Enabling Reliable Detection of Failed Parts in Cyber-Physical Systems Using Unreliable Detection Sensors

Consider a problem of reliable detection of failures in Educational Cyber-Physical Systems (ECPS) consisting of several hundred computers, advanced audio and video devices and control mechanisms, as well as intelligent sensors, which are located at home and in the classrooms of the university; they can interact with each other via the Internet and serve for hybrid teaching and learning, for example, during and after the pandemic. Various probabilistic and statistical methods are known for assessing the probability of failure detection in the system, depending on the reliability of the imperfect sensors. Our approach is based on probabilistic safety assessment. To achieve the required level of safety, our key idea is to check each CPS component several times in succession. Using the formula for total probability and the Bayesian approach, we build a mathematical model for finding the minimum number of necessary repeated tests required for each component. We then develop an appropriate test scheduling algorithm and examine its complexity. Finally, we run computational experiments to detect failures in a real educational CPS serving for hybrid teaching; we compare the proposed algorithm with two known failure-detection algorithms and obtain encouraging practical results.