Reducing the Effects of the Braess Paradox with Information Manipulation

Abstract

The Braess Paradox is a well known phenomenon in transportation engineering: adding a new road to a traffic network may not reduce the total travel time in it. In fact, some road users may be better off but they contribute to an increase in travel time for other users. This situation happens because drivers do not face the true social cost of an action. Previous works have shown that in commuting scenarios, where people use the same traffic network routinely, a continuous learning and adaptation process is a realistic scenario: road users can adapt to the traffic conditions and will eventually learn to avoid the situation in which the cost is higher. However, this learning process can take a long time. Moreover, because the process is very sensible to the cost function and to the number of agents using the network, a more efficient approach to distribute agents in the network is to let the traffic control center to acquire and process data regarding the occupancy of the available roads and compute the optimal distribution (from the point of view of the whole system). With this information, manipulated information can be passed to the road users. The interesting point is what happens when drivers simultaneously receive this kind of information and are involved in learning processes. Thus this paper reports results obtained after simulations of several situations related to the Braess scenario: only uninformed agents using the network; with different shares of uninformed agents; drivers adapting to the traffic conditions under different learning probabilities; drivers receiving forecast; and drivers receiving manipulated information.