The problem is that in many cases, many people want to get off at the same freeway exits, and the local network doesn't have the road capacity to match.
AI doesn't really have the power to change that, and might actually be worse depending on how it reacts to pedestrians and cyclists.
The problem stems from the car AI being crazy aggressive in it's lane switching. They will switch to the lane for their exit the instant they can, even if they still have to travel around the entire city to get to it. Leading to middle lanes not getting as much use (At least without careful planning to ensure there is something they can exit the road from the middle lanes with) and potentially large backups of a single line of cars waiting to turn right (Even when the lane right next to it is also a right turn lane). They also cannot merge properly so I end up placing exits on both side of my highways (For both incoming and outgoing traffic) with one side coming up and over the road to join into the other sides exit to limit congestion.
Not that any of these things are completely unrealistic but it is occasionally annoying when your traffic is backed up solely because they're ignoring the adjacent lane.
Whatever algorithm they use is sometimes a bit too greedy. There have been times where I've been able to manually drag my path to another freeway and make it faster
Unfortunately CS suffers from many "dumb computer tricks". The pathing algorithm is open loop, often causing huge 1 lane jams even though it would be more efficient for through traffic to use other less congested lanes. You end up needing to create crazy unrealistic spaghetti interchanges in order to trick the AI into behaving
If you funnel all the traffic that needs to turn in the next 2 miles into a one-lane access road, you're just going to have them backed up past the exit on the main road.
The reason no one has implemented this system is that it wouldn't work.
> At some point, every available square foot of roadway is filled with a car. AI can help limit the amount of extra space taken up by air, but that will only moderately increase the number of cars that it will take before traffic hits. That number of cars will be reached, easily.
The thing to keep in mind is that AI can limit the extra space between the cars as well as increase the speed those cars can safely travel. The cars will spend substantially less time on the road because they'll be going faster. Double the speed and you've converted a 60 minute commute into a 30 minute commute, doubling the number of commuters the roadway can support even if you don't adjust inter-car distance art all.
A more powerful potential mechanism is that they can be programmed to respect do not enter directives from a traffic control system (keeping human drivers off a freeway would take a lot of infrastructure). So the amount of traffic on the road can be regulated to keep it below the capacity implied by the safe response time of the vehicles.
It's not random, it's poorly engineered entrance and exit ramps. In order to maintain traffic flow, cars need to be able to increase to traffic speeds on on-ramps and decrease on off-ramps. Much have the beltway system put entrance and exit ramps where they definitely don't belong. Thoroughfares for in-city traffic also don't exist, the interstate has taken that role. To compound the problem, often times entrance and exit ramps are the same piece of pavement.
Humans don’t closely approach the theoretical limits of highway capacity. They encounter a much lower limit and get stuck in a range of suboptimal choices.
It only takes a few people missing an exit and swerving to create a bunch of traffic. So many people are used to not navigating manually anymore I can’t imagine it doesn’t have a big effect.
This is meant for self-driving cars, right? Because I can not see this happen with human traffic.
I was also waiting in the video to see busier traffic simulations, but didn't see any. I suspect the model of keeping distance then breaks down completely and creates lots of congestion, ifnot collisions (if the road is full and you need to evade an oncoming vehicle...). On the upside, you use the whole road dynamically instead of only a fixed few lanes, that is a nice improvement.
I think, quite literally, one's mileage may vary. Not all stop-and-go traffic is the same.
It's not always safe to keep the huge distances apart at the speeds the system wants (other drivers will cut in). The system can also have a difficult time with motorcycles lane-splitting.
Anyway the way I see this is the switches are there for the driver to use at their discretion. I would hate for the industry to move towards full automation. I love the system as it works right now. Just my two cents.
It’s also just not realistic. If everyone was that defensive (5+ seconds distance) on such a busy road it would severely lower its capacity, leading to constant traffic jams. Leaving that gap will just mean others taking it.
much higher stakes and much less response time. There are plenty of city drivers that get confused and, well, pull over or even just stop. it's irritating, but a sensible thing to do.
On a freeway, that's terrifying.
It _seems_ like lots of parallel systems, each with the capability to reduce speed or stop could cope with city traffic about as well as a tourist. but, uh, I guess $2.5B says otherwise.
Less a "faulty algorithm" and more a moral hazard. The intent of the rules is to limit usage during periods of congestion not caused by the road operators themselves. The distinction is that human actors are not moral-free mechanisms.
This sort of traffic jam dissipation is only applicable to phantom traffic jams. Here in Los Angeles, traffic jams are caused by too many cars being on the road.
At some point, every available square foot of roadway is filled with a car. AI can help limit the amount of extra space taken up by air, but that will only moderately increase the number of cars that it will take before traffic hits. That number of cars will be reached, easily.
I've seen the same in traffic simulations.
If none of the routes are optimized for tons of traffic, 1 slow driver can gridlock an entire city at rush hour.
I suspect you haven't been in a big city much if you think thousands of stop-and-start intersections would be easier to automate than a far fewer number of continuous lanes and offramps that work via simple merging.
AI doesn't really have the power to change that, and might actually be worse depending on how it reacts to pedestrians and cyclists.
reply