This paper presents an autonomous traffic engineering framework, named ATE, a highly efficient data dissemination mechanism for multipath data forwarding in Wireless Sensor Networks (WSNs). The proposed ATE has several salient features. First, ATE utilizes three coordinating schemes: an incipient congestion inference scheme, an accurate link quality estimation scheme and a dynamic traffic diversion scheme. It significantly minimizes packet drops due to congestion by dynamically and adaptively controlling the data traffic over congested nodes and/or poorer quality links, and by opportunistically exploiting under-utilized nodes for traffic diversion, while minimizing the estimation and measurement overhead. Second, ATE can provide with high application fidelity of the network even for increasing values of bit error rates and node failures. The proposed link quality estimation and congestion inference schemes are light weight and distributed, improving the energy efficiency of the network. Autonomous Traffic Engineering has been evaluated extensively via NS-2 simulations, and the results have shown that ATE provides a better performance with minimum overhead than those of existing approaches.