Testing of Reconfigurable Systems: A Cognitive-Oriented Approach
Generally, a reconfigurable system is a component-based system that consists of several hardware and software components that are independently developed either on-site, or by third parties (Denaro et al. 2003). As a result, reconfigurable systems involve
- PDF / 882,895 Bytes
- 20 Pages / 439.37 x 666.142 pts Page_size
- 44 Downloads / 151 Views
Testing of Reconfigurable Systems: A Cognitive-Oriented Approach Asem Eltaher
11.1 Introduction Generally, a reconfigurable system is a component-based system that consists of several hardware and software components that are independently developed either on-site, or by third parties (Denaro et al. 2003). As a result, reconfigurable systems involve the ability to replace one or more component(s) of the Device-Under-Test (DUT), which permits new possible configurations that make the test process an expensive burden. Indeed, most of the existing test techniques are foiled by the assumption that the internal structure of the DUT is known with at least partial access to the source code (Bezerra et al. 2001). In the reality, the assumption of complete knowledge about a reconfigurable system is not generally true since several components may be supplied by a third party. Therefore, testing of reconfigurable DUTs represents new challenges that cannot be adequately manipulated by traditional testing techniques. In this context, facing the fact that “it is impossible to fully test a product” (Kaner 1997); one central issue is how to define a reasonable end to testing processes. Traditional solutions, e.g. Dalal and Mallows (1992), Gemoets et al. (1994) and Levendel (1990), end a testing cycle when the odds of detecting additional faults are below a certain threshold. Related theories tend to consider bugs as discrete objects, statistically allocated over the entire software space with a known distribution (Levendel 1990). In an effort to achieve further improvements, a statistical optimization approach is developed in Dalal and Mallows (1992) to extend this work to unknown distributions and enrich it by some graphical aids. Later on, research is done in Gemoets et al. (1994) to consider the choice of fuzzy reliability models as an optimization problem and solve it. Indeed, on one side, approaches in Dalal and Mallows (1992), Gemoets et al. (1994) and Levendel (1990) offer reasonable theoretical explanations for the test A. Eltaher (B) Institute of Control Engineering,Technische Universität Braunschweig, Hans-Sommer-Str. 66, D-38106 Braunschweig, Germany e-mail: [email protected] M. Maurer and H. Winner (eds.), Automotive Systems Engineering, DOI: 10.1007/978-3-642-36455-6_11, © Springer-Verlag Berlin Heidelberg 2013
249
250
A. Eltaher
results. But, on the other side, they inherit auxiliary problems to capture some qualitative aspects; e.g. how to define a bug, how to report bugs, the severity of an error, etc. Moreover, classical automatic testing methods discard the ability of skilled human testers to enrich test processes with intuition-based strategies. In response to this limitation, the last decade witnessed an emerging interest in studying human intelligence to mimic it during automatic testing. For example, the contribution in Gras et al. (2006) highlights an approach to interview skilled human testers to get an insight into their experiences. These experiences formulate the training sets for a learning
Data Loading...
