Artificial Life Laboratory, Graz, Austria

  One of the main challenges in automatic controller synthesis is to
  develop methods that can successfully be applied for complex
  tasks. The difficulty is increased even more in case of settings
  with multiple interacting agents. We apply the Artificial
  Homeostatic Hormone Systems (AHHS) approach, which is inspired by
  the signaling network of unicellular organisms, to control a system
  of several independently acting agents decentrally. The approach is
  designed for evaluation-minimal, artificial evolution in order to be
  applicable to complex modular robotics scenarios. The performance of
  AHHS controllers is compared to NeuroEvolution of Augmenting
  Topologies (NEAT) in the coupled inverted pendulums benchmark. AHHS
  controllers are found to be better for multi-modular settings. We
  analyze the evolved controllers concerning the usage of sensory
  inputs, the emerging oscillations, and we give a nonlinear dynamics
  interpretation. The generalization of evolved controllers to initial
  conditions far from the original conditions is investigated and found to
  be good. Similarly the performance of controllers scales well even
  with module numbers different from the original domain the
  controller was evolved for. Two reference implementations of a
  similar controller approach are reported and shown to have
  shortcomings. We discuss the related work and conclude by
  summarizing the main contributions of our work.