Fig。 7(a) and (c) shows time-series photographs and a stick figure of typical locomotion behavior recorded by a high-speed camera。 It can be clearly observed that, at the takeoff of the left leg, the knee joint is extended and that the ankle joint is flexed (upper figures in the photographs)。 Also at the beginning of touchdown, the knee joint is  flexed。

More detailed data of kinematics and ground reaction force were recorded during 10 steps of the robot walking by both   the

Fig。 6。 (a) Schematics of the robot design, (b) photograph of the robot with a supporting rotational bar。

motion capture system and force plate (at the same sampling rates as those in the human experiment)。 The data were then aligned with respect to the ground reaction force to analyze the behaviors during stance phase (Fig。 7(b))。 Although the kinematics of    the

locomotion behavior is not as stable as that in simulation, the salient behavioral features of the proposed model are preserved (i。e。 the body excursion, the knee flex at the beginning of the stance phase, and the multiple peaks in the ground reaction force)。 The quantitative measures of the experimental results are also similar to the simulated ones; Approximately, the body excursion is 5% of the leg length; And flexion of the knee and ankle joints are 20 and 40  degrees, respectively。

Despite the difference in the quantitative measures, it is important to mention that  biped  robots  could  appreciate  a few advantages from the qualitative characteristics of the joint trajectories。 Firstly, the flexion of the knee joint at the beginning of touch down would help absorbing both intensive ground reaction force and deviations derived from the ground surface, by exploiting the elastic properties of the knee joint。 Secondly, the vertical movement of the body can be reduced compared to that of the compass gait model。

4。Discussion and conclusion

This paper presented a novel control scheme of bipedal locomotion and a set of experimental results in simulation and a

robotic platform。 Compared to the conventional PDWs, it is shown that the dynamic behavior of the proposed model is significantly comparable to that of a human, although it was tested only in planar environment (i。e。 yaw, pitch, and rotation movement are fixed)。摘要：文献综述

    双足行走的传统模式一般假设刚体结构的形式，而材料的弹性性质，似乎在本质上起着至关重要的作用。在一个新的双足行走的理论模型的基础上，本文研究了一种从生物系统获得启发的使得结构和弹性控制被动关节的最小使用模型的的双足机器人。通过对运动学和地面反作用力的分析，对模型进行了仿真和物理机器人平台的评价。实验结果表明，与被动动力学和弹性的腿设计，像人一样的一个吸引子状态步行的步态可以实现非常简单的控制，而无感觉反馈。详细的分析也解释了如何了解动态的人一样的步态可以有助于自适应的双足行走。

1。引言

    在生物力学和机器人领域，双足行走进行了研究，对我们进一步理解人类和机器人的自适应运动机制。两足行走，首先利用人工系统工程预定轨迹的腿关节。虽然这种方法证明在运动行为受到一个突出的通用性、适应性的高度的限制，因为这种方法具有精确的环境模型和计算轨迹的计算任务的要求。

被动动态行走的研究提出了传统的方法。

    基于生物力学模型，所谓的` `指南针步态模型”或` `弹道行走模型”[ 1 ]，一些被动步行（PDWS）已经开发并展示了自然行走的行为（例如，[ 24 ]）。受到在人类行走过程中的摆动腿的肌肉活动的启发，这些模型利用没有致动和纯粹的机械摆动力学来实现行走行为。它也表明，通过实施执行器在PDWS，甚至在水平底面状态下步行可实现高能源效率和小规模控制[ 5 ]。