Spatially inhomogeneous evolutionary games provide a rich class of models for describing the interaction and movement of rational agents, such as a flock of birds or pedestrians in a congested area. In these models the agents choose their strategy based on the other agents' actions and a payoff function. Qualitative features of the individual and collective behaviour of agents can be reproduced by careful manual design of this function, e.g. repulsion at small distances. For a more quantitative approach we seek to infer the payoff function from observation data. We discuss several adaptations to the original model to make this inverse problem approachable, give a consistent mean field limit, and a consistency result for the reconstructed payoff functions.

Joint work with Mauro Bonafini and Massimo Fornasier. Supported by DFG programmes SPP 1962, SFB TR 109 and the Emmy Noether programme.