Bounding the inefficiency of logit-based stochastic user equilibrium (Q1037672)

The paper deals with the time inefficiency of drivers' behaviour in a congested road system. In the paper, there is assumed that the drivers decide for their possible route in accordance to their individual estimations of associated travel time, which depends either on the traversed distance or on the traffic flow along the route. If the drivers have perfect information about traffic and minimize their travel times, their behaviour leads to the deterministic user equilibrium traffic assignment. If the information is not perfect and the drivers act accordingly to individually perceived travel times, where perceived values for the same route are randomly distributed, then the drivers' behaviour leads to the stochastic user equilibrium traffic assignment. These consequences of drivers' individual decisions are compared with the system optimum flow and the stochastic system optimum flow, which minimize the total actual travel time and the total perceived travel time respectively. The authors expressed the inefficiency as a relative loss of efficiency comparing the total actual travel times of the stochastic equilibrium flow and system optimal flow and derived an upper bound of the inefficiency. They also expressed the inefficiency in terms of the economic benefit to be able to compare the stochastic equilibrium flow with the stochastic system optimum flow and also provided an associated upper bound. In the last section of the paper, the authors focused on the tightness of the inefficiency bounds and they performed an analysis of the bounds for special cases of cost function.