基于时态测试器的实时分支时态逻辑模型检测

基于自动机理论的模型检测技术在形式化验证领域处于核心地位, 然而传统自动机在时态算子上不具备可组合性, 导致各种时态逻辑的模型检测算法不能有机整合.本文为了实现集成限界时态算子的实时分支时态逻辑RTCTL*的高效模型检测, 提出一种RTCTL*正时态测试器构造方法, 以及相关符号化模型检测算法.证明了所提出的RTCTL*正时态测试器构造方法是完备的.也证明了该算法时间复杂度与被验证系统呈线性关系, 与公式长度呈指数关系.我们基于JavaBDD软件包成功开发了该算法的模型检测工具MCTK 2.0.0.我们完成了MCTK与著名的符号化模型检测工具nuXmv之间的实验对比分析工作, 结果表明MCTK虽然在内存消耗上要多于nuXmv, 但是MCTK的时间复杂度双指数级小于nuXmv, 使得利用MCTK验证大规模系统的实时时态性质成为可能.

Model Checking for Real-time Branching-time Temporal Logic Based on Temporal Testers

Model checking techniques based on automata theory play a central role in formal verification. However classical automata are not composable for temporal operators, such that the model checking algorithms for various temporal logics cannot be organically integrated. To realize efficient model checking for real-time branching-time temporal logic RTCTL* integrating bounded temporal operators, we propose a construction method for RTCTL* positive temporal testers, and further propose the RTCTL* symbolic model checking algorithm based on positive temporal testers. We prove that the proposed construction method for RTCTL* positive temporal testers is complete. We also prove that the time complexity of the proposed algorithm is linear in the size of the verified system and exponential in the length of the given formula. Based on the JavaBDD software package, we successfully develop the model checking tool MCTK 2.0.0 for the proposed algorithm. The experimental data and the analysis results show that although MCTK consumes more memory than the famous symbolic model checker nuXmv, the time complexity of MCTK is doubly-exponentially less than nuXmv, which makes it possible to verify real-time temporal properties of large-scale systems by MCTK.