariat for Education, Research and Innovation (SERI).
