Paper in proceedings, 2017
We present a computational model for artificial animals (animats) living in block worlds, e.g. in Minecraft. Each animat has its individual sets of needs, sensors, and motors. It also has a memory structure that undergoes continuous development and constitutes the basis for decision-making. The mechanisms for learning and decision-making are generic in the sense that they are the same for all animats. The goal of the decision-making is always the same: to keep the needs as satisfied as possible for as long as possible. All learning is driven by surprise relating to need satisfaction. The learning mechanisms are of two kinds: (i) structural learning that adds nodes and connections to the memory structure; (ii) a local version of multi-objective Q-learning. The animats are autonomous and capable of adaptation to arbitrary block worlds without any need for seed knowledge.
Multi-objective reinforcement learning