Socio-economic systems are systems with distributed intelligence: Each agent makes decisions on her/his own, which together makes the system function as a whole. In contrast to many other distributed intelligence approaches, there is however no overarching problem to solve or to optimize - it is enough if the system ``works''.
The transportation system is a part of the socio-economic system, and it functions according to the same principles: travelers make autonomous decisions, and somehow the system conspires to ``work'', i.e. fulfill the transportation needs of each individual. New agent-based simulation approaches to transportation use the same principles: the simulation is composed of agents, and besides driving they also make decisions on the strategic level such as planning their daily activities and choosing their mode and route. Although it is not yet universally accepted and very few implementations exist, it seems that in future these simulations will have an agent database where each agent collects several strategies and corresponding performance knowledge, and in consequence, each agent will have only a partial and individualized view of the situation. In addition, simulation systems will allow for with-day replanning, i.e. that agents are able to change their plans spontaneously and not just over night.
For large scenarios, parallel computing is a necessity. The arguably cleanest way to do this is to have two structures: (1) A parallel traffic micro-simulation, where a regular daily traffic dynamics unfolds according to the tics of some clock. Agents in this simulation are autonomous on the tactical level. (2) Distributed modules which compute strategic decisions of the agents. These modules will compute and update strategies while the dynamics keeps unfolding. Once they have settled down on a strategy, this is communicated to the agent in the micro-simulation, which will implement it if it is still consistent with what has happened in the meantime.