In recent decades, there has been an increasing interest in the use of robotic powered exoskeletons to assist patients with movement disorders in rehabilitation and daily life. Providing assistive torque that compensates for the user’s remaining muscle contributions is a growing and challenging field within exoskeleton control. In this article, ankle joint torques were estimated using electromyography (EMG)-driven neuromusculoskeletal (NMS) model and an artificial neural network (ANN) model in seven movement tasks, including fast walking, slow walking, self-selected speed walking, and isokinetic dorsi/plantar flexion at $60^{\circ }/s$ and $90^{\circ }/s$ . In each method, EMG signals and ankle joint angles were used as input, the models were trained with data from 3-D motion analysis, and ankle joint torques were predicted. Six cases using different motion trials as calibration (for the NMS model)/training (for the ANN) were devised, and the agreement between the predicted and measured ankle joint torques was computed. We found that the NMS model could overall better predict ankle joint torques from EMG and angle data than the ANN model with some exceptions; the ANN predicted ankle joint torques with better agreement when trained with data from the same movement. The NMS model predicted ankle joint torque best when calibrated with trials during which EMG reached maximum levels, whereas the ANN predicted well when trained with many trials and types of movements. In addition, the ANN prediction may become less reliable when predicting unseen movements. Detailed comparative studies of methods to predict ankle joint torque are crucial for determining strategies for exoskeleton control. Note to Practitioners—In exoskeleton control for strength augmentation applied in military, industry, and healthcare applications, providing assistive torque that compensates for the user’s remaining muscle contributions, is a challenging problem. This article predicted the ankle joint torques by electromyography (EMG)-driven neuromusculoskeletal (NMS) model and an artificial neural network (ANN) model in different movements. To the best of our knowledge, this is the first study comparing joint torque prediction performance of EMG-driven model to ANN. In the EMG-driven NMS model, mathematical equations were formulated to reproduce the transformations from EMG signal generation and joint angles to musculotendon forces and joint torques. A three-layer ANN was constructed with an adaptive moment estimation (Adam) optimization method to learn the relationships between the inputs (EMG signals and joint angles) and the outputs (joint torques). In the experiments, we estimated ankle joint torques in gait and isokinetic movements and compared the performance of methods to predict ankle joint torque, relating to how the methods have been calibrated/trained. The detailed analysis of the methods’ performance in predicting ankle joint torque can significantly contribute to determining which model to choose, and under which circumstances, and, thus, be of great benefit for exoskeleton rehabilitation controller design.

中文翻译：

---

使用肌电图驱动的神经肌肉骨骼模型和人工神经网络模型进行踝关节扭矩估计


近几十年来，人们越来越关注使用机器人动力外骨骼来帮助患有运动障碍的患者进行康复和日常生活。提供辅助扭矩来补偿用户剩余的肌肉贡献是外骨骼控制中一个不断发展且具有挑战性的领域。在本文中，使用肌电图 (EMG) 驱动的神经肌肉骨骼 (NMS) 模型和人工神经网络 (ANN) 模型在七种运动任务中估计踝关节扭矩，包括快走、慢走、自选速度行走和等速行走背屈/跖屈位于 $60^{\circ }/s$ 和 $90^{\circ }/s$ 。在每种方法中，都使用 EMG 信号和踝关节角度作为输入，使用 3D 运动分析的数据训练模型，并预测踝关节扭矩。设计了使用不同运动试验作为校准（对于 NMS 模型）/训练（对于 ANN）的六种情况，并计算了预测和测量的踝关节扭矩之间的一致性。我们发现 NMS 模型总体上可以比 ANN 模型更好地根据 EMG 和角度数据预测踝关节扭矩，但有一些例外；当使用来自相同运动的数据进行训练时，人工神经网络预测的踝关节扭矩具有更好的一致性。当通过 EMG 达到最大水平的试验进行校准时，NMS 模型可以最好地预测踝关节扭矩，而 ANN 在通过许多试验和运动类型进行训练时可以很好地预测。此外，在预测看不见的运动时，人工神经网络的预测可能会变得不太可靠。预测踝关节扭矩的方法的详细比较研究对于确定外骨骼控制策略至关重要。 从业者须知——在军事、工业和医疗保健应用中应用的用于增强力量的外骨骼控制中，提供辅助扭矩来补偿用户剩余的肌肉贡献是一个具有挑战性的问题。本文通过肌电图 (EMG) 驱动的神经肌肉骨骼 (NMS) 模型和人工神经网络 (ANN) 模型预测不同运动中的踝关节扭矩。据我们所知，这是第一个比较 EMG 驱动模型与 ANN 的关节扭矩预测性能的研究。在 EMG 驱动的 NMS 模型中，制定了数学方程来重现从 EMG 信号生成和关节角度到肌肉腱力和关节扭矩的转换。采用自适应力矩估计（Adam）优化方法构建了三层人工神经网络，以学习输入（肌电信号和关节角度）和输出（关节扭矩）之间的关系。在实验中，我们估计了步态和等速运动中的踝关节扭矩，并比较了预测踝关节扭矩的方法的性能，这与这些方法的校准/训练方式有关。对预测踝关节扭矩的方法性能的详细分析可以显着有助于确定选择哪种模型以及在什么情况下选择，因此对于外骨骼康复控制器的设计大有裨益。