應用FPGA為基準之感測器系統於不平坦地面運動之人形機器人的動態平衡

機構典藏 > College of Engineering > Graduate Institute & Department of Electrical Engineering > Thesis > Item 987654321/52517

Please use this identifier to cite or link to this item: https://tkuir.lib.tku.edu.tw/dspace/handle/987654321/52517

Title:	應用FPGA為基準之感測器系統於不平坦地面運動之人形機器人的動態平衡
Other Titles:	Dynamic balance of a humanoid robot on an uneven ground by a FPGA-based dynamic sensing system
Authors:	林明龍;Lin, Ming-lung
Contributors:	淡江大學電機工程學系碩士班黃志良
Keywords:	人形機器人;動態平衡;加速度計;陀螺儀;humanoid robot;Dynamic balance;Accelerometer;Gyro;FPGA;Decentralized Control
Date:	2010
Issue Date:	2010-09-23 17:52:26 (UTC+8)
Abstract:	論文將進行小型人形機器人的開發，並探討其於不平坦地面的動態平衡之研究。所應用之小型人形機器人，其高度為58公分，重量為3.5公斤，共有21個自由度，包括雙手的8個自由度、腰部的1個自由度以及雙腳的12個自由度。並在身體中心位置安裝一組三軸加速度計及二組單軸陀螺儀，組合而成的姿態感測器，以量測人形機器人於(不)平坦地面的Pitch 及Roll兩軸的響應，卽時進行動態平衡的控制。以三軸加速度計量測X、Y及Z軸的重力加速度值，經過計算後可得到人形機器人當時相對於重力方向的傾斜角度，但因加速度計容易受外力干擾造成姿態誤判，必須加上陀螺儀輔助降低姿態誤判的狀況。根據以上的需求，本論文亦將應用Butterworth及Kalman 濾波器改善人形機器人姿態評估的精確性。此外，為了達成動態平衡必須快速獲得感測器資料，避免因各感測器回傳速度不同造成無謂的時間損失，是故本論文將應用FPGA(Field Program Gate Array) 進行各感測器資料接收的任務，藉由FPGA可達到平行處理的優點，並利用SoPC (System on a Programmable Chip)的架構先進行感測器的資料處理，以減少決策端的資料處理量，讓各感測器資料能以最快的速度更新，並使用具有高速鮑率的通訊埠在最短的時間回傳資料至決策端，以進行動態平衡的控制。在機器人行走的過程中，可區分為單腳支撐(Single Support Phase (SSP))及雙腳支撐(Double Support Phase (DSP))，其動態平衡會有所差異。首先，令機器人於平坦的地面走動，以獲得其中心位置之姿態響應，依據此設計動態平衡控制的參考訊號。所設計的動態平衡策略，會判斷機器人當時處於SSP或DSP狀態，進行不同的動態平衡之對策，以達到有效的平衡，避免因地面不平坦造成機器人跌倒的狀況。最後，將以包括(不)平坦地面及(沒)有啟動態平衡之相關實驗，驗證所建議方法的有效性。 In this thesis, a small size of humanoid robot (HR) is developed such that it can walk on uneven ground with the reduction of falling down. The proposed HR is height of 58 cm, weight of 3.5 kg, and degrees of freedoms (DOFs) of 21, including both hands with 8 DOFs, the waist with one DOF and both feet with 12 DOFs. A posture sensor formed by the combination of a tri-axial accelerometer and two single-axis gyros is installed in the center of gravity of an HR to measure its postures in the pitch and roll directions as it is on (un)even ground. Based on the measured posture, a decentralized balance control is on-line applied to prevent an HR tilt down. The use of tri-axis accelerometer can measure the acceleration of X, Y and Z axes. After the suitable calculation, the posture of an HR (i.e., the pitch and roll angles with respect to gravitational direction) is obtained. However, accelerometers are vulnerable to external interference caused by system noise. Therefore, two single-axis gyros are added to avoid the inaccurate measurement of the posture of an HR. Because the dynamic balance needs the quick access of sensor data for different sensors to prevent the unnecessary loss of time. Due to the advantages of parallel processing of FPGA, it is employed to receive the sensor data and to send the command for the balance. In addition, the use of SoPC (System on a Programmable Chip) for the processing of the data can reduce the computation time of decentralized balance control. For example, the communication port with high-speed baud rate can speed up the signal transmission for enhancing the effect of dynamic balance. In general, one period of the walking for an HR possesses single support phase (SSP) and double support phase (DSP). The strategies of dynamic balance for these two phases are different. It must separately discuss. In the beginning, the responses of pitch and roll axes of an HR with a stable walking on an even ground. After appropriate analysis of the corresponding posture, a reference posture for the decentralized balance control of an HR on uneven ground is obtained. Finally, the experiments of an HR on (un)even ground with (or without) balance control are arranged to confirm the effectiveness of the proposed methodology.
Appears in Collections:	[Graduate Institute & Department of Electrical Engineering] Thesis

Files in This Item:

File	Size	Format
index.html	0Kb	HTML	216	View/Open

Loading...