Abstract: The swing-up control of an inverted pendulum，a Diagonal Recurrent Neural Network（DRNN） is built to identify the system and self-tuning the PD gains.At the same ranges of swing-up，the system parameters ranged from -50% to 30% for self-tuning PD control based on DRNN，from -40% to 20% for conventional PD controller.The simulation results show that the system，compared with conventional PD controller，the presented control system has great anti-jamming，adaptability and robustness.

Key words: inverted pendulum, diagonal recurrent neural network, PD control, swing-up

摘要： 针对小车一级倒立摆的起摆控制，以DRNN神经网络作为辨识器，在线自适应调整PD控制器的两项参数。在起摆范围相同的情况下，DRNN神经网络控制的倒立摆系统其模型参数变化范围为-50%～30%，传统PD控制倒立摆系统其参数变化范围为-40%～20%。结果表明，基于DRNN神经网络的PD控制器比传统的PD控制器具有较强的抗干扰能力和自适应能力，系统鲁棒性增强，效果明显优于传统的PD控制器。

关键词: 倒立摆, DRNN神经网络, PD控制, 摆起