制御系と機構系の同時最適化から探る適応的ロコモーションの設計論

有腿运动的稳定性的标准定义尚未达成一致，而且是迫切需要的，特别是为了量化实际硬件，而不是高度抽象的理论模型。由于固有的噪声、系统的高维性以及固有的非线性和非光滑动力学，这是一项具有挑战性的工作。开发了一种从高维运动数据中提取嵌入极限环相位的算法，并开始了数据驱动的Floquet乘法器和Lyapunov指数估计的初步工作。这些方法提供了极限环稳定性的可量化测量，允许比较不同的步态，控制器和形态。对剑桥大学一位同事收集的数据集进行的初步测试显示出了希望。

A standard definition of stability in legged-locomotion has not yet been agreed upon and is sorely needed, especially in order to quantify actual hardware as opposed to highly abstract theoretical models. This is challenging due to the inherent noise, high-dimensionality of the system, and inherent nonlinear and non-smooth dynamics. An algorithm for extracting the embedded limit-cycle phase from high-dimensional kinematic data was developed, and preliminary work for data-driven estimation of Floquet Multipliers and Lyapunov Exponents was started. These methods provide a quantifiable measurement of stability of the limit-cycle, allowing a comparison of different gaits, controllers and morphologies. Initial test on datasets collected by a colleague at Cambridge University show promise.