Article Preview
Top1. Introduction
The use of robotics systems is widespread and spans a variety of application areas. From healthcare, to manufacturing, to nuclear power plants, to space missions, these systems are typically conceived to perform difficult, rote, dangerous or critical tasks. The nature of such tasks —e.g., surgery operations, radioactive waste clean-up or space mining— places high demands on the dependability of robotics systems. Dependability is an umbrella concept which associates reliability, safety, availability, security and maintainability. However, in this paper emphasis is mainly put on the reliability and safety of the robotics systems. The latter can be seen as an extension of the former. That is, a system is in a safe state when it is in any state of correct service (the system is still reliable), or when it is in a state of incorrect service but without catastrophic consequences on the environment (Avižienis, Laprie, Randell, & Landwehr, 2004). For example, a robot arm-based hazardous waste retrieval manipulator can still function safely after failure of one of two redundant sensors in a critical joint, assuming that no compromising event occurred somewhere else.
The preoccupations in the dependability of robotics systems are not new. Fault Tree Analysis (FTA, Vesely 1981) and Failure Modes and Effects Analysis (FMEA, IEEE Std.352 1987) are among the most often used techniques in various domains of robotics. For instance, Visinsky, Walker, and Cavallaro (1993) describe the use of FTA for robots operating in remote and hazardous environments. This technique is also emphasized by Walker, and Cavallaro (1996) in the context of a radioactive waste clean-up robot manipulator. Moreover, Guiochet, Tondu, and Baron (2001) describe the importance of both FTA and FMEA in the assessments of medical robots. Other fields of application include industrial robots, like in (Karbasian, Mehr, & Agharajabi, 2012); modular and swarm robots, like in (Winfield & Nembrini, 2006), (Murray, Liu, Winfield, Timmis, & Tyrrell, 2012), and in (Bjerknes & Winfield, 2013); as well as exploration and target searching robots, like in (Yakymets, Dhouib, Jaber, & Lanusse, 2013).
The widespread use of FTA in the dependability assessment of complex systems is mainly due to the flexibility and ease of use of the fault trees. These enable the use of efficient Boolean calculus in the elimination of component failures that are irrelevant to the total failure of the system, and thereby simplifying the process to produce overall probabilities of system hazards. In this direction, a number of authors proposed to generate fault trees from system models. For example, Rauzy (2002) described an approach to compiling fault trees from mode automata which are used to capture complex behavioral aspects of systems. Another example in the same vein consists of the work in (Joshi, Vestal, & Binns, 2007). The authors produced a static fault tree generator prototype based on AADL models. AADL is an Architecture Analysis and Design Language which is intended to be an aerospace standard.