雙足機器人站立姿態擾動平衡控制

Abstract

本論文的主要目的為為實驗室展開中形人型機器人的研究，以期在機器人研究領域上有新的一步。 人形機器人開發後的第一步首先要能成功站立，並可應付外界一定程度的干擾還能維持不倒，因此站立姿態擾動平衡恢復控制將為本文首要探討的運動規劃。人形機器人成功站立後，鎖定機器人膝關節，僅踝關節和髖關節可制動，由軌道能量的概念，針對外界水平向擾動後系統質心的位置和速度狀態去計算系統軌道能量，由此導出系統能量和地面壓力中心的關係，並賦予此壓力點為腳底的瞬時擷取點，當瞬時擷取點都維持在壓力中心臨界值所包圍的腳底支撐多邊形內部時則應用倒單擺系統所推展出來的踝關節穩定策略來維持機器人的恢復穩定；當超過壓力中心可能承受的臨界後，則切換到線性飛輪倒單擺所推展出來的髖關節穩定策略來導回機器人的平衡。 踝關節和髖關節的制動扭矩由虛擬模型控制導出。針對俯仰軸面而言，上軀幹的質心僅受水平向、垂直向和俯仰軸的旋轉三個虛擬作用力，依虛擬腳指點合力矩為零的概念去推算水平向虛擬作用力，加上垂直向補償重力和關節策略下髖關節對應俯仰軸的旋轉虛擬力矩，三者即可依正向運動學概念導出各關節的相應扭矩。 本論文應用Solidworks對機器人平台的機構做規劃與設計，以此設計好的組合件賦予材質等物理特性後，匯入matlab之simulink工具庫中所提供的SimMechanics模組，快速建立符合設計需求的虛擬實體，以此進行運動模擬可加快整個機電系統開發設計的速度與模擬可視化的直覺性。The thesis aims the development of the human-like robot for the lab, and hopes to set a new milestone of the research on humanoid robot. After developing the bipedal robot, the first step is to make robot stand up successfully and be able to fight against the disturbances to preventing falls. To this end, the balance control for fall prevention of the bipedal robot upright posture is the prime consideration of motion control. With the concept of the orbital energy, we figure out the system state of the position and the velocity to calculate the orbital energy and define it as the Instantaneous Capture Point. When the Instantaneous Capture Point stays in the support polygon under the foot, the ankle strategy derived from the linear inverted pendulum model is applied to maintain the balance of the robot. If the Instantaneous Capture Point is over the boundary of the center of pressure, it switches to the hip strategy derived from the linear inverted pendulum model plus fly wheel to get push recovery. The values of the ankle joint and hip joint are derived under the approach of Virtual Model Control. To the pitch axis side, the torso is applied by the virtual forces caused by the horizontal virtual force, the vertical virtual force of gravity-compensation, and the virtual torque. With the concept that the resultant torque to Virtual Toe Point is zero, it could be used to calculate the horizontal virtual force. Combine these three virtual forces, the related joint torques would be derived with the forward kinematic. I design the structure of the bipedal robot, define material properties to the structure by Solidworks, and transport it into SimMechanics, one of tool boxes plugged in Simulink of matlab. It shortens the time to build the virtual model for simulation and the result will be more visualized.