Investigation of a Robust Software Architecture for Safety-Critical Applications

安全关键控制系统的软件必须高度可靠，因为故障可能会造成灾难性后果。 提出了一种新的方法，以确保这些系统的可靠性要求，使用多个程序版本，满足不同的功能。一个版本实现了系统的安全关键功能，而其他版本提供了更优化和更精细的控制。 每个程序都是一个通用关系，它计算输入的所有可能输出。 系统输出是从所有输出集的交集中获得的，并保证安全关键版本永远不会被违反。这种方法允许系统的可靠性进行评估，具有成本效益。 它还有助于将先进的自适应和学习控制算法用于非安全关键版本。 该项目包括分析的架构，语法和语义的关系程序，有效的关系组合方法，成本效益的可靠性评估方法，支持工具的编写，执行和动态组成的关系程序，和评估的多样性能力架构使用几个问题的属性。 这项研究的成功完成将为安全关键系统实现可认证的高保证提供科学严谨和实用的方法。

Software for safety-critical control systems must be highly reliable since failures can have catastrophic consequences. A novel approach is proposed for assuring ultrahigh reliability requirements for these systems using multiple program versions that satisfy different functionalities. One version implements the safety-critical functions of the system while other versions provide more optimal and refined control. Each program is a general relation that computes all possible outputs for an input. The system output is obtained from the intersection of all the output sets with a guarantee that the safety-critical version will never be violated. This approach allows the reliability of the system to be assessed cost-effectively. It also facilitates the use of advanced adaptive and learning control algorithms for the non-safety-critical versions. The project includes analysis of the properties of the architecture, syntax and semantics of relational programs, efficient relational composition methods, cost-effective reliability assessment methods, supporting tools for writing, executing, and dynamically composing relational programs, and evaluation of the diverse-capability architecture using several problems. The successful completion of this research will result in a scientifically rigorous and practical approach for achieving certifiable high assurance for safety-critical systems.