Fridays

جهت مشاهده Writing تصحیح شده فایل PDF را دانلود نمایید.

In this work, we aim at solving the barrier-based safety problem for a potentially unknown stochastic dynamics through finite data collected from trajectories of the system ; and providing a lower bound on the safety of the system accompanied by an a priori confidence. To do so, in the first step/initially, we reformulate the barrier-based safety problem as a robust convex problem (RCP) and then we stick to the scenario-driven approach to tackle this problem. We take samples from the system and construct a scenario convex problem (SCP) and then webefore establishing a connection between the optimal solution of the obtained SCP and safety of the stochastic system. There are two sources of uncertainty, ; one comes from data and another the other one is caused by the stochasticity in the system. We incorporate the effect of both of them in our provided confidence. In continuesubsequently, the proposed approach is extended for a class of non-convex safety problems. Finally, we provide an algorithm in order to synthesize a safety-guided controller based on collected data. We conclude bring the paper to the end with two case studies to elucidate the effectiveness of our approach in practice.