We evaluate the effectiveness of real planning and control about the traffic in Hiroshima city by simulating the traffic flows in the entire city. Our massively parallel simulation of minute behavior of individual drivers is made credible by the use of novel techniques for learning the parameters of simulation models from available data.
