A multi-agent-based short-cycle reliability evolution model for unmanned swarm systems

The reliability of Unmanned Swarm Systems (USSs) represents a critical research domain essential for ensuring the safety and stability of system operations. Given the dynamic nature of missions and environments, it is essential to equip the reliability models of USSs with appropriate evolutionary capabilities to enhance the credibility of the evaluation process. However, much of the work in this field is primarily designed for Unmanned Equipment (UE) or components, limiting its applicability to USSs. This study proposes a multi-agent-based short-cycle reliability evolution model for USSs. An evolution framework for the agent-based reliability model is developed to effectively integrate the evolution elements, evolution timing, and evolution strategies of the USS reliability model. The internal modules and data flow of agents are designed to describe the associated elements of USSs and support model evolution. Moreover, the study emphasizes the Adaptive Adjustment of Failure Propagation Paths (AAFPP) and the Online Addition and Removal of System Agents (OARSA), supported by associated trigger mechanisms and evolution strategies. A case study involving a 6-UAV swarm for surface reconnaissance validates the approach, demonstrating an average of 23 model evolutions per simulation, with AAFPP accounting for 54.56 % of adjustments.