Design and evaluation of norm-aware agents based on Normative Markov Decision Processes

Moser Silva Fagundes, Sascha Ossowski, Jesús Cerquides, Pablo Noriega

Abstract: Highlights•The NMDP framework, an extension of MDPs, for representing the knowledge of norm-aware agents in stochastic environments.•Two norm-aware agent models, which use domain-independent standard methods for estimating the rewards of executing plans and using this information to make decisions in complex stochastic environments regulated by norms.•Experiments with homogeneous and heterogeneous populations of agents, which allowed us to illustrate how our framework can be used to quantitatively evaluate and compare the agents' performance in terms of the variation of parameters of the agents and environment.