Studies on Collision Detection and Control of Multi-link Flexible Manipulators for Their Safety Operations

在本研究中，获得了以下结果：针对柔性机械臂的非线性特性，引入Unscented卡尔曼滤波器作为非线性状态估计器，以获得柔性机械臂的状态估计。基于UKF的创新过程，成功实现了多连杆柔性机械臂的碰撞检测。由于将碰撞产生的冲击力视为未知扰动，当碰撞发生时，状态估计器的估计误差增大。当检测到碰撞时，估计误差的恶化会影响柔性机械臂的控制性能。为了提高柔性机械臂在未观察到的障碍物与臂侧碰撞时的控制性能，引入了最优干扰解耦滤波器，将不期望的碰撞力影响与估计误差解耦。结果表明，碰撞发生时的控制性能得到了成功的改善。

In this research, the following results have been obtained. Considering the nonlinearity of flexible manipulators, Unscented Kalman Filers(UKF) as nonlinear state estimators have been introduced in order to obtain the state estimate of the manipulator. The collision detection for multi-link flexible manipulator has been succeeded based on the innovation process of the UKF. Since the impact force caused by collisions are regarded as unknown disturbance, the estimation error of the state estimator increases when the collision occurs. The deterioration of the estimation error affects the control performance for the flexible manipulator when the collision is detected. To improve the control performance of the flexible manipulator when unobserved obstacles collide with the arm's side, Optimal Disturbance Decoupling Filters that decouples undesired collision force effects from the estimation error have been introduced. As the result, the control performance when the collision occurs have successfully improved.