Execution environment for real-time rule-based decision systems

In data-intensive real-time applications, it is often desirable and sometimes inevitable to explicitly compromise accuracy in order to perform fast deliberation. This lack of control of deliberation time in most expert system shells is a key aspect that limits their applicability to time-critical applications. In order to facilitate the tradeoff between execution time and the quality of a computation in rule-based reasoning, we propose an execution model for intelligent control of deliberation which provides performance guarantees subject to resource constraints. Our proactive real-time rule-based architecture budgets its resources by trading off. An example rule-based expert system--Collision Avoidance System (CAS) is used to demonstrate the proactive architecture.; The key components of our architecture are the budget planner and the MSP.RTL scheduler. The budget planner solves the top-level resource optimization problem and the MSP.RTL scheduler deals with the actual job scheduling in accordance with the computation time allocation. MSP.RTL is a tool for producing real-time schedules for a wide variety of timing constraints which are expressible in the language RTL. MSP.RTL has been used to solve a number of benchmark problems, including a sanitized version of the Boeing 777 Integrated Airplane Information Management System (AIMS) which contains 1859 task instances and 15703 timing constraints.