人形機器人之節能步態生成、多目標全身控制與安全性

Abstract

首先，藉由最小化由關節力矩，關節角度限制，關節角度速度限制所構成的成本函數，提出了一個質心高度軌跡最佳化的方法。質心高度軌跡的最佳化過程是採用了梯度下降法。由於最佳化的過程無法即時進行，本文提出了將機器人步伐切分為單元步的方式，並利用此方式建立對應各單步的質心高度軌跡資料庫，運用此資料庫可以線上即時生成所需的質心高度軌跡，用來使步態生成器得以生成節能步態。經過不同步速的實驗證明，與固定質心高度軌跡相比，由此方法生成的節能質心高度軌跡可以減少能量的消耗至14%，因此方法確實是有效的。 接著我們提出了基於二次規劃的整合型控制器，藉由比例－微分控制與線性二次狀態增量步態生成器等多種控制方法在機器人動量上的規劃，可確保機器人在執行任務期間的穩定性。除此之外，根據重心力矩樞軸 (Centroidal Moment Pivot, CMP)，機器人可以算出所需要的水平補償力矩，並透過整合控制器進行補償來增加機器人行走的穩定性。 最後，我們為人形機器人提出了一套智慧型的錯誤偵測、診斷和健康評估系統。此系統使用主成分分析擷取感測器資訊並使用尼爾森法則 (Nelson rules) 進行線上的偵測，一旦偵測出異常現象，機器人會開始進行設定好的診斷動作，取得特徵並使用多類別支撐向量機來進行錯誤的診斷。另外，我們也設計了一個基於模糊邏輯的機器人健康指數產生器，可以用來量化機器人目前的健康狀態。接著將本系統套用在我們的人形機器人上進行測試，並且加入了幾個與老化相關的測試條件。

Firstly, a Center of Mass (COM) trajectory optimization method is proposed to minimize the cost function of joint torque, joint limit, and joint speed limit. The COM height trajectory is optimized by finding the derivative of the cost function with respect to the COM height offline. Then the proposed walking pattern generator builds the COM height trajectory database of different walking steps for online connection of a walking pattern. The walking pattern generator is verified by experiments and simulations of different step cycles with our humanoid robot, NINO, and it can clearly reduce the required joint torque of the robot while walking. In addition, compared with the fixed COM height trajectory, the energy consumption is reduced by 14% from the experimental results. Thus, the method succeeds in generating a more energy-saving walking pattern. The control framework based on quadratic programming (QP) method is designed to keep the balance of the robot during task period by combing several controllers, including proportional-derivative (PD) control to regulate the task goal and the linear quadratic state incremental (LQSI) walking pattern generator to regulate the linear momentum rate change of the robot. Besides, according to the centroidal moment pivot (CMP) to calculate the compensatory horizontal angular momentum rate change, the walking stability is improved with unexpected disturbance. Finally, an intelligent fault detection, diagnosis and health evaluation system for real humanoid robots is proposed. The system uses principal component analysis based statistical process control with Nelson rules for online fault detection. Several suitable Nelson rules are chosen for sensitive detection. When a variation is detected, the system performs a diagnostic operation to acquire features of the time domain and the frequency domain from the motor encoder and motor torque sensor for fault diagnosis with a multi-class support vector machine. Additionally, a fuzzy logic based robot health index generator is proposed for evaluating the health of the robot, and the generator is an original design to reflect the health status of the robot. Finally, several aging-related faults are implemented on our humanoid robot, NINO, and the proposed system is validated effectively by the experimental results.