Dropping Off Passengers Could Be a Difficult Task for Self-Driving Cars

News Self Driving Cars Dropping Off Featured

We’ve been spending so much time thinking about the safety of autonomous cars as they drive, that we haven’t stopped to think about what happens when they drop off passengers.

How will they negotiate the best place to drop someone off? We all know the best place to be dropped isn’t necessarily right immediately in front. Sometimes it’s close to the side door over here. Or it may be that there are other vehicles where the self-driving car needs to drop you off. How will it decide what is the best place to drop you off?

Self-Driving Cars’ Next Problem

Isn’t it annoying when you plug a destination into a navigation map, and it puts you across the street, down the block, etc? Will self-driving cars have this same problem? If they can’t program the navigation apps to get it exactly right, how are self-driving cars going to do it?

The dropoff question was raised in a Fobes article by Lance Eliot, an artificial intelligence, machine learning, and autonomous cars expert.

Even an expert on autonomous cars isn’t sure of the answer here, as he referred to it as “a quite vexing problem,” recognizing that it’s “not yet seemingly very high on the priority list of AI developers for autonomous cars.”

Eliot discusses whether self-driving cars will at some point break the law when they drop someone off, posing that if it ends up double-parking even though it’s illegal, then that’s not true that they won’t ever break the law.

News Self Driving Cars Dropping Off Parking

He writes that sometimes humans need to break the law, so the question is if autonomous cars will as well, specifically when it comes to finding a place to drop you off.

The best drop-off spot, according to Eliot, should:

  • be close to the desired location
  • allow you to get out of the car easily
  • ensure your safety
  • ensure the safety of the car
  • not interfere with traffic

This means the self-driving car will need to make a decision. Helping it do so will be its cameras, radar, LIDAR, ultrasonic, and other sensors. These will collect data in real-time about the situation, leaving the AI to ultimately figure out the best spot.

Of course, that above algorithm also needs to take into effect the time of day and the weather. A big rain puddle could change the dropoff slightly and so could a huge snow mound. Road repair could affect this as well

ODDs

Perhaps this is where an ODD will take over. These are Operational Design Domains. Each car can have its own set of ODDs. These determine the standards the car will obey. Once they reach their limit and the situation is outside its ODD, the car will pull over or find a “minimal risk condition” setting.

When parking, the car could have an ODD. If it couldn’t find an acceptable spot, it would need to pull over. But if it’s not finding a spot to pull over with you, this means it would struggle to find a place to pull over as well to obey the ODD, and that wouldn’t be good either.

How do you suggest self-driving cars handle the drop-off situation with passengers (other than just not having autonomous cars on the roadway in the first place)? Tell us your thoughts in a comment below.

One comment

  1. “Even an expert on autonomous cars isn’t sure of the answer here, as he referred to it as “a quite vexing problem,” recognizing that it’s “not yet seemingly very high on the priority list of AI developers for autonomous cars.””
    It seems that the highest priority is to just get the autonomous cars out on the road and to the market. Minor annoyances as the “drop-off” problem or hack-proofing them or making sure they can survive stupid-driver tricks, and/or dozens of other annoyances that have crept up so far, will be taken care of at some future date.

    “Once they reach their limit and the situation is outside its ODD, the car will pull over or find a “minimal risk condition” setting.”
    Isn’t this the same as the “drop-off” problem? The car will still have to find a safe place to park or to pull off.

    “this means it would struggle to find a place to pull over as well to obey the ODD, and that wouldn’t be good either”
    I can just envision the car’s AI going into an infinite logic loop and either stopping right on the spot or continuing on its merry way unresponsive until stopped by some external force.

    The autonomous car developers are presented with an interesting conundrum. The cars will supposedly be much safer because the AI controlling them will be devoid of any human emotions. All decisions will be logical. However, to be able to react to ANY situation, the AI must be as complex, as developed, as experienced as a human brain. It is impossible for the developers to foresee and program for all situations the AI can possibly face. So, just like a human, the AI will have to learn through experience. Are we ready for thousands, if not millions, of cars running around in the wild, learning? It would be analogous to having millions of 14-15 year olds driving around,. Of course, the process can be sped up by having ALL autonomous cars linked together so that what one “learns” all the rest will “learn”. Unfortunately, there are two problems with that idea: 1) There will still be a learning period of some duration, and 2) there needs to be only one AI maker or, otherwise, ALL autonomous car makers will have to agree to use the same standards and protocols. Somehow I do not see that happening. Each car maker will have their own proprietary technology that it will not want to give up or share with others. How many “Auto Pilot”-like products are in use today? How many OnStar-like products are there in use today? I dare say each major car manufacturer (GM, Ford, Nissan, Tesla, etc) has one.

    A lot of MAJOR changes in attitudes, policies, technology, law will have to take place before autonomous cars become viable an as safe as the the proponents claim they can be.

Leave a Comment

Yeah! You've decided to leave a comment. That's fantastic! Check out our comment policy here. Let's have a personal and meaningful conversation.