Higher coordination with less control - A result of information maximization in the sensorimotor loop

This work presents a novel learning method in the context of embodied\nartificial intelligence and self-organization, which has as few assumptions and\nrestrictions as possible about the world and the underlying model. The learning\nrule is derived from the principle of maximizing the predictive information in\nthe sensorimotor loop. It is evaluated on robot chains of varying length with\nindividually controlled, non-communicating segments. The comparison of the\nresults shows that maximizing the predictive information per wheel leads to a\nhigher coordinated behavior of the physically connected robots compared to a\nmaximization per robot. Another focus of this paper is the analysis of the\neffect of the robot chain length on the overall behavior of the robots. It will\nbe shown that longer chains with less capable controllers outperform those of\nshorter length and more complex controllers. The reason is found and discussed\nin the information-geometric interpretation of the learning process.\n