The controller area network (CAN) is the dominant protocol for in-vehicle network (IVN) systems because it provides bounded transmission delay among electronic control units (ECUs) at data rates between 100 Kbps and 1 Mbps. Many automotive companies have chosen the CAN protocol for their chassis network system of intelligent vehicles. However, the increasing number of ECUs in intelligent vehicles and the need for more intelligent functions require a network system with more network capacity and real-time capability. As one approach to enhance the network capacity of a CAN system, this paper introduces a CAN system with dual communication channels. This paper also presents a traffic-balancing algorithm that predicts the traffic of each channel and allocates frames to the most appropriate channel. An experimental testbed using commercial off-the-shelf microcontrollers with two CAN controllers was used to demonstrate the feasibility of the trafficbalancing algorithm.