Applying Formal Methods to Build a Safe Continuous-Control Architecture for an Unmanned Aerial Vehicle

Abstract

Cyber-Physical Systems (CPS) are systems composed of computational and physical processes where constant interaction with the surrounding environment exists. Unmanned Aerial Vehicles (UAVs) can be highlighted as a typical example of CPS. It contains devices that sense the surrounding environment (e.g., IMU, GPS) and provide data for the embedded continuous-control software to compute the CPS reactions. Such reactions are, in fact, actions in the physical (electro-mechanical) process, which occur using actuators (e.g., motors' speed controllers). Such a CPS is typically classified as safety-critical because a failure might have severe implications. Therefore, providing safety guarantees is of utmost importance when designing this application. This paper presents a solution for offering safety guarantees during the design of the continuous-control architecture, which is one of the most critical parts of the CPS. The present proposal applies formal verification (FV) techniques to detect software errors and verify if the architecture is suitable to cope with the real-time requirements coming from the system specification. The first verification round targets individual elements of the architecture, especially the continuous-control algorithm. Therefore, the ESBMC model checker is used; it receives the element's source code as input and can check for a set of language-specific properties, such as memory safety and concurrency vulnerabilities. After making all the individual analyses and performing subsequent corrections, another verification process is started using the UPPAAL model checker, aiming to make the schedulability analysis of the proposed architecture. Finally, we conduct a runtime monitoring analysis using our recently developed RMLib tool. This proposal was successfully used within the design process of a UAV, where different classes of design and implementation problems were detected and further corrected, as detailed in the paper.