AI for Home Garage Automatic Parking with Self-Driving Cars


By Dr. Lance B. Eliot, the AI Trends Insider

I park two cars into my home’s attached two-car sized garage. When I first moved into this house, the garage was empty and so it was pretty easy to park the cars. One of the cars is kept to the far left of the garage and the other car is aimed at the far right of the garage. This allows for a couple of feet between the two cars. It is just enough of a gap between them to allow the driver and the backseat passenger on the driver’s side of the right parked car to get out without bumping the car to the left of it. Likewise, the front seat passenger that sits next to the driver of the car parked to the far left is able to fairly readily exit from the far-left parked car. Well, this all assumes that any occupants in either car are mindful of not opening their car doors at precisely the same moment as someone in the other car parked next to it (if they do so, it is like the famous quote from Ghostbusters that crossing the two streams will cause utter devastation).

After living in this house for a while, I sheepishly admit that the garage has increasingly become more crowded. There are boxes of school items, cartons of old electronic equipment that needs to be e-wasted, bikes that I am too lazy to put up on the garage racks, and so on. What this has done has forced me to gradually try to park the two cars closer and closer together. That earlier gap between the two cars has narrowed significantly. Doing so has allowed for more space opening up on their respective far sides, which is where the family-related junk now sits. Imagine that when you try to get out of one of the cars, you need to very carefully open your car door, doing so slowly and gingerly, otherwise you will bump either into boxes or dent into the side of the other car.

Parking the cars into the garage has become the equivalent of docking together the International Space Station and a Soyuz rocket capsule. If one car is already parked in the garage, the effort to get the other car into the garage requires a keen eye and a steady hand. Inch at a time, you need to just have the second car crawl into the leftover space. Depending on your starting position on the garage pad, you can end-up in a really bad spot. Thus, knowing where to first sit the car on the garage pad is crucial to getting into the space. You need to get the car seated just right, then proceed forward at a snail’s pace, meanwhile looking earnestly at each side of the car to make sure that you don’t bump into something. Often, the side view mirrors take the brunt of this tight squeeze and get pushed into the in-tight position by sliding or bumping on the other car or some other obtruding object.

I sometimes feel that it is a great accomplishment to simply get both cars parked into the garage without striking anything. This is reminiscent of the old Operation board game that required you to pick-up a piece of a skeleton and not hit the sides of the board when doing so (else a light would light-up and a kind of electrical shock noise would be heard). There are some nights that I arrive at home after an especially tiring day at work, and I don’t even try to park the car into the garage, instead just leaving it out on the garage pad overnight. Why endure the agony of trying to park it in the garage, I figure, when I’ll just need to back it out the next morning. Of course, overnight the car will sit outdoors and need to deal with the vagaries of the world that might come at it, but the exertion needed to get it into the garage seems worth those chances. Plus, I figure too that since I have car insurance that if the car is somehow gets vandalized it would be easier to deal with the insurance company than it would to park the car into my garage (that’s how bad the garage parking situation has become!).

What does this have to do with self-driving cars?

Answer, we are all going to want to have a self-driving car that is smart enough to park in our home garages. At the Cybernetic Self-Driving Car Institute, we are developing AI that will be able to park your self-driving car into your home garage, no matter how tough your garage parking situation might be.

The aim is to ensure that a self-driving car can park a car into a home garage, assuming that it is possible and that a human driver could do so. I emphasize this aspect because if a car just won’t fit, it won’t fit. The AI cannot magically make a car become thinner. There has to be some reasonable means of getting the car into your garage. The self-driving car is not going to magically transform into some kind of snake-like creature and weave its way into your garage. The physics of your car and the physics of your garage are still going to be factors.

Self-driving cars are considered to be self-driving capable as rated on a scale of 0 to 5, with the Level 5 being a self-driving car that can drive in any situation that a human driver can drive. I mention this due to the aspect that if an auto maker claims to have a Level 5 self-driving car, which is considered a true self-driving car, it must be able to park your car into your garage, assuming that a human could do so. In other words, by definition, if a human can park your car into your garage, the self-driving car if a Level 5 must be able to do the same thing.  There are some auto makers that are claiming they are devising a Level 5 car, but those self-driving cars aren’t able to park a car into a home garage to the same skill level as a human driver, thus, I would assert that those self-driving cars are not a Level 5.

Some might quibble with me and argue that if a self-driving car could drive on the freeways without human intervention, and drive on city streets without human intervention, and even drive up to your garage door without human intervention, doesn’t that seem sufficient to say that the car is a Level 5 car? I say no. I say that if it cannot do whatever a human driver can do, including being able to park the car into your garage, assuming that a human could do so, then that self-driving car can be considered a Level 4 perhaps but not a Level 5. I am a stickler about this. The definition is clear cut about a Level 5. It must be able to drive in whatever circumstance a human could drive a car. You might consider parking into your crowded home garage to be the so-called “last mile” of what a self-driving car needs to do.

Suppose you have a home garage that has ample space and no obstructions that prevent a car from being parked in there. That’s an easy situation and indeed many of the upcoming self-driving cars can achieve that parking situation. If you’ve got ample clearance on either side of a car, and if the car can fit without any particular tightness, this is a pretty easy driving task. I mention this because some auto makers are assuming that a home garage will be spacious and a snap to park into. In an idealized world, that would be true. I am sure there are many homeowners perfectly keep their garage well cleared out and make it as easy to park a car as if a two-year-old child could do it. On the other hand, I am not alone in having a home garage that requires some really surgeon-like skills to park a car into.

I would wager that maybe one-third of the home garages in my neighborhood present the same challenges of parking. I see the contents of their garages as I drive past those homes during my driving throughout the neighborhood, and can’t help but notice how many are using their garages in a pack rat kind of way. Old sofas are in those garages, surfboards are in those garages, and you name it. I suppose an auto maker would say that those people are wrong to use their garages in this manner. They should toss out that stuff or put it someplace else. In the minds of some self-driving car makers, they believe that the “user” is wrong for having a garage some jam packed that it is nearly impossible for self-driving car to park into it.

I say hogwash to those auto makers. You cannot change the world to suit the capabilities of your self-driving car. You need to make the self-driving car match to the needs of the real world.

Alright, how can you get a self-driving car to properly park the car into a jam packed garage. The answers is by having an AI component that undertakes that particular need. Using the generalized AI of the self-driving car will not be sufficient for the true home garage parking problem. The generalized AI assumes that there is a lot of space to park the car. It assumes that the task of parking the car is straightforward. The AI merely drives the car into the garage as though it is parking into a mall parking spot. Nothing special to consider.

Our AI specialized component has learned about the variants involved in a tightly woven home-garage parking situation.

There are several notable aspects:

  • When first trying to park into a more rigorous home garage parking situation, the self-driving car uses it various sensors to try and detect the nature of the garage space itself. The dimensions of the space are crucial. Other objects in the space need to be detected. As much of a map of the garage needs to be scanned as can be ascertained.
  • In some instances, the self-driving car needs to do this by entering into the garage. If the self-driving car is only sitting outside on the garage pad, its sensors might not be able to sense the intricacies of the inner aspects of the garage. As a result, the self-driving car will start into the garage, putting a toehold into the garage to be able to get a better scan of what’s there and how the self-driving car can park into it. It is not trying to park at this stage, and merely doing a closer scan.
  • The self-driving car also often needs to interact with the human of the car, finding out whether for example the garage is being used to park more than one car. If it’s a two-car garage, and suppose no cars are yet parked into it, and if the self-driving car parks smack dab in the middle of the garage, it’s not much of an accomplishment, since it means that the other car (whether human driven or self-driving) will not be able to park at all. The default for the self-driving car is to park to one side or the other, allowing for maximum available space for another car, while also ensuring that the self-driving car can get into and out of the garage.
  • The basis for interacting with a human also includes the circumstance when suppose the garage normally has to park two cars, but suppose one of those cars is not going to be parked there for now (maybe that car is in the shop being repaired). If the self-driving car does its usual thing of squeezing to one side, it is perhaps stupid to do so, when it could more readily use the rest of the space. The tight squeeze method might not always be needed. The self-driving car does not have any heavenly means to divine this, and would need to interact with the human to find out the circumstance.
  • The self-driving car needs to know how much space around itself it needs to leave to be able to allow humans to enter into and get out of the self-driving car. As such, it needs a certain amount of self-awareness about the doors of the car and how far they swing when opened. Again, this might require interacting with a human, since it is possible that Grandma Jones can only get into the car if the left side has lots of clearance, while those entering into the right side can squeeze in, as needed.

The self-driving car further needs to be able to do this:

  • Obstacles that are in the garage need to be detected. There can be clearance aspects involved from all directions. There might be items hanging from the ceiling of the garage. There are objects in front of where the car will be parked. Objects to the right and left. The car also needs to fit into the garage such that the garage door can be closed. It does no good to have squeezed into a spot and then the garage door won’t close (though, this is allowed as a desired exception by the human, if so desired).
  • Once the self-driving car learns about the idiosyncratic aspects of a particular home garage, it is then able to more routinely park there. It is akin to a human driver that learns over time how to park in their garage. In my own case, I pull into my garage at a pretty fast clip. I noticed the other day though, upon allowing a friend to park in my garage, I could see how slowly he parked in comparison to my speed. Notably, the first time involves wanting to make sure that you are getting a feel for where the car goes. After multiple such parking achievements, it becomes easier. A good AI component has this learning capability.
  • The self-driving car though cannot make an assumption that the garage parking situation is identical each time the self-driving car parks the car. For example, the other day, I opened my garage door and realized that one of my children had parked their bike into the spot where my car would normally go. If I had proceeded on the basis of my earlier learned static map of the garage, I would have run into the parked bicycle. As such, the self-driving car needs to re-assess each time, identifying whether the parking situation has changed from what it earlier knew.

For some homeowners, they put various devices into their garages to make parking easier to do. For example, a low-tech method involves hanging a tennis ball from a string, placing this into the middle of the garage to help a human driver realize they must park to the left or right of the hanging marker. There are more high-tech devices such as a signal light with a sensor. The signal light is hung onto a wall of the garage and shines the usual colors of green, yellow, red, and will then try to help guide you as a human to park your non-self-driving car.

Our viewpoint is that you should not need to trick-out your garage for purposes of having a self-driving car park into your garage. I realize that some auto makers are going to try and go this route, namely get owners of self-driving cars to buy special add-on devices that need to be mounted or placed into a garage, thereby allowing a “normal” self-driving car to park into the garage. We don’t think this is needed and also view it as a brazen attempt at solving a problem by simply upselling the consumer. The AI of the self-driving car should have the smarts needed to park that self-driving car into the garage, and no added gimmicks or expensive add-ons should be needed. Raise the bar on self-driving cars, we say, and not dumb them down.

This content is originally posted in AI Trends.