人形機器人系統識別

System Identification of Humanoid Robots

黎純蕙(Astra Lee Chun Hui)

指導教授 ： 黃漢邦

國立臺灣大學/工學院/機械工程學研究所/碩士(2020年)

https://doi.org/10.6342/NTU202002768

摘要

人形機器人的特點在於其浮體座標系能使機器人在三度空間中移動，也能達成類似人類的動作。此一動作在進行運動計畫與控制時，需要使用完善的數學模型，在建構機器人數學模型時，其動力學參數對系統數學描述準確性扮演關鍵的角色。一般來說，此一參數係透過電腦輔助設軟體 (CAD) 設計機器人的實體模型後提供。然而，由於未建模的零件(例如電纜，螺釘和注油)，使給定的參數並不完全正確，從而導致不預期與不穩定的運動問題。本文通過提出一個二次規劃的回歸模型，克服了現有不真實參數的局限性。為了克服人形機器人的穩定性，本文也提出 Single Support Motion （單腳支撐動作）方法使人形機器人在達到穩定性的情況下同時給予不同特性之軌跡進行系統識別。首先識別系統類型並考慮物理一致性條件，其次運用物體的幾何近似來設計另一組約束條件。利用所設計之約束條件使得二次規劃運可以獲得新的識別參數。此一新參數在模擬平台 (MSC ADAMS) 和人形機器人 NINO 上進行了測試。結果顯示本文所提出的方法適用於識別人形機器人系統。

關鍵字

浮體座標系 ； 人形機器人 ； 系統識別 ； 二次規劃 ； 物理一致性 ； 單腳支撐動作

並列摘要

A humanoid robot is an extensive robot protocol from the conventional robot arm, its flexibility on moving in three-dimensional space allows robots to have natural movements like human beings. Locomotion and planning and control require a well-formulated mathematical model of robots for the implementation of studies. Generally, a robot is initially designed via computer-aided software to obtain the object’s parameters. However, the given parameters are not entirely true due
to those unmodelled parts such as cable, screws, and oiling; therefore, the imprecise parameters lead to the inaccurate mathematical model and cause undesirable motions and serious stability problems. System identification (SI) is needed and physical conditions are crucial for the guarantee of the solutions’ feasibility. The constraints are designed based on the geometric approximation of link objects. This thesis overcomes the limitations in extant unsure parameters by proposing a quadratic programming regression model accompanied by physical consistency constraints with designated excitation motion. The proposed model considers the physical conditions and design constraints generated by the geometric approximation of link objects. In order to maintain the stability of humanoid robots, the single support motion method is proposed, this method is aimed to generate exciting trajectories where the robot is under stable situations. The identified parameters were tested in the simulation platform (MSC ADAMS software) and on a humanoid robot—NINO, proven to be feasible.

並列關鍵字

Floating-base system ； humanoid robot ； system identification ； quadratic programming ； physical consistency ； single support motion

參考文獻

C. H. An, C. G. Atkeson, and J. M. Hollerbach, “Estimation of inertial parameters of rigid body links of manipulators,” in 4th IEEE Conference on Decision and Control, Ft. Lauderdale, FL, 11 – 13 Dec. 1985, pp. 990–995.

C. G. Atkeson, C. H. An, and J. M. Hollerbach, “Rigid body load identification for manipulators,” in 24th IEEE Conference on Decision and Control, Ft. Lauderdale, FL, 11 – 13 Dec. 1985, pp. 996–1002.

K. Ayusawa, G. Venture, and Y. Nakamura, “Identification of humanoid robots dynamics using floating-base motion dynamics,” in IEEE/RSJ International Conference on Intelligent Robots and Systems, Nice, France, 22 – 26 Sep. 2008, pp. 2854–2859.

C. G. Atkeson, C. H. An, and J. M. Hollerbach, “Identification of the inertial parameters of a humanoid robot using unactuated dynamics of the base link,” in Humanoids 8th IEEE-RAS International Conference on Humanoid Robots, Daejeon, South Korea, 1 – 3 Dec. 2008, pp. 1–7.

C. G. Atkeson, C. H. An, and J. M. Hollerbach, “Symbolic proof of inertia-parameter identifiability of legged mechanisms from unactuated base-link dynamics,” IFAC Proceedings Volumes, vol. 42, no. 10, pp. 693–698, 2009, 15th IFAC Symposium on System Identification.