We develop a model for two-layer traffic flow in which a batch of n_{o} packets are dispatched from the source at regular intervals. This scenario is applicable in various communication and transportation systems, such as TCP congestion control mechanisms and congestion management through traffic lights. We demonstrate that implementing stochastic switching between the shortest-path and greedy approaches in the routing strategy of packet transmission results in a significant reduction in the total transmission weight when n_{o} exceeds a certain threshold. This phenomenon mirrors Parrondo's paradox, in which two games or strategies, individually yielding losses, produce a winning outcome or optimal results when combined. In addition, we observe that the influence of layer 1 on overall dynamics surpasses that of layer 2. Furthermore, we find that the impact of the structural characteristics of layers is more significant for switching probability γ<0.5.