Robust biomimetic models of human legs to solve high-dimensional real-time control problems

项目摘要 人体四肢的解剖学验证的肌肉骨骼模型计算速度比实时更快 这些工具将有助于推进神经修复术的控制、康复和运动神经的研究。 控制原则。然而，目前最先进的模型不能既准确又快速。我们 建议开发新一代经过验证的实时人体腿部模型， 在多维运动的整个范围内具有鲁棒性的动力学。第一个目标是验证一个 下肢模型中的全方位静态姿势。第二个目标是验证下肢模型 在动态运动任务中。 在我们之前工作的基础上，我们将使用OpenSim模型库作为 使用已发表的解剖学数据验证肌肉解剖学和功能的迭代过程。 我们希望用经过验证的模型重新创建腿部姿势的全部范围。我们将收集数据 在分裂带跑步机上由健康人执行的运动任务期间， 地面反作用力、全身运动捕捉和表面肌电图的记录， 腿部肌肉发达。该模型将在丰富的实验数据集上进行验证， 在自行步速跑步机上进行不对称踩踏时所需的能量。我们希望验证动态 通过从肌肉活动和地面实时估计观察到的全身运动学来建立模型 反作用力逆解将使我们能够估计正在进行的时空模式， 肌肉活动 在这项研究的最后，我们将开发详细的下肢模型与高迪， 肌肉路径模拟，以实时预测肢体运动。这个结果 建议将告知未来的工作，使用实时肌肉骨骼模型的发展- 增强装置的安装和运动缺陷的临床评估。

Project Summary Anatomically validated musculoskeletal models of human limbs computed faster than real-time are tools that will help advance the control of neuroprosthetics, rehabilitation, and the study of motor control principles. However, the current state-of-the-art models cannot be both accurate and fast. We propose to develop a new generation of validated real-time human leg models with musculoskeletal dynamics that are robust over the full range of multidimensional motion. The first aim is to validate a lower-limb model in the full range of static postures. The second aim is to validate a lower-limb model during dynamic locomotor tasks. Building on our previous work, we will use OpenSim model repository as a starting point for the iterative process of validating the muscle anatomy and function using published anatomical data. We expect to recreate the full range of leg postures with the validated model. We will then collect data during locomotor tasks performed by healthy humans on the split-belt treadmill with simultaneous re- cordings of ground reaction forces, full-body motion capture, and surface electromyography from rep- resentative leg muscles. The model will be validated over a rich experimental dataset for locomotor pat- terns required in asymmetric stepping on a self-paced treadmill. We expect to validate the dynamic model by estimating in real-time the observed full body kinematics from muscle activity and ground reaction forces. The inverse solutions will allow us to estimate the ongoing spatiotemporal patterns of muscle activity. At the conclusion of this study we will develop the detailed lower-limb model with high-di- mensional robust muscle path simulations to predict limb motion in real-time. The outcomes of this proposal will inform future work on the use of the real-time musculoskeletal models for the develop- ment of augmentation devices and the clinical assessment of locomotor deficits.