Future safety-critical systems, used in, f … Future safety-critical systems, used in, for example, the aerospacial, aeronautic and
automotive industries, call for innovative computing architectures, with increased complexity. These systems must still cope with strict requirements, not only in terms of safety
and reliability, but also in terms of size, weight and power consumption (SWaP).
Traditional approaches used in the design of such critical systems, rely on proving and
guaranteeing, at design time, the safety and predictability of their applications. However,
with the emergence of new technological solutions and the increase of the complexity of
applications, it gets harder or even infeasible to prove their safety by design, limiting the
scope and possible features to include in such systems. For instance, the use of wireless
communications opens a new world of possibilities: it may be used to develop smart
vehicles that cooperate with each other to achieve some common goal. However, due to
its uncertainty, the development of such applications for safety-critical systems turns out
to be a challenging task.
In this thesis, we propose a hybrid architecture, in which simple and predictable components coexist with complex and unpredictable ones, without compromising safety, despite the unavoidable uncertainty. The inclusion of complex components into safety-critical systems allows the emergence of new applications that provide new features or
that improve the existing ones. Furthermore, we want to deal with the uncertainty that
characterizes wireless communications and provide mechanisms which allow systems to
cooperate with each other in a safe way.
We rely on a component called Safety Kernel, in charge of monitoring and managing
the runtime conﬁguration of the system, forcing it to adapt to faults and runtime constraints in order to avoid hazardous situations. We describe the architecture and role of
such Safety Kernel, and how they interact with other components in the system architecture, including the functional components of the control system. Finally we present a
prototype implementation of such Safety Kernel over AIR, an architecture based on the
concept of Time- and Space Partitioning (TSP) developed for aerospace systems. ing (TSP) developed for aerospace systems.