For the problem of complex heterogeneous network congestion, an adaptive RED algorithm based on Bell type membership function is proposed, in the new algorithm, the probability function of packet loss is nonlinear by using the Bell type Fuzzy membership function. On one hand, it can enhance the adaptability of network and avoid cyber source waste, on the other hand, it automatically adjust the probability of packet loss according to the network environment to keep the queue length. Simulation shows that the new algorithm can effectively control the occurrence of IoT and has strong adaptivity.