Supply Chain Competitive Balancing Using Finite‐Time Zeroing Neural Networks: A Dynamic System Perspective

ABSTRACT

Inventory equilibrium in supply chains can be formulated as a time‐varying multi‐agent competitive resource‐allocation game between supply and demand agents. Its real‐time solution is challenging due to heterogeneous, coupled inequality constraints, and time‐varying parameters. This paper develops a zeroing neural network (ZNN)‐based computational framework for this class of problems. The key feature of the framework is an exponential auxiliary‐variable transformation, through which the original time‐varying matrix inequalities are reformulated as equivalent time‐varying matrix equations suitable for ZNN computation. A finite‐time convergence mechanism based on a sign‐bi‐power activation function is then employed to obtain finite‐time tracking of the transformed system. Lyapunov analysis establishes finite‐time convergence and provides an explicit upper bound on the settling time. Numerical simulations and comparative studies are conducted to evaluate convergence accuracy and computational performance, while application experiments demonstrate its capability to compute feasible competitive decisions online under time‐varying operational constraints and examine robustness to disturbances and problem scale. The study therefore contributes a unified inequality‐to‐equation ZNN framework and its application to real‐time competitive resource allocation in supply chain systems.