Cooperative Adaptive Cruise Control (CACC) is a promising technology for improving the capacity and energy efficiency of the ground transportation system. In this paper, a novel CACC control scheme is proposed to deal with the adverse impacts of both inter- and intra-vehicle network delays. First, a hetero-integration poly-net (PN) loop delay analysis method is presented to clarify the system delays in CACC considering both inter- and intra-vehicle network influences. A mathematic equation is put forward to calculate the upper bound of the PN loop delays. Then a collaborative software-defined network scheme is presented to deal with the PN loop delays, which consists of the application/strategy, network-control and network-data planes. In the network-control plane, a fraction-type basic period scheduling method is adopted. In the application/strategy plane, a delay-tolerant model predictive controller is designed for decision-making while a combination of an H ∞ controller and a linear quadratic regulator is adopted for acceleration tracking control against local intra-vehicle network delays. Finally, the proposed scheme is verified under a variety of scenarios based on comprehensive Hardware-in-the-Loop tests.