Design issues of adaptive and fault-tolerant real-time intelligent systems

The main objective of this paper is to describe several fundamental issues in the development of a practical and unified technology for building fault-tolerant, adaptive and distributed real-time intelligent decision systems (RTIDS) for monitoring and control of safety-critical real-time environments. These embedded systems include new generation airplane avionics, high-performance network and telephone switching systems, Space Station control software, weapons control systems, and medical monitoring equipments. This technology shall be based on a rigorous framework which facilitates formal verification and validation of these intelligent time-critical embedded systems. Real-time intelligent decision systems exhibit three characteristics which distinguish them from other decision systems: (1) bounded response time, (2) adaptive to the changing environment, and (3) fault-tolerant. An initial effort is made to tackle the difficult analysis problem of determining whether a rule-based decision system has bounded response time, but more work remains to be performed to yield a practical response time verification technology. This paper describes research results and ongoing investigations concerning the analysis problem.