Abstract: The operational characteristics of a cyber-physical system (CPS) may deviate from the certified system due to intended (in case of cheating) or unintended (wear and tear) reasons. A computer-implemented certification system makes test case search aware of the multidimensional interactions between the cyber and physical subsystems and addresses practical dynamical system problems like Zeno behavior and dynamical model divergence. The system dynamically modulates an input search space by predicting and limiting input variations that can potentially cause divergence or Zeno behavior.

Abstract: Complexities arising from interaction between multiple inter-connected components in an autonomous cyber physical system can potentially result in differences in intended and observed operations of a CPS. To perform fault analysis of a CPS it is imperative to explain this discrepancy in terms of the components of CPS control code. A system estimates a “mined” hybrid system representation of a CPS based on observed input/output traces and extracts a state machine representation of the CPS control code. The system compares the “mined” hybrid system representation with the extracted state machine representation simplification of the CPS code to identify discrepancies between expected and observed operation of the CPS. The system explains discrepancies in terms of call conditions as binary or unary operations on input and output variables and status of function call arguments.

Abstract: Safety certification of cyber-physical system (CPS) such as autonomous cars or medical control systems are complicated especially due to the lack of transparency between the manufacturer and certifying authority. The manufacturer has significant restrictions in sharing knowledge with the certification authority. Further, given time constraints, it may not be feasible for the certification authority to examine internal details of the CPS. A system models the safety certification of CPS as an agile iterative game, where a manufacturer agent acting on behalf of the manufacturer and a certifier agent acting on behalf of a third-party certification entity aims to find an optimal subset of operating data to share for accurate safety certification. The certification agent, armed with CPS model mining methods and safety assessment analysis tools, aims to accurately assess safety of the CPS with the information shared.

Abstract: Methods and systems for utilizing co-simulation of a physical model and a self-adaptive predictive controller using hybrid automata are described. For example, there is disclosed a self-adaptive predictive control system to generate an initial patient model with initial patient model settings (k1, k2, . . . , kn) and execute a program that takes as input, the initial patient model settings as configuration parameters that specify at least a plurality of initial states. Such a system further receives new patient inputs and generates hybrid automata as a variation of the initial states by applying the new patient inputs to the initial patient model. The system further derives reachable states from the hybrid automata and outputs all reachable states computed. The system further detects changes and responsively generates a new patient predictive model with new parameter settings (k?1, k?2, . . . , k?n), iteratively repeating until a termination criterion is satisfied. Other related embodiments are disclosed.

Abstract: Systems and methods for mining hybrid automata from input-output traces of cyber-physical systems are disclosed herein.

Abstract: Methods and systems for utilizing co-simulation of a physical model and a self-adaptive predictive controller using hybrid automata are described. For example, there is disclosed a self-adaptive predictive control system to generate an initial patient model with initial patient model settings (k1, k2, . . . , kn) and execute a program that takes as input, the initial patient model settings as configuration parameters that specify at least a plurality of initial states. Such a system further receives new patient inputs and generates hybrid automata as a variation of the initial states by applying the new patient inputs to the initial patient model. The system further derives reachable states from the hybrid automata and outputs all reachable states computed. The system further detects changes and responsively generates a new patient predictive model with new parameter settings (k?1, k?2, . . . , k?n), iteratively repeating until a termination criterion is satisfied. Other related embodiments are disclosed.

Abstract: Systems and methods for mining hybrid automata from input-output traces of cyber-physical systems are disclosed herein.