Gallay, Olivier2010-09-302010-09-302010-09-302010https://infoscience.epfl.ch/handle/20.500.14299/54541To analyze the potential outcomes resulting from interaction between autonomous decision-making "smart parts" in logistics networks, we propose here an exactly solvable stylized model that is able to quantify how much the dynamics can be enhanced by (fully decentralized) agent-based mechanisms. Cost functions are introduced in order to compare the performances of centralized versus decentralized organization and we are enable to conclude that for time horizons shorter than a critical value, multi-agent interactions generate smaller costs that an optimal effective central controller.Transportation logistics"Smart parts" dynamicsMulti-agent systemsMean-field approximationBurgers' nonlinear evolutionEmergent cooperative patternOptimal stochastic controlKullback-Leibler relative entropyComplex Adaptive SystemsSupply NetworksPerspectivetext::conference output::conference presentation