In this paper, we present a control scheme for robotic exoskeletons to gain the desired transparency for a wide range of tasks, which leads to the provision of sufficient support to the wearer in different situations. The main goal of the study was to evaluate the performance of adaptive neural networkbased controllers in gaining transparency for powerassist robots. As these devices experience a large dynamic change during handling different tasks/loads, we studied whether neural networks (NNs) can promptly learn the dynamics and provide sufficiently smooth commands to achieve the desired transparency. Our evaluations were performed through experiments conducted on an upperlimb robotic exoskeleton with unknown dynamics handling external loads. We tested a commonly used structure of NNs with different layers of adjustable weights and numbers of neurons. We also tested the smoothness of the system response and concluded that to gain natural and comfortable feeling the desired transparency needs to be selected in accordance with the task.

在本文中，我们提出了一种用于机器人外骨骼的控制方案，以在各种任务中获得所需的透明度，从而在不同情况下为佩戴者提供足够的支持。这项研究的主要目的是评估基于自适应神经网络的控制器在获得助力机器人透明性方面的性能。由于这些设备在处理不同的任务/负载时会经历较大的动态变化，因此我们研究了神经网络（NN）是否可以迅速学习动态并提供足够平滑的命令以实现所需的透明度。我们的评估是通过在上肢机器人外骨骼上进行的实验进行的，该外骨骼的动力学未知，无法处理外部载荷。我们测试了神经网络的常用结构，该结构具有可调整的权重和神经元数量的不同层。我们还测试了系统响应的平滑度，并得出结论，要获得自然舒适的感觉，需要根据任务选择所需的透明度。