Bounding the inefficiency of the C-logit stochastic user equilibrium assignment

Compared with standard logit-based stochastic user equilibrium assignment model, the C-logit model describes route choice behavior in a more realistic way by considering the overlapping effect between routes. This paper investigates the inefficiency upper bounds of this model against the deterministic system optimum and the C-logit stochastic system optimum in terms of the total network travel time. It is found that the commonality factor of overlapping routes significantly affects the inefficiency bound, besides link congestion degree, total demand and the number of feasible routes. If the commonality factor is not considered, the efficiency loss resulting from selfishly stochastic travel behavior will be to large extent underestimated.