This paper describes a three-layer model for organising systems of interacting agents based on situated action. The top layer of the model consists of agents performing roles, the middle layer provides the skills which agents need to perform their roles, and the bottom layer consists of the behaviours which are needed to realise these skills. The model is realised computationally by means of a production rule language called RTA which is compiled to an efficient asynchronous digital logic circuit representation so allowing multiagent simulations to operate in real time.
Wavish, P., & Graham, M. (1995). Roles, skills and behaviour: A situated action approach to organising systems of interacting agents. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) (Vol. 890, pp. 371–385). Springer Verlag. https://doi.org/10.1007/3-540-58855-8_24