An agent-based model of the emergence of cooperation and a fair and stable system optimum using ATIS on a simple road network

Traffic congestion threats the growth and vitality of cities. Policy measures like punishments or rewards often fail to create a long term remedy. The rise of Information and Communication Technologies (ICT) enable provision of travel information through advanced traveler information systems (ATIS). Current ATIS based on shortest path routing might expedite traffic to converge towards the suboptimal User Equilibrium (UE) state. We consider that ATIS can persuade drivers to cooperate, pushing the road network in the long run towards the System Optimum (SO) instead. We develop an agent based model that simulates day-to-day evolution of road traffic on a simple binary road network, where the behavior of agents is reinforced by their previous experiences. Scenarios are generated based on various network designs, information recommendation allocations and incentive mechanisms and tested regarding efficiency, stability and equity criteria. Results show that agents learn to cooperate without incentives, but this is highly sensitive to the type of recommendation allocation and network-specific design. Punishment or rewards are useful incentives, especially when cooperation between agents requires them to change behavior against their natural tendencies. The resulting system optimal states are to most parts efficient, stable and not least equitable. The implications for future ATIS design and operations are further discussed.