Modeling and analysis of software specifications for an autonomous aerial vehicle

Abstract

The present increase in the usage of software for controlling safety-critical systems drives the need for the development of software safety analysis techniques. This thesis describes the use of a state-machine based approach for modeling and analyzing software requirements specifications, for safety related properties, of the supervisory architecture of an autonomous aerial vehicle. The requirements specification model, developed using concepts from controls engineering, is in a tabular format and specifies the intended blackbox behavior of the system in terms of discrete events. The goal of modeling was to produce a set of specifications that clearly describe the desired system behavior for ease of human review. Analysis focused on identifying safety-related errors in design philosophy rather than implementation. This was done because software requirements errors (or high level design errors) that are allowed to propagate into later phases of the development process can result in tremendous costs and are often the most difficult to correct. Results show that the symbiotic nature of the modeling and analysis techniques was useful in improving software safety. Their application led to the identification and elimination of several hazardous conditions not found during traditional software unit and system testing.