Task reliability evaluation method for autonomous intelligent system based on multi-agent approach

Objective Aiming at the problem that the task reliability of autonomous intelligent systems is difficult to evaluate accurately due to such characteristics as the decoupling of components, changing systems, and complex fault propagation laws, this paper proposes a multi-agent-based task reliability evaluation method for autonomous intelligent systems. MethodsBy constructing a reliability simulation framework based on a double-layer network structure, the separation expression and coupling correlation of the autonomous intelligent system's information propagation process and fault propagation process are realized. By designing the attributes, behaviors, and interfaces of various components in the autonomous intelligent system, a universal fault propagation and state switching mechanism is built. Based on an analysis of the factors affecting the task reliability of the autonomous intelligent system, a multi-stage task success criterion is established. Finally, based on the Anylogic simulation platform, a task reliability evaluation model for autonomous intelligent systems is developed which is capable of single process simulation and multiple Monte Carlo simulation. ResultsThe simulation results show that the proposed method can quantitatively evaluate the reliability of the typical tasks of intelligent autonomous systems, as well as exploring the key fault factors affecting such reliability. ConclusionThe proposed method can support the quantitative evaluation of the task reliability of intelligent autonomous systems and the improvement of their design.