Auto-park is one of those "tech features" in cars that I really, really, want to see work, but I'm convinced no one has nailed it yet.
In theory, it's the perfect candidate for an autonomous driving feature:
1) Watching people for many years try to parallel park, I'm convinced it's not a skill humans were meant to master. I find it easy, but the vast majority of drivers simply suck at it. I no longer think it's a matter of bad training. I really think most people don't have the ability to do the geometry in their heads. Nor should they bother to try to learn. It's like spelling. Nice to be good at it, but it's not necessarily a marker of great intelligence. More like a parlor trick.
2) There are a finite set of variables. Unlike full-self driving, parking is a singular set of problems to solve. Detecting a useable spot, getting the car into starting position, wheel movement as the car is moved into place. Yes, you have to be aware of cars trying to pass, cyclists, pedestrians, other obstacles. Plus incline/decline and how that might effect acceleration, etc. But it's all mostly contained and predictable compared to driving out on the road for several miles. And the movements are, again, simple geometry. Humans may suck at that, but computers were made for this sort of problem.
3) Given number 1, people would gladly pay to never have to park ever again.
Knowing this is the ultimate convenience feature, it baffles me that every car I've ever tried to use auto-park on has done a terrible job. Or, I should say, it does a fine job under perfect circumstances, but I end up in those circumstances about once a year.
The rest of the time: Can't detect spots. Often gets the angle wrong. Curbs the rims. Almost hits other cars/pedestrians. And on and on.
I am curious to try the Lucid auto-park. But I will do it in a very safe place.
I get that companies have a tendency to do a half-crap job on a feature and then simply "check it off" as done. But man, I wish someone would forget full self driving and just focus on parking for five years.
But who needs to invest in that, when everyone is already selling the feature as if it works?