双足机器人平衡控制及步态规划研究

发布时间：2018-07-01 13:32

本文选题：双足机器人 + 平衡控制　；参考：《电子科技大学》2017年硕士论文

【摘要】：驱动技术,人工智能,高性能计算机等最新技术已经使双足机器人有了粗略模拟人体运动的灵巧性,能够进行舞蹈展示,乐器演奏,与人交谈等。然而这与投入实际应用所需求的能力还有不小差距。主要体现在缺乏与人类相近的平衡能力和步伐协调能力,对工作环境要求高,在非结构化环境中适应能力差。因此,本文以自主研制的双足机器人为研究对象,重点研究了双足机器人的平衡控制,阻抗控制以及步态规划等内容。本文首先简要介绍了自主研制的双足机器人的软硬件构架,建立了ADAMS和Gazebo仿真来协助对控制算法性能预测和优化并减少对物理机器人的危险操作。接着分析了双足机器人的正逆运动学并引入运动学库KDL来简化运动学运算。稳定的平衡控制对于双足机器人而言在目前还是个不小的挑战。本文就此研究了两种处理平衡的阻抗调节方案。一种是基于LQR的固定阻抗模型,这种方案简单有效,但存在易产生振动的问题,本文结合滤波改善了平衡控制效果。另一种是基于增强学习的自适应阻抗模型。该方法可以在不知道系统内部动态信息的情况下利用迭代策略在线得到最优解,是对前述LQR方法的进一步优化。随后本文通过仿真和实验进行了验证并分析了优缺点。步态规划是机器人运动控制中最基础的一环。本文从五连杆平面机器人入手对其运动控制进行了研究。首先采用基于ZMP的多项式拟合法实现了机器人平地行走的步态规划。然后分析其动力学模型并利用PD控制器进行运动仿真,就仿真中出现双腿支撑阶段跟踪误差较大的问题提出了PD与径向基神经网络混合控制的新策略。再次通过仿真证实该方案能够减小跟踪误差。最后,本文利用前述多项式拟合法对实验平台的物理机器人进行静态行走和上楼梯的步态规划。针对上楼梯的步态规划的特殊性,本文提出了分段拟合来实现各关节的协同规划,并引入了躯干前倾角来辅助身体平衡。由于时间所限,本文实现了双足机器人的稳定步行实验,上楼梯实验还尚缺稳健性,这将作为下一步的工作。
[Abstract]:The latest technologies, such as driving technology, artificial intelligence and high performance computer, have enabled biped robots to have the dexterity of simulating human motion roughly, to perform dance displays, to play musical instruments, to talk to people, and so on. However, there is still a big gap between the capacity required for practical applications. It is mainly reflected in the lack of balance ability and step coordination ability which is similar to that of human beings, the high requirement of work environment and the poor adaptability in unstructured environment. Therefore, this paper focuses on the balance control, impedance control and gait planning of the self-developed biped robot. In this paper, the software and hardware architecture of the self-developed biped robot is briefly introduced, and Adams and Gazebo simulations are established to help predict and optimize the performance of the control algorithm and reduce the dangerous operation of the physical robot. Then the forward and inverse kinematics of biped robot is analyzed and the kinematics library KDL is introduced to simplify kinematics operation. Stable balance control is still a big challenge for biped robots. In this paper, two kinds of impedance control schemes are studied. One is a fixed impedance model based on LQR, which is simple and effective, but easy to produce vibration. The other is an adaptive impedance model based on reinforcement learning. This method can obtain the optimal solution online without knowing the dynamic information of the system. It is a further optimization of the LQR method mentioned above. Then, the paper verifies and analyzes the advantages and disadvantages by simulation and experiment. Gait planning is the most basic link in robot motion control. In this paper, the motion control of five-bar planar robot is studied. Firstly, the plane-walking gait planning of the robot is realized by using the polynomial fitting method based on ZMP. Then the dynamic model is analyzed and the PD controller is used to simulate the motion. A new hybrid control strategy of PD and radial basis function neural network is proposed to solve the problem of large tracking error in the leg support stage. The simulation results show that the scheme can reduce the tracking error. Finally, this paper uses the polynomial fitting method to carry out the static walking and gait planning of the physical robot on the experimental platform. In view of the particularity of gait planning for stairs, a piecewise fitting method is proposed to realize the joint coordination planning, and a torso anteversion angle is introduced to assist the body balance. Due to the limitation of time, the steady walking experiment of biped robot is realized in this paper, and the stair experiment is still short of robustness, which will be the next step.
【学位授予单位】：电子科技大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP242

【参考文献】