Road to Robotic Parking Is Littered with Faulty Projects

The design problems are mostly basic mechanical design problems. Some systems park a car on a flat pallet, then move the pallet. This has the risk of the car rolling if the brakes don't hold. The more robust systems have a curb on both sides of the pallet and depressions for the wheels, or curbs that rise at each end. (This should be mandatory in earthquake zones.) This also gives the driver more guidance on getting the car positioned properly. The modern systems all check with light beams for proper positioning, but it's better if there are also physical obstacles.

There are palletless systems, where rows of rails rise up under the wheels and lift the car for transport, but those seem to be less popular.

The better designs have more than one retrieval mechanism that can access each slot, so if one fails, the system can still operate, but more slowly.

The lifts have to have the usual safety devices of a freight elevator. There is no excuse for an elevator falling, which has happened.

The usual problems of exposed machinery apply. Everything has to be built for a wet, hostile environment. Optical sensors must survive dirt. Limit switches must survive being banged on. Motors must be weatherproof. Some of these parking systems are repurposed automated warehouse designs, and that's a marginal approach. You want someone on the design team who's designed for a harsh environment, like railroading. "Move fast and break things" will not work.