Мультиагентна архітектура децентралізованого керування міським трафіком в умовах мінливості

Abstract

The report solves the urgent scientific and applied problem of increasing the efficiency of urban traffic control under conditions of high spatio-temporal volatility. The shortcomings of existing strictly periodic control systems are analyzed, and the expediency of transition to an event-driven paradigm is substantiated. A novel multi-agent architecture for decentralized control using adaptive fuzzy logic is proposed. A specific feature of the architecture is the selective activation of adaptation mechanisms (updating membership functions and rule weights) solely upon the occurrence of critical events in the traffic network. Numerical simulation proves that this approach ensures a 25.4% reduction in average vehicle delay, a 28.9% reduction in queue length, and an 18% increase in overall throughput while simultaneously reducing the computational load on the system by more than 2.3 times.