Sean
Ernstes

Fig. 1 — The sensor

Good localization, off the shelf.

A VR headset already knows exactly where it is: inside-out tracking fuses camera frames and IMU data into a full 6-DoF pose, with no markers on the walls. It's mass-produced, in stock and calibrated at the factory.

The usual way

QuestNav

Money

specialist sensors

one consumer headset

Parts

niche vendors, lead times

any electronics store

Skill

calibration and SLAM tuning

mount it, add a library

Fig. 2 — Teardown

QuestNav points that stack at a robot.

I lead the open-source project. It was built for FIRST Robotics teams, students on tight budgets, and has since spread to other robots that need localization without the price tag. Four parts, four languages:

Headset

Unity / C#

reads the pose

Wire

Protobuf

ships it

Robot

Java

fuses it

Pit

Vue

configures it

Fig. 3 — On the field

Wheels slip. Cameras don't.

Wheel odometry drifts with every bump and skid. The headset keeps locking onto features it has already seen, so the error stays near zero, with no lidar, no motion-capture room and no SLAM expert on the team.

• QuestNav (VSLAM)
• Wheel odometry
• Mapped features

Stars

140+

Forks

80+

Commits

400+

Releases

20+

A vendor dependency for FRC teams worldwide, and now low-cost localization for robots beyond FIRST.