在机械臂、精细化工等工业生产过程中广泛存在着点对点(Point-to-point, P2P)跟踪问题，它要求在执行重复任务中系统输出必须经过某些特定点的跟踪目标。针对点对点跟踪问题，迭代学习控制（Iterative learning control, ILC）是最为常见的算法。当前的研究主要集中在利用点对点跟踪问题中非关键跟踪点的自由度来提升算法在批次方向的收敛速度，而缺少对点对点跟踪问题中可能面临的过程干扰和初始状态误差影响的研究。本文基于轨迹更新策略，把点对点跟踪问题描述为二维(2D)模型，以充分利用2D模型可以同时考虑两个维度的特性来分析各种算法在批次方向的收敛性和在时间方向的鲁棒性，然后进一步在2D模型下提出了针对过程干扰和初始状态误差的点对点跟踪控制方法。论文主要贡献如下：1、采用轨迹更新的策略将点对点跟踪问题描述为2DP2P模型，并分析了该时变2D模型的状态转移矩阵的性质和系统响应。在此基础上，通过在时间方向上整合模型预测控制，提出了基于轨迹更新的2D点对点综合预测迭代学习控制算法（2DP2P-IPILC）。该算法利用轨迹更新策略提升了点对点跟踪问题在批次方向上的收敛速度，还利用批次内的预测控制提升了时间方向的鲁棒性。进而利用2D理论分析了该算法的收敛条件和过程干扰情况下的跟踪误差收敛边界。在机械臂等仿真模型上进行了算法的验证。2、提出了基于状态反馈的2D点对点综合预测迭代学习控制算法，并基于2D理论分析了该算法的收敛条件和收敛边界。该算法在批次间把两个维度的状态信息引入反馈，在批次内与IPILC相结合，从而改善了系统在过程干扰和初始状态误差都存在时的鲁棒性。在机械臂模型上进行了算法的仿真验证。3、对于输入输出受限的点对点跟踪问题，提出了一种在输入约束解空间优化求解参考轨迹的可行更新策略，从而保证在任意的初始参考轨迹情况下，系统都可以收敛到特定的跟踪轨迹。进一步将该受限条件下的轨迹更新策略与2DP2P- IPILC结合，仿真结果表明，在受限的情况下该算法可以满足系统的跟踪误差要求，并提升算法在时间方向的鲁棒性。

There are many point-to-point (P2P) tracking problems in industrial processes, such as robot manipulator, fine chemical reactor, and so on. In these processes, some specific tracking points of system outputs must be reached in each batch when they carry out repetitive tasks. For the point-to-point tracking problems, iterative learning control (ILC) is a kind of most useful methods. The main research works focus on utilizing the freedom of none-critical tracking points in order to accelerate convergence of control method in batch direction. However, the disturbances and initial state errors of these process have been seldom discussed by now. This dissertation builds the two-dimensional (2D) model of the point-to-point tracking problem based on the trajectory updating strategy. The 2D model is then used to analyze the convergence in the batch domain and robustness in time domain of ILC method, because the model can exploit advantages of the method both in time domain and batch domain. Different control algorithms have been proposed based on this model, aiming at addressing the problems with disturbances and initial state errors. The main contributions of the dissertation are as follows:1. A two-dimensional point-to-point (called as 2DP2P) model is proposed by describing the point-to-point tracking problem based on the trajectory updating strategy. The characteristics of time-varying state transition matrix of the model is analyzed, and the system response of this model has been derived theoretically. Then, a 2D point-to-point integrated predictive iterative learning control (2DP2P-IPILC) method is proposed by combining model predictive control in the time domain. The algorithm improves the convergence speed in batch direction by trajectory updating, and it also enhances the robustness in time direction by integrating model predictive control. Convergence principle of the proposed method is analyzed based on 2D theory, and the boundary of output tracking errors is computed in detail. Simulations on a numerical model and a robot manipulator model are demonstrated.2. By adding the state variables to the feedback control law in the 2DP2P-IPILC, a new 2D point-to-point tracking control method is proposed. Convergence principle and tracking error boundary are analyzed with 2D theory. Because the method introduces all information of two states in different domains into the batch feedback and combines with the IPILC in time direction, the proposed method contributes to better robustness when both initial state errors and process disturbances exist. Simulations on the robot manipulator model are also illustrated.3. To address the problem of the constrained input and output in the point- to-point tracking problem, a feasible trajectory updating strategyof optimization in input constrained space is proposed. In the case of an arbitrary initial trajectory, system output can automatically converge to a specific trajectory even under the constrained conditions by using the proposed scheme. By combining with 2DP2P-IPILC, the tracking error can reach the optimal solution under the constrained conditions, and robustness and stability of the algorithm is also guaranteed at the same time.