本文针对航天领域以及工业机器人领域中广泛存在的柔性变惯量特性负载以及其对伺服电机控制性能的影响进行研究，主要内容包括：提出了一种可以统一的对单柔性关节系统、多柔性关节伺服驱动系统以及柔性连杆类伺服驱动系统等进行一致的描述的动力学建模方法，并对参数的影响进行了分析。在此基础上，分别通过有限元分析、仿真与实验等对上述建模方法和分析过程进行了验证。提出了一种分两阶段对高阶柔性变惯量伺服驱动系统进行辨识的方法。通过将系统中参数分为变化较为缓慢的静态参数以及不断变化的动态参数两类，然后分别采用低动态高精度以及高动态的方法对系统中静态参数和动态变化等进行辨识，所提出的辨识方法能够有效对系统中的快变参数以及慢变参数进行准确而快速的辨识。基于传统PI调节器对高阶柔性变惯量伺服系统控制策略进行了研究，通过将柔性考虑在内对参数进行确定。提出了一种基于模型的陷波滤波器控制方法，相对于传统陷波滤波器设计方法能够更好的对负载柔性引起的谐振进行抑制。为了在兼顾系统动态性能的同时有效实现对系统谐振的抑制，将柔性阻尼、负载转矩等干扰因素考虑在内，提出了一类基于状态反馈与转矩补偿的优化控制策略。在系统参数准确的前提下，所提出的控制策略能够在有效的保持系统动态特性的基础上实现对系统谐振的抑制，仿真和实验结果验证了所提出控制策略的有效性。针对柔性变惯量伺服驱动系统，将系统的动力学模型考虑在内，提出一种基于模型预测控制的控制策略，实时的根据当前的运行状态对接下来系统的控制输出进行估计。其中，在控制器的设计中，将负载惯量的变化等效为变化的扰动转矩，通过对扰动转矩的实时辨识反馈到模型预测控制器中进行补偿。通过引入显式模型预测控制方法大大降低了控制器的运算量。仿真和实验结果验证了所提出的基于显式模型预测控制的控制策略的有效性。

This paper focuses on the modeling, system identification and control strategies of servo system with flexibility and variable inertia load characteristic widelyspread in the field of aerospace and industrial robot applications.Firstly, a unified dynamic model for the servo system with flexible manipulator and variable inertia is proposed, and then based on the unfied dynamic model, the influence of system flexibility and variable inertia on the system performance is analyzed. Compared to the traditional modeling method, the proposed method can uniformly describe the servo system with flexible joint, multi-flexible joint and flexible links system, which greatly reduce the analysis and controller design cost for the flexible manipulator motor driven system. The proposed modeling method and analysis procedure are verified with finite element analysis, simulation and experiment results.In this paper, we propose a two-stage identification method for the high-order flexible manipulator servo system with variable inertia. Firstly, the system parameters are divided into the static parameters which remains constant during operation and the dynamic ones which keep changing during operation. Then the static parameters such as the flexible joint elasticity coefficients are identified with a method with low dynamic response but high precision. Then based on the static parameters identification results, since the number of the dynamic parameters are much smaller than the static ones, thus the system dynamic parameters such as the load inertia can be identified with much higher dynamic response. In this paper, an adaptive Kalman filter, which features high accuracy and high dynamic response, is proposed for the dynamic system identification procedure. As a result, the proposed identification method can effectively identify the static and dynamic parameters with both high accuracy and dynamic response. The proposed system identification method is verified via simulation and experimental results.As for the speed loop of high order flexible manipulator motor drive system, the application of the tradition PI controller is analyzed. Due to the limited degrees of freedom, the traditional PI regulator could not completely suppress the resonance caused by the flexibility of the load, and the controller performance of the PI regulator can only be improved with better pole placement strategy. Then a model-based notch filter which takes into consideration of the overall system dynamics is proposed. Compared to the traditional design method of notch filter, the proposed method can better suppress the resonance brought by system flexibility. Although the model-based notch filter and other passive control strategies can suppress the system resonance caused by flexibility under the condition of accurate parameters, they are at the cost of sacrificing the system dynamic response and usually are sensitive to the variance of the system parameters. In order to effectively suppress the resonance of the system while taking into account the dynamic performance of the system, in this paper, we take the disturbance factors such as flexible damping and load torque into account, and propose an optimal control strategy based on state feedback and load torque compensation. On the premise of system parameters accuracy, the proposed control strategy can effectively suppress the resonance while still maintain the high dynamic response. The simulation and experimental results verify the effectiveness of the proposed controller.As for the parameter variation problem, this paper takes the dynamic model of the system into consideration and then propose a control strategy based on model predictive control. With the MPC controller, the controller output is calculated realtime according to the current operating state. In the design of the controller, the variance of the system parameters is equivalent to the variance of the load disturbance torque, which is then estimated and fed back to the model prediction controller. In order to reduce the computation cost in the real-time model predictive control, the explicit model predictive control method is introduced to determine the parameters of the controller off-line, which greatly reduces the computation cost of the controller. The simulation and experimental results verify the effectiveness of the proposed controller based on explicit model predictive control for the flexible manipulator servo system with variable inertia.