Are you looking forward to self driving cars?

Yesterday, I was driving to Kaanapali beach walk and tried to look for parking in a very congested area with some public parking spaces jammed in between hotels and shops. All were occupied, and I had to settle for an expensive paid slot in a multi-level parking garage ($3/half-hour).

I thought about a robot car, and wonder how it could even begin to navigate this mess. You will understand if you see the map that the LIDAR produces. It is not that detailed and precise. The LIDAR gives you distances to surrounding obstacles. It does not tell you what those obstacles are. And then, a human eye can see between parked cars, and his mind can map out the layout of the parking lots with other cues such as rows of trees, fences, lamp posts, etc... What one needs is sensor fusion software to combine the LIDAR with the vision cameras.

I have not seen anyone doing the above. It takes a lot of very smart software. In time perhaps, but they do not have it yet.

I have yet to see someone demonstrating a car navigating a parking lot that has not been mapped out in advance. I do not think the current robot cars know how to find the entrances or exits of a parking lot (I spotted the toll booth), let alone a parking garage. The software needs to be a lot smarter than its current state.
 
Last edited:
What one needs is sensor fusion software to combine the LIDAR with the vision cameras.

I have not seen anyone doing the above. It takes a lot of very smart software. In time perhaps, but they do not have it yet.
Thoughts:
1) If they can do the sensor fusion right, it will actually be better than humans can do. Especially since the computers can do signal processing to detect changes (from frame-to-frame, doppler/phase detection, and months apart) that lends new info.
2) When the cars can each collect and contribute to building a shared, common picture of a place based on all their input, it won't take long to build and maintain a "map" of even the most detailed stuff (where is the tool booth? Which parking spaces are frequently empty?). But that's a lot of date to be made into knowledge, and I doubt we are near having that capability, and it wil be expensive.
3) Lots of privacy and law enforcement tie-ins to #2. License plate readers, facial recognition are already common. "Scan the Google world, provide locations of Ohio plate PGY 1234 for last 14 days. And was it ever known to be collocated with Illinois plate TRP 5678?" Will Google collect info like this? What does Google do >now< to make money? It's all publicly available for anyone to see, right? I'm not sure present Constitutional protections will pertain to this. New territory for sure.
4) If you had a self-driving car, you wouldn't have needed a parking space in the first place. Have it drive around the city for awhile. Gas and maintenance sure would have been cheaper than $3/30 minutes:)
 
Last edited:
1 through 3) Sure, but in time. We will not be there in a few years as they promise.

4) And to cause more traffic congestion and energy wastage?


PS. I think people misunderstood when I pointed out the deficiencies of current systems. I never said that an autonomous car is impossible, only that there are a lot of problems that need to be solved before a truly smart robot car becomes a reality. So, I am not holding my breath for one. I'd rather be more excited if someone announces some lower fruit, such as inexpensive renewable energy cost so I can lower my thermostat in the summer.

In the mean time, some easier and ready options like autobraking, blindspot detection, and lane deviation warning would be welcome to become standard features in cars. These are implemented with simpler proximity sensors and vision cameras, and require no high-power onboard computers.
 
Last edited:
There was an news paper article today about Google testing their cars in Austin, Texas, and how Texas did not appear to have any law against cars without human controls.

This got me thinking. How do Google's cars handle large parking lots and malls? If there are no human controls, how do they select a parking space? How do they know which door or store to park closest to? There are most likely a whole lot of other little decisions in the finale mile of a journey that would seem to need human control.
 
... This got me thinking. How do Google's cars handle large parking lots and malls? If there are no human controls, how do they select a parking space? How do they know which door or store to park closest to? There are most likely a whole lot of other little decisions in the finale mile of a journey that would seem to need human control.

I don't see that they do any of that right now, and only try to navigate the streets and avoid collision with other cars, watch out for jay walkers, bicyclists who swerve into their lane, people about to cross at a crosswalk, etc...

I can see that large shopping mall parkings, once they are mapped out, can be handled with a large touch screen showing the mall. You select where you want the car to go by pointing on the screen.

But there are so many other smaller business and residential places that must be mapped. And what to do if the map is outdated? How about you want to pull off the road to stop at a fruit stand?

And then, I see that in high-density urban areas, people run out of space and park their car part on a driveway, part on the lawn. If a car has no steering wheel, they will have to provide a way for the rider to tell the car what to do.
 
I am not looking forward to self driving cars. Prefer them in the rear view mirror.
 
I am not looking forward to self driving cars. Prefer them in the rear view mirror.


You may change your mind though when you can't drive anymore


Sent from my iPhone using Early Retirement Forum
 
They can't come soon enough for me. Driving in LA Metro is a nightmare.
 
40 years too early

As a young programmer, I was doing software for a self driving electric car project back in 1972-1975. The company built a test and demonstration course with garage/terminal and three stops on a 14 acre site. Getting everything to work as expected was pretty tough back then - there are endless "possibilities" that have to be accounted for. The company came reasonably close to selling the concept to the city of Los Vegas to provide transportation between the airport, the strip, and downtown. However, simulations showed it would be hopelessly congested. The project was canned when one of the vehicles took off on it's own cross country, ran into a creek, turned over and caught fire.

Of course, we were trying to do this with a PDP11 with 24Kbytes of memory :)

Bottom line, I think this is a tougher problem than many assume. I've also heard that the biggest issue is how to turn control back over to a human when things go wrong and the human is unprepared to handle the situation.
 
Back
Top Bottom