Cooperative performance assessment for multiagent systems based on the belief rule base with continuous inputs

By dint of the advantage of deeply integrating empirical knowledge and monitoring data, the belief rule base (BRB) is widely used to assess the performance of complex systems, including multiagent systems. However, the existing paradigms for designing the inputs of BRB systems necessitate sampling discretization when faced with continuous signal inputs, which poses the risk of information loss and reduces the effectiveness of performance assessment. As such, the BRB with continuous inputs (BRB-CI) is constructed based on the continuous wavelet transform, a typical joint time -frequency analysis technique, enabling BRB systems to handle continuous signal inputs directly. The stability of the BRB-CI is proven through output error analysis. A new structure optimization strategy aimed at simplifying the BRB-CI by removing redundant belief rules is developed. Moreover, a new parameter optimization approach based on the improved gray wolf optimizer with interpretability reinforcement is devised, contributing to the interdisciplinary research on BRB systems and metaheuristic algorithms. Several experiments are conducted, demonstrating the novelty, superiority, and engineering practicability of the proposal.