Toshifusa Sekizawa

Verification of the Deutsch-Schorr-Waite Marking Algorithm with Modal Logic

We have proposed an abstraction technique that uses the formulas of variants of the modal μ-calculus as a method for analyzing pointer manipulating programs. In this paper, the method is applied to verify the correctness of the Deutsch-Schorr-Waite marking algorithm, which is regarded as a benchmark of such analysis. Both the partial correctness and the termination are discussed. For the former, we built a system on top of the proof assistant Agda, with which the user constructs Hoare-style proofs. The system is an optimum combination of automatic and interactive approaches. While a decision procedure for a variant of modal μ-calculus, which is available through the Agda plug-in interface, enables the user to construct concise proofs, the run time is much shorter than for automatic approaches.

A Case Study: Verification of Specifications of an Embedded System and Generation of Verification Items Using Pairwise Testing

Ensuring the reliability of embedded systems has become very important. Reliability may be ensured by a number of formal methods. We study one such verification technique by applying it to an in-house development product in Mitsubishi Space Software Co., Ltd. This is a practical industrial case study, we describe our approaches and present verification results. Our aim is to check the correctness of specifications which include a set of constraints on parameters individually called an evaluation item. To that end, we adopt model checking and satisfiability checking. In our study, we set conversion rules from specifications to formal models. Part of the conversion is done by hand in this study. Manual generation limits the preparation of individual evaluation items. To overcome this limitation we present an approach for automatically generating combinations of parameters for verification by applying the pair wise testing method. Finally, we present experimental results. Note that the application of formal techniques, in this setting, is still in its preliminary stages. It is intended to develop formal techniques to the point where products may be automatically verified.

Analysis of Specification in Japanese Using Natural Language Processing

A requirement specification for software is usually described in a natural language and thus may include sentences containing ambiguity and contradiction. Problems due to the ambiguity often occur at the stage of the verification process of software development, and this forces developers to go back to the design process again. In order to prevent this kind of rework, a method of automatically converting a required specification written in Japanese to a state transition model is desired to help detect ambiguity and contradiction points of the specification. This paper proposes a method for this purpose, and reports on the result of applying the method to a specification example of an electric pot.

Equivalence Checking of Java Methods: Toward Ensuring IoT Dependability

IoT devices are software-rich and Java is sometimes chosen as the developing programming language. Although Java is highly productive in constructing large advanced programs, application or user-defined Java classes must be responsible for safety and security issues. In particular, two fundamental methods hashCode and equals play key roles in safety and security assurance. Some existing studies for ensuring the correctness of these two methods rely on static analysis, which are limited to loop-free programs only. This paper proposes a new solution to this important problem, based on equivalence checking of methods or functions. The proposed approach makes use of software analysis workbench (SAW), an open source tool. The approach is also useful in reducing the cost of regression testing when program refactoring is conducted.

Towards Verification of Robot Design for Self-localization

CPS plays important roles along with popularization. In this study, we handle an autonomous robot which estimates its position by observations in discrete two-dimensional field. Probabilistic behaviors are modeled in MDPs, and model checking results validate robot’s design.

A Case Study: Verification of an Embedded System Using Abstraction Refinement with Requirements

Along with popularization of embedded systems, reliability of them has become important. Model checking has been successfully applied to many systems for ensuring reliabilities. We had studied model checking by applying it to an in-house development product in Mitsubishi Space Software Co., Ltd. In our previous study, models are constructed in straightforward manner, and verified using model checker SPIN. However, one-step transitions have only been verified. Such verification is important to ensure fundamental behaviors, but not sufficient to ensure overall behaviors. One problem is the state explosion problem which limits verification of overall behavior. In this study, we adopt abstraction and refinement approaches to overcome the limitation. In the approach, values of signals are aggregated for abstraction, and signals of interest obtained from requirements are extracted for refinement. To demonstrate the effectiveness, we show experimental verification.

Behavior Verification of Autonomous Robot Vehicle in Consideration of Errors and Disturbances

Reliability of embedded systems has become important along with their popularization in many fields. Model checking is one of formal methods which has been successfully applied to many systems for ensuring reliabilities. In this study, we handle an autonomous robot vehicle in consideration of errors and disturbances, as an concrete example of embedded systems. Behaviors of the vehicle become essentially probabilistic in such an environment. We show that such probabilistic behaviors can be modeled in Markov Decision Processes (MDPs). To demonstrate the applicability of the modeling, we show experimental verification using the probabilistic model checker PRISM. Note that the approaches are still in preliminary stages. It is intended to establish qualitative and quantitative validations from the standpoint of design phase.

Safety Verification of Multiple Autonomous Systems by Formal Approach

We have studied verification of a line tracing robot using model checking. In this paper, we extend the model to multiple autonomous systems, and describe the advantages of applying model checking and difficulties. The targeted line tracing robot usually has only one or two sensors to detect a line painted on white background, and it traces the line according to the read value of the sensors. It is easy to trace if the line is simple straight line. However, lines sometimes become complicated by existence of random sequential corners. Those robots are often used in robot competitions for university students in Japan. Driving time, accuracy and robustness are evaluated in such competitions. The robot is usually designed as a stand-alone. Here, we extend such line tracing robots to multiple autonomous robots by adding communication functions and proximity sensors. We consider multiple lines to be crossed where robots might hit each other. Although the introduced model is simple, it has enough power to provide a structure where we can discuss safety and robustness using model checking. Our proposed method can also treat time constraints of robot controls.