CAREER: Understanding the role of sensory feedback in gait stability through neurorobotic modeling

本项目旨在建立一个综合现有理论的运动模型，并提供一个更完整的神经系统，身体和外部环境如何相互作用的图像。神经科学在机器人领域的应用滞后于每个领域的能力和知识；这项工作通过统一这两个领域的最新技术来弥补这一点。通过神经控制器的发展，机器人技术将得到进步，神经控制器可以复制动物运动的鲁棒性和适应性。神经科学将通过在物理系统上体现神经模型而得到发展。这些进步将改善影响运动的疾病的治疗和康复方法，以及可以在非结构化环境和辅助环境中使用的机器人系统。该项目旨在通过确定稳定运动所需的最小神经回路，为四足机器人建立基于神经控制系统的基础。此外，这项工作将通过研究力学、感觉反馈和脊髓回路之间的相互作用，深入了解这些回路如何对内部和外部扰动提供强大的抵抗力。本研究的方法是将最新的运动电路的神经科学数据封装到综合的神经模型中，并在实时计算平台上运行，使用它们来控制仿生机器人，并将实验结果与神经和生物力学动物数据进行比较。具体目标是测试动态步态调节和步态转换的计算神经科学模型，并评估它们在机器人中保持稳定运动和解释动物行为的能力。为了实现这一目标，本项目的目标是：1)通过表征神经机器人、神经力学模拟和纯神经模型之间的差异，量化全身力学对理解神经控制的影响；2)建立感觉反馈和脊髓回路如何在不同扰动下影响稳定性；3)发现维持稳定性和保证成功的步态转换所必需的神经机制。该奖项反映了美国国家科学基金会的法定使命，并通过使用基金会的知识价值和更广泛的影响审查标准进行评估，被认为值得支持。

This project aims to construct a comprehensive model of locomotion that unifies current theories, and to provide a more complete picture of how the neural system, body, and external environment interact. The application of neuroscience to robotics lags capabilities and knowledge in each field; this work remedies this by unifying the state-of-the-art in both fields. Robotics will be advanced through the development of neural controllers that copy the nature of what makes animal locomotion robust and adaptive. Neuroscience will be advanced through the embodiment of neural models on a physical system. These advances will lead to improved treatment and rehabilitation methods for disorders that affect locomotion, and robotic systems that can be used in unstructured environments and assistive contexts.This project seeks to build the foundation of neural based control systems for use in quadrupedal legged robots by identifying the minimal neural circuits required for stable locomotion. Additionally, this work will provide insights into how these circuits provide robust resistance to internal and external perturbations by investigating the interplay between mechanics, sensory feedback, and spinal circuits. The approach for this research is to encapsulate the latest neuroscience data of locomotion circuits into comprehensive neural models that run on real-time computing platforms, use them to control a biomimetic robot, and compare experimental results with neural and biomechanical animal data. The specific goal is to test computational neuroscience models of dynamic gait regulation and gait transitions, and evaluate their ability to maintain stable locomotion in a robot and explain animal behavior. Towards this goal, the objectives for this project are to: 1) Quantify the impact of full-body mechanics on understanding neural control by characterizing the differences between a neurorobot, neuromechanical simulations, and pure neural models, 2) Establish how sensory feedback and spinal circuits effect stability in response to different perturbations, and 3) Discover what neural mechanisms are necessary to maintain stability and guarantee successful gait transitions.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.