本项目以双足机器人为研究对象，以多体动力学、非光滑动力学、随机非线性动力学、神经网络动力学以及控制理论和方法为基础，开展对双足机器人多源随机不确定性建模和智能控制研究。结合双足机器人实际步行环境，模拟产生双足机器人多源随机不确定性信号，并将此类信号映射到机器人各驱动关节的力矩中，实现对多源随机不确定性信号与确定性数据的有效融合，生成机器人各驱动关节的实际输入力矩；分别建立含多源随机不确定性的双足机器人动力学模型以及神经网络模型，并对系统进行稳定性分析，进而对模型进行优化；基于对双足机器人动力学特性的分析结果，发展基于神经网络模型的控制策略，编写相应的控制算法以对双足机器人进行实时在线控制；对双足机器人进行动力学仿真实验，验证上述理论和方法的可行性以及数值计算结果的正确性，以实现双足机器人在多源随机不确定性环境下的稳定行走。

Based on the theories and methods of multi-body dynamics, non-smooth dynamics, stochastically nonlinear dynamics, neural network dynamics and control, this project aims at the multi-source random uncertainty modeling and intelligent control for biped robot. Taking the real walking surroundings of the biped robot into account, random uncertain signals from single and multiple sources will be simulated, which are then mapped into the actuating torques on the joints of the biped robot to generate the real input torques by mixing the random uncertain signals with the deterministic data effectively. By establishing the dynamical model and neural network model of the biped robot with multi-source random uncertainties, the stability of the biped robot will be studied in detail, from which the models will then be optimized. Based on the dynamical analysis on the biped robot, the neural network model will be employed to develop the control strategy, and the corresponding control programs will be written to accomplish the real-time online control on the biped robot. Simulation experiments will be performed on the dynamics of the biped robot to validate the theoretical predictions and numerical results and realize the stable bipedal walking under multi-source random uncertain surroundings.