使用地面特性分類來實現雙足機器人於不同環境之行走

Abstract

要讓雙足機器人能穩定地行走，必須要能夠處理不同地面的情境。但是因為人形機器人的行走控制必須考慮動態平衡，這個問題十分困難。由於複雜的演算法會需要強大的計算能力或者不同的機器人結構，而這在市面上能取得的機器人來說十分困難。我們選擇較為基本的走路模型。為了適應不同地面，我們收集壓力傳感器的數據，使用SVM來偵測地面類型，並把相應的參數提供給走路模型，以達到最佳的穩定表現。

To accomplish reliable walking of biped robot, the ability to handle various terrain is very important. However, humanoid robot walking has been a challenging issue in the control field since the balance can only be achieved by considering the dynamics of the system. Instead of using sophisticated algorithms that are limited by computing power and specific control units, we choose simple walking model. To handle different ground characteristic, the class is detected by SVM with data from force sensors, and the corresponding parameters are assigned to walking model to reach optimal stability.