Highly reliable relational control programs for robust rapid transit systems

High reliability, evolvability, and adaptability are important criteria for safety-critical rapid transit systems due to the catastrophic cost of failures. The conventional organization of digital control systems makes it difficult to achieve a high level of reliability since failures in noncritical components can trigger failures of safety-critical functions. Also, repairing or evolving a component is difficult due to the complex interactions between components. In this paper, we present an architecture consisting of a society of Independently Developable End-user Assessable Logical (IDEAL) agents. We use relational composition to achieve rigorous component composition and system adaptation. It facilitates the development of fine-grain agents whose behaviors are directly traceable to the requirements specification and achieves highly reliable software, fault isolation, and controllable component evolution. The approach is applied to a case study specified by the High-Integrity Systems group at Sandia National Labs. It involves the control of a future version of the Bay Area Rapid Transit (BART) system where trains must be run closer together without compromising system safety and reliability