As a CSP researcher I find this motivating, because we are working on a system that does not require calibration at all.

Nevertheless, such self-aligning approaches are not explained at all which is quite annoying.

Our particular approach is based on machine learning for object detection using an embedded computer.

We have published several papers about this idea and also an article explaining our approach: https://mechatronicsblog.com/smart-solar-tower-power-mockup-...

We did not get much attention but we hope that we can keep working on this and get funding to develop this approach for real systems.