Any technical system can fail and cause damages to its users and other people at large; the more damages a failure can cause, the more hazardous the system is. Of course, engineers have always striven to lower failure probabilities, damage levels, and related hazards. But to lower something one should first measure it, so over the years they developed many methods to measure and classify all three aspects. Measuring and classifying all possible failures is an outright nightmare for laypeople, but just a difficult (they say NP-hard) problem for a rational engineer. Such problems have been tackled extensively in the operation research field, and many solutions are available today for a number of specific problems in many scientific and technological environments.

In the automotive field, this need gave rise to automotive safety integrity levels (ASILs), sets of safety requirements used to lower both the hazard level in case of failure and failure occurrence. This paper presents a hazard reduction method based on ASILs that rapidly converges to acceptable solutions, thus minimizing time, computational power, and costs. It takes its moves from the ant colony optimization (ACO) algorithm, an algorithm inspired by how ants forage around their colonies. Such an algorithm is very detailed on the one hand; on the other, it causes an explosive growth in search space size. The method proposed in this paper finds acceptable solutions in two steps: first it reduces the search space size, and then it finds acceptable results in this reduced space with a swarm intelligence algorithm derived from the ACO algorithm.

The most notable takeaway is that, although the paper refers to the automotive safety field, it seems very easy to extend all concepts and solutions described here to other fields and technological systems.