This paper focused on the analysis of combining several metrics criteria for the implementation of RPL ob-jective function, the new routing standard for the Internet of Things. The general problem is known as NP-complete, we propose the use of fuzzy inference system for finding a good trade-off among the various chosen metrics. Many routing solutions tend to favour increase on network lifetime, neglecting other network perfor-mance aspects. In this work, we consider : the expected number of transmission needed to successfully send a packet to its final destination, to meet reliability; the latency, to minimize end-to-end delay; in addition to the remaining power draw by node, for network lifetime extension. Implementation was done on Contiki and simulations were carried out on its emulator Cooja. Obtained results show improvements compared with those from the most common implementation, namely the one that uses ETX as unique routing metric.