Documentation request: how to (approximately) reduce UXsim to a point-queue / vertical-queue model

[takala4]

Could the official tutorials or documentation include a section describing how, if at all, UXsim can be configured to approximate a point-queue + FIFO model — i.e., a setting in which each link has a single bottleneck at its downstream end and no spillback to upstream links?

Comparing UXsim with point-queue DTA solvers conflates the difference in routing principle with the bias from spillback, which the point-queue side does not model.

A documented "reduced" mode would help in two situations:

1. Cross-method benchmarking, by placing UXsim and point-queue solvers on the same physical footing so the routing component of any gap can be isolated.
2. Quantifying spillover bias by running the same scenario in standard mode and in reduced mode and comparing the results.

If such a reduction is not intended to be supported, a brief note to that effect would also be useful, so that users do not invent ad hoc parameter recipes that may not match the model's design intent.

[toruseo]

Thanks for doing that analysis.

You can mimic a point queue model by simply setting huge jam_density.

from uxsim import *

W = World(
    name="",
    deltan=5,
    tmax=1200,
    no_cyclic_routing=True,
    print_mode=1, save_mode=0, show_mode=1,
    random_seed=0
)

W.addNode("orig", 0, 0)
W.addNode("mid", 1, 0)
W.addNode("mid2", 2, 0)
W.addNode("dest", 3, 0)
W.addLink("link1", "orig", "mid", length=1000, free_flow_speed=20, number_of_lanes=1)
W.addLink("link2", "mid", "mid2", length=1000, free_flow_speed=20, number_of_lanes=1, capacity_out=0.4, jam_density=1000) #<-This one
W.addLink("link3", "mid2", "dest", length=1000, free_flow_speed=20, number_of_lanes=1)

W.adddemand("orig", "dest", 0, 500, 0.8)

W.exec_simulation()

W.analyzer.print_simple_stats()

W.analyzer.time_space_diagram_traj_links(["link1","link2","link3"])

I am not sure where exactly to add this to the document. For now, I will move this issue to the Discussion page.