Modeling Human Real Time Decisions: An Approach Based on Automatic Learning and Visual Interactive Simulation

Decisions taken in real time (control decisions) often have to be modeled in considerable detail when one wants to simulate such systems as manufacturing systems. Unfortunately, these decisions (e.g., resource allocation, job assignment, etc.) may be complex and thus difficult to model. A few approaches based on Artificial Intelligence methods have already been proposed. Unfortunately, these approaches need explicit knowledge about the decision logic, which is often difficult to obtain or formalize. To tackle this difficulty, we propose an approach that places the decision maker “in situ”. For this, a VIS is used. The real-time decisions taken by the decision-maker are collected and processed by a machine learning program. The resulting rules or decision tree characterize the decision logic and can be embedded in the simulation model to simulate decision-making processes.