RIASSUNTO
Resilience is a rehashed concept in natural hazard management - resilience of cities to earthquakes, to floods, to fire, etc. In a word, a system is said to be resilient if there exists a strategy that can drive the system state back to "normal" (acceptable states) after a shock. What formal flesh can we put on such malleable notion? We propose to frame the concept of resilience in the mathematical garbs of control theory under uncertainty. Our setting covers dynamical systems both in discrete or continuous time, deterministic or subject to uncertainties. Our definition of resilience extends others, be they "a la Holling" or rooted in viability theory. Indeed, we require that, after a shock, the system returns to an acceptable "regime" , that is, that the state-control path as a whole must return to a set of acceptable paths (and not only the state values must belong to an acceptable subset of the state set). More generally, as state and control paths are contingent on uncertainties, we require that their tails processes must lay within acceptable domains of stochastic processes. We end by pointing out how such domains can be delineated thanks to so called risk measures.