Robots are an indispensable part of modern production lines. Usually working in designated areas, they are separated from human workers in order to avoid collisions. A safe collaborative work environment is enabled by so-called soft robots, which constitute a human-friendly alternative to classical rigid industrial robots. However, modeling soft robots proves difficult and results in a lack of accuracy for control tasks. In this contribution, a disturbance observer-based control is proposed in order to improve tracking behavior. In doing so, unknown model errors are considered as disturbances and estimated by an observer. This estimate is used to actively reject the perturbation. Different disturbance observers are designed and implemented for a soft quasi-continuum manipulator, amongst them two nonlinear disturbance observers and two extended state observers. The distinction in the versions of the same observation concept lies in the disturbance dynamics assumption, which is either considered static or as a damped double integrator. All implemented concepts combined with a PD control show superior performance compared to an existing benchmark concept based on PID-like control. Thanks to the modularity of this approach, disturbance observers can be further investigated in combination with other control strategies.

机器人是现代生产线不可或缺的一部分。通常在指定区域工作，他们与人工分开，以避免碰撞。所谓的软机器人实现了安全的协作工作环境，它构成了传统刚性工业机器人的人性化替代方案。但是，对软机器人建模非常困难，并且导致控制任务的准确性不足。为此，提出了一种基于干扰观测器的控制，以改善跟踪行为。这样做时，未知的模型误差被视为干扰，并由观察者估算。该估计用于主动拒绝扰动。为软准连续机械手设计和实现了不同的干扰观测器，其中有两个非线性干扰观测器和两个扩展状态观测器。同一观测概念的版本之间的区别在于干扰动力学假设，该假设被认为是静态的或被视为阻尼双积分器。与基于类PID控制的现有基准测试概念相比，与PD控制相结合的所有已实现概念均显示出卓越的性能。由于这种方法的模块化，可以结合其他控制策略进一步研究干扰观测器。与基于类PID控制的现有基准测试概念相比，与PD控制相结合的所有已实现概念均显示出卓越的性能。由于这种方法的模块化，可以结合其他控制策略进一步研究干扰观测器。与基于类PID控制的现有基准测试概念相比，与PD控制相结合的所有已实现概念均显示出卓越的性能。由于这种方法的模块化，可以结合其他控制策略进一步研究干扰观测器。