For almost 10 years, Statoil’s IT has worked on how to best develop and deploy what we have called systems of action, systems that are described by three key capabilities: First, they can observe a phenomena, process, or machine. Second, they process their observations in search of anomalies and deviations that must be dealt with. Third, they identify and execute the best possible actions to bring the observed phenomena, process, or machine back to its desired state. In parallel they monitor the effects of their actions and re-plan and adapt their actions based on observed effects.
Systems of action will operate with some level of autonomy; they will interact with human operators in trust-based collaborations, and they can be tasked with missions. They build heavily on the theory and concepts of rational agents and multi-agent systems as defined by Shoham and Leyton-Brown: “a combination of multiple autonomous entities, each having divergent interests or different information, or both.”
Their development involves use of cybernetics and AI technologies, including decision theory, learning, and belief representation. Architecturally they have a lot in common with what is called microservices. To ease development of systems of action, we have developed a stack model defining a capability hierarchy that we used to position applicable technologies.