有脊椎气动肌肉四足机器人机构与滑坡面步态CPG调节机理

Disasters relief is the main application field of the bionic robot research. Compliant crawling mechanism and gait planning are the keys of the bionic robot steady crawling on the landslide slopes. Based on the analysis to the hunting dog motion, we will design the compliant body mechanism and virtual prototype for the quadruped robot inspired by the dog, including the spine, the parallel trunk and the four limbs actuated by agonistic-antagonistic pneumatic muscles. Firstly, we will use Lagrange function and the floating based cooidinate method to establish the dynamics and get the system state space. Next, we will decuple the crawling motion to get basic gait using the coordinate transformation, and design the neural oscillators based on the Fourier series. The three-layer hierarchical central pattern generator (CPG) network is proposed for crawling gait planning. Spine undulation layer uses the reticular connective neures; Trunk layer is symmetrical structure; Limb layer consists of the upper and lower sub-layer connected under mono-directional coupling principle; We will study the intermittent- Laplacian- coupling method to set the upper CPG parameters; The lower CPG adopts the three agonistic-antagonistic neure pairs; We will design the posture energy reflection function to adust the trunk CPG parameters,and the reinfrocement learning method to obtain the coupling strength matrix between the limbs and trunk CPG; We will caclulate the gait stability margin using the characteristic roots of the Poincare map. Then, we will add the limit-cycle-behavior adjustment to knee joints for the local motion adjustment based on the contact force-feedback. The adjustments include the touch buffering control based on the time-domain passivity constrain, and the stress reflection mechanism to get rapid flexion of knee joints. Finaly, we will develop the robot prototype and test gaits on the landslide. This project will upgrade the bionic robot mechanism to the spine-trunk-limb coordination level, and extend the CPG network to the three-layer hierarchical structure with the contact force feedback, and increase the ability of crawling on the uneven surfaces. This robot also can be used as the locomotion platform for military robots.

灾难救援是仿生机器人应用方向.爬行机构和步态是机器人滑坡面爬行的基础问题.1)分析猎狗运动,设计脊椎,并联躯干和气动肌肉关节肢体四足机器人虚拟样机;基于浮动基建立拉格朗日动力学;2)对运动和步态解耦,基于傅里叶级数设计神经元振荡器;搭建与机构拓扑相对应的三层递阶中枢神经模式发生器CPG网络步态规划器;脊椎层为网状,躯干层结构对称,肢体层为单向耦合的上下层,间歇性拉普拉斯算子耦合分析上层CPG参数;下层CPG用3对拮抗神经元;利用能量反射调节躯干CPG参数;利用强化学习调节肢体-躯干CPG耦合矩阵;并用庞加莱映射特征根量化步态稳定裕度;3)基于极限环对膝关节引入力反馈局部调节,分时域无源性触地缓冲控制和气动肌肉关节应激反射屈曲;4)搭建机构和控制器样机测试步态.研究将仿生机构提升到脊椎-躯干-肢体协调层;将CPG扩展为力反馈三层递阶网络;提高机器人灾难现场爬行能力,为军用机器人提供移动平台.

具有CPG的仿生四足机器人具有强动态运动控制能力，在滑坡环境具有应用前景。研究工作有：1）建立Kimura振荡器CPG模型，膝髋信号由振荡器同步生成。MATLAB仿真分析了CPG参数对步态周期、幅值的影响。对气动肌肉四足机器人的反射运动建模。通过牵张反射阈值和增益调节，实现机器人的位置控制；基于神经反射机理和虚拟模型控制VMC，以肢体摆动相位和足端触地为状态切换条件，采用摆线函数规划足端轨迹，建立沟壑地形自适应步态规划方法；利用机器人躯体姿态角与关节平衡位置关系，建立了坡面稳定运动的前庭反射模型。Adams虚拟样机和实物样机实验证明了算法有效性。2）设计3-RPS型并联脊椎。利用零力矩点分析四足机器人稳定运行条件；基于相互耦合的Hopf振荡器，建立肢体层的CPG网络；建立有脊椎四足机器人的运动学模型，分析脊椎层与肢体层的耦合关系，规划脊椎扭转对角步态；利用矢量纵向稳定裕度分析得到脊椎层CPG参数值；仿真和实验验证了主动脊椎可提高四足机器人动态稳定性、协调性。3）改进Hopf振荡器，建立规划脊椎节运动轨迹的双层CPG神经网络。节律层用于规划步态，模式层用于规划肌肉屈伸。基于猎豹脊椎在对角小跑(trot)步态和疾驰(gallop)步态下对应的偏航(yaw)及俯仰(pitch)运动，设计自适应信息输入的双层CPG闭环系统；构建双层CPG输出与脊椎目标角度的拟合映射关系以及角度与肌肉驱动间的反正切运动方程。开展了Webots与MATLAB的联合仿真和气动肌肉脊椎节样机测试。4）设计含脊椎层、肢体上层和下层的三层递阶CPG网络拓扑结构，建立脊椎层与肢体层之间、肢体层内的耦合机制。推导出bound步态下速度与髋关节幅值之间的映射关系。在Webots中开展有脊椎四足机器人虚拟样机bound步态仿真，证明同时改变髋关节幅值和脊椎幅值相比只改变髋关节幅值更合理、更具仿生性。5）根据Jacobian矩阵，建立气动肌肉机械腿关节力矩模型。设计位置外环、拉力内环的串级控制、计算力矩控制策略。开展机械腿摆动控制实验和弹跳控制实验，证明机械腿可实现周期性摆动和弹跳。本项目研制的CPG和有脊椎气动肌肉四足机器人具有滑坡面应用潜力。