With recent advancements in communication networks, congestion control remains a research focus. Active Queue Management (AQM) schemes are normally used to manage congestion in routers. Random Early Detection (RED) is the most popular AQM scheme. However, RED lacks self-adaptation mechanism and it is sensitive to parameter settings. Many enhancements of RED were proposed and are yet to provide stable performance under different traffic load situations. In this paper, AQM scheme called Flexible Random Early Detection (FXRED) is proposed. Unlike other RED’s enhancements with static drop patterns, FXRED recognizes the state of the current network’s traffic load and auto tune its drop pattern suitable to the observed load situation in order to maintain stable and better performance. Results of the experiments conducted have shown that regardless of traffic load’s fluctuation, FXRED provides optimal performance and efficiently manages the queue. © 2020, Springer Nature Switzerland AG.