This paper concerns the prevention of faults in discrete event systems modeled with partially observed Petri nets (POPNs) that include the definition of sensors used to measure the events and markings. Observation sequences result from this modeling, and the firing sequences and initial marking consistent with a given observation sequence are systematically obtained. The degree of confidence of past and future states and events are computed: state estimation fault diagnosis and fault prediction result from this computation. Finally, diagnosability, detectability, and predictability are defined for observation sequences and POPNs and are discussed with respect to the sensor configuration.
1 November 2019 in Journal of Loss Prevention in the Process Industries
This article develops a novel design principle for the observer of timed discrete event systems behaving under specific time semantics. Observers devoted to discrete event systems usually ignore the timing aspects of underlying systems but this can have implications to many applications, in particular, for refinement of estimation and inference tasks. The techniques of this article use the time stamps of observations to refine the state estimation process for a class of labeled and timed automata where events occur based on constant time values. The resulting timed observer is beneficial for refining privacy and security issues. Current-state opacity is discussed as a promising application of the timed observer.
21 March 2012 in International Journal of Machine Learning and Cybernetics
Recently, considerable attention has focused on enhancing the security and safety of industries with high-risk level activities in order to protect the equipment and environment. In particular, chemical processes and nuclear power generation may have a deep impact on their surroundings. In the case of major events, such as chemical spills, oil rig explosions, or nuclear accidents, collecting accurate and rapidly-evolving data becomes a challenging task. So, coordinating a fleet of autonomous mobile robots is a very promising way to deal with unpredicted events and also prevent malicious actions. This paper addresses the problem of assigning optimally a set of tasks to a set of mobile robots equipped with different sensors to minimize a global objective function. The robots perform sensing tasks in order to monitor the area and to facilitate firefighters and inspectors work if a disaster occurs by providing the necessary measures. For this purpose, a centralized Genetic Algorithm (GA) is proposed to determine the task each robot will perform and the order of execution. The proposed approach is tested through a simulation scenario of a grid map environment that represents an industrial area of the city of Le Havre, France. Moreover, a comparative study is conducted with the Hybrid Filtered Beam Search (HFBS) approach and the Mixed-Integer Linear Programming (MILP) solver Cplex. The results demonstrate that the GA approach offers a favorable balance between optimality and execution time.