Objective Hypoxia is one of the greatest threatens to those who come to plateau for living and working, therefore a proper amount of hypoxic training is essential for them to reduce altitude sickness caused by hypoxia before going there. This paper designs a control system for a new hypoxic chamber, which can simulate the oxygen concentration on plateau of different altitudes for hypoxic trainings and experiments. Methods The control system is based on fuzzy self-learning algorithm combined with expert algorithm and PD algorithm, with the ability to justify parameters itself to fit the environment of time-delay and instability. Results During a 12-day-test of hypoxic training, the simulating altitudes inside the chamber are controlled within expected limits, which is able to recover after interference occurred. Conclusions This system solves the problem of controlling difficulty due to the complex environment of the chamber, and provides strong robustness as well as high accuracy, which supplies a significant platform for hypoxic trainings and experiments.