欠驱动农业轮式机器人广义动力学与实时控制研究

The generalized efficient recursive dynamics model、real-time simulation and neural network-based sliding mode adaptive control system for agricultural mobile robot are researched in this study. The method of describing the under-actuated rigid-flexible multibody system based on Spatial Operator Algebra (SOA) theory was proposed in early work. In order to solve the problem of efficient dynamics model and Intelligent control of Non-holonomic constrainted Mobile Robot, the purpose of this study include four parts: Firstly, the generalized multibody is described in according to the type of the joint (active or passive), and then the generalized articulated inertia-matrix of robots system is gained based on Kalman filter algorithm, the residual forces and the generalized acceleration and torque are computed through twice tip-to-base recursive and once base-to-tip recursive, so the O(n) hybrid dynamics was designed;Secondly, the real time solver for the large differential-algebra equation is studied based on the linear multi-step method; Thirdly, an approach to adaptive trajectory tracking of mobile robots which combines a feedback linearization based on a nominal model and neural network-based sliding mode adaptive dynamic compensation is presented; Fourthly, a special simulation program of symbol software will be designed. The technique explored here should be applicable to a wide variety of dynamics model and nonlinear control problems in agriculture.

本项目的研究是围绕农业轮式移动机器人的广义动力学高效率建模、实时仿真、智能控制等工作展开。在前期研究中，基于空间算子代数理论建立了链式刚柔性多体系统广义动力学高效率建模算法。在此基础上本项目进一步扩展到带有非完整约束的轮式农业机器人应用中，主要任务包括：使用序贯滤波和光滑化最优估计理论方法对带有非完整约束轮式移动机器人系统的广义质量矩阵进行分解与求逆，探讨欠驱动轮式机器人广义动力学O（n）阶效率的建模规律；基于线性多步积分算法进行快速精确的积分算法处理大型的动力学微分-代数方程达到实时动力学仿真；采用神经网络分层滑模变结构智能控制算法，实现机器人本体的轨迹运动以及机械臂的规划控制；最后编制与开发用于农业轮式移动机器人系统可视化通用分析管理平台。该项目的是为工程设计人员提供动力学分析与综合、机械优化和设计、精确控制的良好软件，可推广领域包括微机电领域，生物领域以及航天领域等。

本项目的研究是围绕农业轮式移动机器人的广义动力学高效率建模、实时仿真、智能控制等工作展开。首先，利用图论和低序体阵列对带有6自由度的4轮式机器人进行规则性描述，使用序贯滤波和光滑化最优估计理论方法对带有非完整约束轮式移动机器人系统的广义质量矩阵进行分解与求逆，建立了欠驱动轮式机器人广义动力学O（n）阶效率的建模算法，进一步基于广义-α积分算法进行快速精确的积分算法处理大型的动力学微分-代数方程达到实时动力学仿真；其次，采用RBF神经网络对欠驱动机器人的不确定因素和关键参数进行预测，进而利用滑模变结构智能控制算法，实现机器人本体的轨迹运动以及机械臂的规划控制；最后编制与开发用于农业轮式移动机器人系统可视化通用分析管理平台。该项目的是为工程设计人员提供动力学分析与综合、机械优化和设计、精确控制的良好软件，可推广领域包括微机电领域，生物领域以及航天领域等。