機器人實驗平台的建構與控制方法的實驗比較

Abstract

本研究建立一個兩軸平面的機械手臂，利用這個實驗平台去估測模型參數、設計控制器以及測試控制器的強健性。推導出包含致動器的機械手臂動力學模型，並以線性最小平方參數估測法求得機械手臂動力學模型中的參數。接著設計出三種控制器：計算力矩的控制器、具有背隙和低速補償的適應性控制器及不需要速度估測的控制器。第一種設計是一個已經被廣泛運用的控制器，可以作為和其他控制器比較的基準。第二種設計能夠補償在本研究中特別明顯的背隙(backlash)現象，並且也能夠降低死區(dead-zone)在低速時所造成的影響。第三種設計是一個新的方法，它只需要角度的量測，就能夠完成控制器的設計。避免使用差分的方式去求得角速度。差分得到的角速度有大幅度震盪，會造成控制器增益調太高時使得機械手臂劇烈震動。實驗的結果顯示第二種設計消除了背隙現象並且縮小追蹤誤差。第三種設計也能得到不錯的追蹤誤差。這三種控制器的強健性也在實驗中被證實。

This thesis sets up a two-joint planar robotic platform and experimentally evaluates the performance and robustness of various kinds of controllers on this platform. The dynamic model of the robotic platform consists of the rigid body motion of the mechanical links as well as the dynamics of the actuators. The linear least square method is applied to find out the unknown parameters of the model. Then we design three kinds of the controllers: the computed-torque controller, the adaptive controller with backlash and low speed compensation, and the controller without velocity estimation. The first controller has been used extensively. It serves as the baseline controller in this thesis. The second controller can compensate the backlash phenomenon which diminishes the performance of the control system. It also can reduce the influence caused by the dead-zone of the actuators at the low speed region. The aforementioned two controllers use backward difference method to estimate angular velocities, resulting in large variations in the estimated values. As a consequence, the robotic arms start vibrating in the case of high controller gains. To overcome this problem, the third controller which needs no velocity estimation is proposed. Experiments are conducted for these three controllers on the robotic platform and the results show that the second controller eliminates the backlash phenomenon and decreases the tracking errors. The third controller also can get good tracking errors. These three kinds of controllers are all robust and have been verified in the experiments.