Title: Learning parametric STL specifications from execution traces: a journey to the promised land.
Speaker: Jose Ignacio Requeno Jarabo, HVL
Cyber-Physical Systems (CPS) are complex environments composed of embedded computers that supervise physical processes. Nowadays, they are ubiquitously present in most aspects of our live, ranging from power stations to communications, transport or finances. Thus, guaranteeing a correct behaviour and a good quality of service (QoS) for such critical systems is imperative. To this end, runtime verification checks the satisfaction of CPS requirements specified in a proper language and warns when the properties are violated. Current researches are focused on learning characteristics directly from execution traces, in order to validate that the current implementation aligns with the CPS design. A first approach consist on, given a template with a parametric specification, finding the valuations that fulfill and falsify the property. In this talk, I will present Signal temporal logic (STL), a suitable specification language that is adapted for dealing with continuous time real-value properties which are endemic in CPSs. I will show the recent improvements in the STL language; including a new method and tool for inferring the validity domain of parametric STL specifications.