Abstract

This paper investigates Cohen-Grossberg neural networks with generalized piecewise constant delay (DEGPCD), motivated by hybrid and sampled-data neural systems. The problem addressed is the existence and global exponential convergence of a-periodic solutions under nonuniform piecewise constant delay structures. By employing Mawhin's coincidence degree theory instead of classical Schaefer-type fixed-point methods, we establish the existence and uniqueness of periodic orbits without imposing global upper-slope bounds on the leakage functions. Furthermore, a Gronwall-type inequality adapted to the DEGPCD framework is developed to derive explicit and verifiable exponential stability criteria. Numerical examples for two-and three-dimensional networks are provided to illustrate the effectiveness and applicability of the theoretical results. The proposed approach offers a unified and computable framework for periodic dynamics in hybrid delayed neural networks.