My understanding of this function in English:

Move an agent forward to the next patch after it spent enough time on the current patch on a road(a road consists of patches). If the agent is at a corner of the road, turn to an appropriate direction. If the agent is at the finished destination, execute end-commute procedure of this agent.

rabbott wrote:

Thanks, Tanya,

Here are a couple of questions.

1. What is delay? Ricardo says it's a global variable. But the NetLogo code says its a Patch instance variable, i.e., something that patches-own. What is its function?

2. What is last-here? (in the fifth line down) It appears only once in the code on this page. How is it used elsewhere?

1. I also agreed that delay is a Patch instance variable. This variable is used in commuter-move function to compare its value with the agent's variable called tick-here(the agent has to be on the patch) to determine whether or not the agent is allowed to advance to the next patch.

2. last-here is also a Patch instance variable. Although each and every patch has it, only patches that defined as dynamic patches will utilize this variable in determine-congestion function(it appears to be the only place where last-here variable is used).

rabbott wrote:

Thanks, Tanya,

One way to find out where a variable is used is to go the actual NetLog code and search for that variable. The other use of last-here is in the procedure determine-congestion. The variable last-here records the most recent time that a commuter moved off that patch. At the start of the model, last-here is set to 0. But once an agent goes to a patch, the patch's last-here is set to the time that the agent leaves the patch. (That's what we see above.)

In that case, as Tanya said, its value is used to set the delay for the patch. Look at line 138 in the NetLogo code. (You can turn on line-numbers under Tools > Preferences.) The line in question is:

set delay floor (((250 / spawn-rate)/(ticks - last-here + 1)) * base-delay)

Can anyone explain this expression?

 

I think I can. Please correct me if I am wrong.

set delay floor (((250 / spawn-rate)/(ticks - last-here + 1)) * base-delay)

You are essentially setting the delay for a patch (or how many ticks a turtle has to stay on that patch before it can move), but it varies depending on the spawn rate and (ticks-last-here +1).
We use (250 / spawn-rate) because that is the constant used in spawn-commuters. If we use 250 as a constant for the spawn-rate, and 1000 in determine-congestion, we will end up with a lot of commuters spawning (and stacking up a lot) because the cars are stopping way too long. (It's a traffic jam!)  
In order to keep the flow nice and steady, we use 250 as the constant number between spawn-commuters and determine-congestion because it lets cars spawn and wait at relatively the same rate. (Which also means we can kind of see more of the individual cars).

The second part of the computation, or the divisor, (ticks - last-here + 1)  is the current number of ticks minus the number of ticks that was previously recorded before the last turtle left the patch. We add 1 because I'm assuming we don't want any divisions by 0, if ticks and last-here were both the same. Basically, the frequency at which turtles enter and leave the patches determines the delay. The higher the frequency, the more of a delay that patch will get. 

The final part of the computation involves multiplying the division by the base-delay , which is a constant of 10. It seems like the number 10 is used because it's makes a reasonable number when it comes to dynamic delays. The dynamic delay does not go too high nor too low which may look unnatural had static patch delays not been 10 ticks as well.

Once we have the final result, we floor it and get the largest integer less than or equal to whatever that number is and make it the final patch delay.