Résumé : A decentralized model structure for representing groups of coupled dynamic agents is proposed, and the Least Squares method is used for fitting model parameters based on observed position data. The physically motivated, difference equation model combines effects from agent dynamics, interactions between agents, and interactions between each agent and its environment. The technique is implemented to identify a model for a group of three cows using GPS tracking data. The model is shown to capture overall characteristics of the group as well as attributes of individual group members. Applications to surveillance, prediction, and control of various kinds of groups of dynamical agents are suggested.