Formal verification applied to attitude control software of unmanned aerial vehicles

Abstract

During the last decades, model checking techniques have been applied to improve overall system reliability, in unmanned aerial vehicle (UAV) approaches. Nonetheless, there is little effort focused on applying those methods to the control-system domain, especially when it comes to the investigation of low-level implementation errors, which are related to digital controllers and hardware compatibility. The present study addresses the mentioned problems and proposes the application of a bounded model checking tool, named as Digital System Verifier (DSVerifier), to the verification of digital-system implementation issues, in order to investigate problems that emerge in digital controllers designed for UAV attitude systems. A verification methodology to search for implementation errors related to finite word-length effects ( e.g., arithmetic overflows and limit cycles), in UAV attitude controllers, is presented, along with its evaluation, which aims to ensure correct-by-design systems. Experimental results show that failures in UAV attitude control software used in aerial surveillance, which are hardly found by simulation and testing tools, can be easily identified by DSVerifier.