Congestion Games and Selfish Routing

How latency on shared edges drives selfish routing to a Wardrop equilibrium, how to compute the social cost of a flow, and how the equilibrium compares to the socially optimal flow via the price of anarchy.

Step 1 of 157%

Tutorial

Congestion Games and Path Latency

In a congestion game, multiple users (here, aircraft) share a network of edges -- sectors, corridors, or arcs -- and the travel time on each edge depends on how many users select it. The travel time on edge ee when its flow is xex_e is called the latency of ee, written

e(xe).\ell_e(x_e).

A standard model is the affine latency

e(xe)=aexe+be,\ell_e(x_e) = a_e\, x_e + b_e,

where be0b_e \geq 0 is the free-flow travel time and ae0a_e \geq 0 measures congestion sensitivity.

The latency of a path PP at flow x\mathbf{x} is the sum of its edge latencies:

P(x)=ePe(xe).\ell_P(\mathbf{x}) = \sum\limits_{e \in P} \ell_e(x_e).

For example, if an aircraft traverses two sectors with latencies S1(x)=x+3\ell_{S_1}(x) = x + 3 and S2(x)=2x\ell_{S_2}(x) = 2x, and currently xS1=2x_{S_1} = 2, xS2=4x_{S_2} = 4, then the path latency is

P=(2+3)+(24)=5+8=13.\ell_P = (2 + 3) + (2 \cdot 4) = 5 + 8 = 13.
navigate · Enter open · Esc close · ⌘K/Ctrl K toggle