Monday, April 2, 2018

OTR: Autonomous Vehicles

I spend a couple hours of each day rolling up and down the Tri-State Tollway going to and from work. The road is pretty busy when I'm out there, with lots of trucks, but the flow of traffic, when it is moving well, is usually quite a bit above the posted speed limit. It's not quite a "black diamond" driving experience, but certainly one that will keep your attention.

I've often thought that these commutes would be perfect for autonomous driving. I'd love to be able to handle email, participate in conference calls (shared screens and all), or catch up on my reading. There would be two hours of time turned productive! Instead, I surf SiriusXM for some good music, think about issues at work, mentally compose blog posts, and so forth. I also react to the drivers around me. Trucks have a bad habit of wandering from lane to lane and driving in the lanes that they aren't supposed to drive in. Automobiles show evidence of distracted drivers, with sometimes very severe evidence in the form of a crash or two. I tend to drive very defensively and leave a fair amount of space between myself and the car ahead of me -- probably to the chagrin of other drivers. I try to moderate my speed and use that space to slow down gradually, rather than slamming on the brakes. It keeps the adrenaline level moderate -- and the brakes last longer.

In my musing, I think about all the things a driver needs to be aware of -- speed, distance, cars proximate to your own car, road hazards, things falling off of other vehicles, badly marked lanes, changes in lane availability due to construction or accidents. When I think about programming software to manage all of those variables, my head hurts. But clearly, the software and surrounding technology has been developed, because it exists in a number of forms.

The question is how good that software might be. In recent weeks, we've seen an accident involving a pedestrian who was hit by an autonomous vehicle. We've also seen a fatal accident where a vehicle utilizing some aspects of autonomy allegedly did not perform as expected. In both instances, a human driver was at the wheel, but did not take control in time to avoid the accidents. The root cause of each accident has not yet been determined. It could be software; it could be sensors; it could be integration of software and sensors; or it could be human error. It also could be something that the sensors or the software weren't prepared to identify and react to.

I think about driving on Chicago-area roads, where moon-crater potholes bloom before the tulips. Many of the same roads suffer from worn out paint demarking the lanes, or paint that was clearly not mixed to specification for use on roadways. There's a manhole in a travel lane on I-290 that was not constructed to be level with the roadway and I've learned to avoid rolling over it each day. I've learned that a vehicle having trouble maintaining its lane and speed is probably being driven by a distracted or impaired driver and needs to be given plenty of room. An exploding truck tire in front of you will result in a lot of flying debris in a hurry. Mix in some unplowed snow or something that has leaked out of a vehicle, and there are lots of things to consider when letting the vehicle find its own way down the road. How will sensors "see" through heavy snow piling up in front of them? Think about the number of cars rolling down the road with inches of snow on the roof and a just a small spot on the window for the driver to see out. What about cleaning of salt or mud or other substances that get in the way of the sensors?

Perhaps the biggest problem and barrier to autonomous driving is that the roads will remain full of human-piloted vehicles for the foreseeable future. That adds a ton of unpredictability and variability. In a homogeneous environment when all vehicles are automated, the vehicles can "talk" to each other and the rate of predictability should increase -- although that will likely require standardized communications protocols and standardized reaction programming.

Looking forward, I think about the future vehicle (possibly a shared vehicle) with no human controls on board. Each day, I let Waze guide me on the quickest route to and from work. It's a remarkable tool, but I often deviate from it. I see a train blocking the road that Waze wants me to go on. I see flashing lights on a different road and realize that an accident just happened. Whatever. I also know that Waze can't get me to the front door of the building I work in. For some reason, it points me to a nearby building.

And here's the practical question -- where will that autonomous vehicle park? I guess in that future, the vehicle will drop you off in front of the building, then go find a place to park that really doesn't matter to the human. And who will fuel or charge the vehicle? I guess maybe there will be automated fuel or charging stations that the vehicles will seek out during the work day.

The other thing is that there really will be no more opportunity to "just go for a ride", with no destination in mind, just exploring back roads or seeing things you've never seen before. If you're not steering, trying to set up a wandering route will likely cause you to spend more time with your head down in a navigation app.

I'm not saying that the autonomous future will never come -- we're just at the birth of the technology. In ten years, I'll probably laugh at this post.
Powered By Blogger