Doctoral Meeting: 'The Missing Points Mystery: Solving LiDAR's 10-Year Information Loss Problem'

Ever wondered how autonomous vehicles "see" the world? LiDAR sensors shoot millions of laser pulses every second to create 3D maps of their surroundings. However, there's a dirty secret: when this data gets converted to images for processing, points literally disappear. For years, researchers have accepted this information loss as inevitable.

In this talk, I'll introduce you to the fascinating world of LiDAR technology (and yes, I'll bring a real sensor to demonstrate), then show how our research challenges this fundamental assumption. What if we could achieve perfect conversion to images without knowing anything about the sensor?

Using techniques from classical computer vision and geometric modeling, we've developed a method that automatically infers how any LiDAR sensor works, just by looking at its data. The result: zero information loss across millions of real-world scans, with measurable improvements in downstream applications. Turns out, even minimal information loss matters more than you would expect.

• Supervisors: Francisco F. Rivera, Óscar García Lorenzo and David L. Vilariño
• Moderator: Catherine Andreu Cafati