Case study
How Auve Tech calibrates 15 camera and lidar sensors on MiCa 2.0, an autonomous shuttle bus
Problem
- Sensor calibration, 15+ sensors / vehicle
- Needs to happen in different on-road municipalities
- Simple process that generates reliable results
Solution
- Calibrates any number and combination of sensors
- Works anywhere: targetless, no checkerboards
- MSA engineers validate every calibration
Company
- Name
- Auve Tech
- Domain
- On-road autonomy
- Platform
- MiCa, autonomous shuttle bus
- Fleet
- Deployed to 6+ countries
Featured engineer
Board Member & Head of Development, Auve Tech
Role: Leads MiCa HW and SW deployment
Backstory: Wrestled with multi-lidar alignment, ground plane detection, and object detection with poor calibrations
Main Street Autonomy accelerated our development plans by more than a year and enabled us to build and deploy vehicles to Estonia, Japan, the UAE, the UK, greater Europe, and now the United States.
Previous solution
Auve Tech wrestled with a large number of sensors that made manual calibration time consuming and complicated. Auve Tech engineers calibrated individual sensors, but with so many sensors on the MiCa shuttle bus, it could take more than a full day to fully calibrate. Auve Tech’s previous calibration process was manual and target-based: the process required moving checkerboards or targets through the sensor FOV while engineers manually adjusted sensor extrinsics.
Limitations of Previous Solution
- Required multiple engineers and took more than a full day per calibration
- Required targets, checkerboards, and manual adjustment of sensor extrinsics
- Did not address sensor intrinsics or time offsets
- Generated a result that was operator-dependent
The Calibration Anywhere solution