I love these little simulators because while they purport to show that slower more polite driving is more efficient, more often than not they show the exact opposite. Want to shove the most cars through per hour? Remove the speed limit. Accelerate and brake like every intersection is a drag race. Leave absolutely no distance between cars. And dump any notion of polite lane changes. NASCAR was right: Minimize unused pavement. The most efficient way to move large numbers of cars down a road is at 200mph with only inches between each car.
That's because these simulations don't include inattentive drivers. They assume everyone behaves ideally.
Normally all those things you said would result in accidents. If the simulators randomly added delays for cars braking or starting up, and then kept a death count, it would be more accurate...
A pet peeve of mine when I still drove was when drivers would leave 1-3 car length's in front of their car when stopped at a stoplight. And then after the car in front starts moving, it'd take them a solid second to start inching forward also.
The bigger problem was that this type of behavior was the norm to the point where at slow speed intersections with stoplights (e.g. city streets around college campuses), only a single digit number of cars would be able to go through per green light.
Or worse yet, they leave so much space that they don't trigger the sensors. That can mess with the traffic management scheme, sometimes preventing a light from ever changing.
> And then after the car in front starts moving, it'd take them a solid second to start inching forward also.
Indeed, the most influential slider in the simulation seems to be the “max acceleration” one... if people accelerate quickly up to their max speed it seems to prevent traffic waves from propagating backwards. It meshes well with my observations 213929387 years ago when I used to drive in traffic.
Putting more people through per stoplight doesn't really change much. You just get more people at the next stop light. The light is by far the limiting factor, not the drivers. That's why polite driving actually helps, because you get less people blocking each other and less mishaps, the things that actually slow everything down.
Or more infuriating, when you want to make a left turn, but the dedicated lane for that is unreachable until the queue in front of you moves through the intersection. And some inattentive driver ahead misses the light and leaves you with an opening that is exactly 1 foot narrower than your car.
That still doesn't mean someone leaving a bit of space or a few seconds time between you and the light changes anything. You're still just going on to another stop light afterwards, which is a much larger limiting factor and the de facto limiting factor in traffic design. Too many people misunderstand this, leading to unnecessary aggressive driving. Understanding the basics of traffic mechanics should really be part of the exam in getting your licence.
One solution might be to characterize individual drivers (agents) by running an RL experiment that mimics the interactions of individual car drivers. If the reward function is made to compare traffic patterns of the simulation vs real life, then eventually the RL model (a combination of agents) should converge to how humans drive.
So you would have various transition probabilities of a car driver moving from an attentive state into various inattentive states, with drivers having different reactions in each state.
It would also help to have a level of "variance" in individual drivers. So instead of having a bunch of drivers who are just as likely to make mistakes, you have some who transition far more easily into inattentiveness and some whose likelihood of damage/nuisance is higher than the standard driver when inattentive.
