頁籤選單縮合
題名 | 基於神經振盪之兩足機器人的步態產生器=Walking Pattern Generator of Biped Robots Using Neural Oscillators |
---|---|
作者 | 張志鋒; 廖世謨; 彭偉鈞; | 書刊名 | 高雄應用科技大學學報 |
卷期 | 39 2010.05[民99.05] |
頁次 | 頁121-150 |
分類號 | 448.94 |
關鍵詞 | 兩足機器人; 神經振盪器; 步行模式; Biped robot; Neural oscillator; Walking pattern; |
語文 | 中文(Chinese) |
中文摘要 | 本文發展一個只由三個神經振盪器所組成之步態產生器,它可用來產生兩足機器人之步態使其穩定步行。首先,採用基因演算法以決定各振盪器和感測器反饋訊號之間的連接方式。由於適應函數對於基因演算法之求解結果具有關鍵性的影響,故提出一個系統性的方法以導出最合適之適應函數,經由實際應用可發現,該適應函數對於最佳解之快速收斂極有助益。此外,也提出一個簡單可靠的策略以求得具有穩定循環之神經振盪器,使其依指定之自然頻率產生振盪。最後,根據上述方法和策略求得一個滿足要求之步態產生器。經由動力模擬和硬體實驗,已證實該步態產生器可用來驅動機器人使其平順地由靜止站立狀態轉換為穩定步行。 |
英文摘要 | A walking pattern generator composed of three neural oscillators is developed to generate the walking patterns of biped robots such that they can walk stably. The connections between the neural oscillators and the sensory feedbacks are determined by using a generic algorithm. Since the fitness function of the generic algorithm is the key factor to the evolution results, a systematic method is thus proposed for determining the most suitable fitness function. That fitness function is found very helpful for the rapid convergence of optimal solutions. Besides, a simple and reliable algorithm is also proposed to find the neural oscillators that have limit cycles with desired natural oscillating frequency. As a result of the proposed method and algorithm, a stable walking pattern generator is finally determined. Dynamic simulations and hardware experiments of a biped robot are used to verify that the walking pattern generator can drive the robot to walk smoothly from stand still. |
本系統之摘要資訊系依該期刊論文摘要之資訊為主。