Compositional Branching-Time Measurements

Abstract

Abstract. Formal methods are used to increase the reliability of software and hardware systems. Methods such as model checking, verification and testing are used to search for design and coding errors, integrated in the process of system design. Beyond checking whether a system satisfies a particular specification, we may want to measure some of its quantitative properties. Earlier works on system measurements suggest extending model checking techniques to measure quantitative artifacts, based on weights associated with the transitions of a transition system. Other works allow counting while performing model checking or runtime verification. This paper presents a simple and efficient compositional measuring framework based on quantitative state testers. The framework allows combining multiple measures, such as distance and power consumption, using a variety of functions, such as min, max, and average. This supports calculation of interesting compound measures that quantitatively characterize a system's behavior.

Using Statistical Model Checking for Measuring Systems

“…We show now (see also [10]) an implementation of CTL model checking using our measuring principles. This algorithm resembles the original Clarke and Emerson algorithm [6], with the help of bounded model checking.…”

“…In [10], we propose a measuring specification based on neighborhoods for a state space. We associate with each state the following components:…”

“…In [10], we showed how to use such a measuring specification for calculating whether a robot can be reached into some docking station before its battery is critically discharged. There, we used the backwards propagation of values to check whether the shortest paths from each state are still short enough we used forwards propagation to check whether critical discharge (including over cycles in the path of the robot) occur.…”

“…In [10], we showed a neighborhood-based measurement scheme that is based on CTL-like specifications: creating a set of nodes in a graph corresponding to each state in the state space, and performing a measurement in a distributed manner propagating values from one node to its neighbors. When performing real-valued measurements, that scheme was limited to finite structures (weighted automata).…”

“…Specification of measurement computations. Following [10], we describe a specification formalism closely resembling a synchronous dataflow language for measurement computation. At a high level, we associate a set of measurement variables with each state in the state space.…”