Behavior Based Estimation of Dependability for Autonomous Mobile Systems Using Particle Filter

The dependability of a system is particularly important when dealingwith autonomous or semiautonomous systems. With an increasing degree of autonomy and safety requirements, the requirements for dependability increase. Hence, being able to measure and compare the dependability of a system is more and more inevitable. Since autonomous mobile systems are usually described by their behavior it is straightforward to also define the dependability of such a system in a behavioral context. Thus, in this paper, the approach of a behavioral based definition of dependability is used together with a Particle Filter to predict the dependability of an autonomous mobile system at runtime. In Avizienis et al. [2004a] a taxonomy of Dependability and its threats is presented where the term dependability is seen as an integrated concept that further consists of attributes like availability, reliability, safety integrity etc., threats and means (see Fig. 1). The list of attributes needed for a dependable system is, however, not fix, see e.g. Candea [2003], Dewsbury et al. [2003], where slightly different attributes for dependability are defined. To evaluate the dependability of a given system one or more of the attributes is usually used, depending on the application. Depending on the research community and application of the system further attributes are added while other are neglected (see e.g. Rüdiger et al. [2007b]). Up to now two main approaches can be distinguished: • the evaluation and validation approach and • the design approach. While the first approach tries to measure the dependability on an already built system; the second approach tries to develop techniques for designing dependable hard and software systems. In Wilson et al. [2002] a classification scheme for dependable systems is proposed based on availability, data integrity, disaster recovery, and security. The system is evaluated against a Dependability Attributes Availability