低功耗單關節驅動之小型仿生機器人運動控制與設計

Abstract

本論文研製之四足單關節小型仿生機器人，為結構簡單、低功 耗、輕重量之小型機器人，目地為克服傳統多關節機器人體積過大， 無法進入之狹小空間或複雜地形、工業管路，同時提升機器人的可 靠性。解決方案即為本論文之四足單關節小型仿生機器人。具備學 習功能，可以經由學習程式運算，修正輸出訊號，於未知環境中， 藉由障礙偵測與距離偵測，自行穿越未知地形。 實際執行過程分為機構設計、程式發展、電路設計、系統整合；機 構設計:單關節,驅動軸傾斜安裝以獲得X軸、Y軸、Z軸移動分量。程 式發展:具備機器學習功能，與納入ZMP 法則的步態控制。電路設計: 系統控制電路採主控板與驅動電路板分離獨立方式，可因應需求可 個別強化主控板或驅動電路板。系統整合:整合後之硬體規格需小於 成人手掌尺寸，且小型仿生機器人須能獨立執行任務，無法仰賴遠 端主機運算資料後再回傳，因此簡化機器學習神經元數量，使符合 要求並縮短機器人反應時間。優化步態控制，使多數時間只運行1 或2個馬達，有效降低耗電提升續航力。－－－－－－－－－－－－ 通過角度測試、距離測試、實體迷宮行走測試，驗證本論文之單關 節仿生機器人，可利用現行市售模組、搭配步態控制，有效達成機 器人微縮化與降低功耗。

A procedure is a small-sized bionic robot with four single-joint feet, simple structure, low power consumption and light weight was developed and tested. The goal of these designs is to overcome the problem that the traditional multi-joint robot is too large to enter into the narrow space or complex terrain and industrial pipeline. For enhancing the reliability, learning function has been implemented into this small-sized bionic robot with four single-joint feet. With this learning program, this bionic robot can go through any unknown environment by barrier detection, distance detection and signal modification. The actual implementation process include: mechanism design, program development, circuit design and system integration. The size of this bionic robot is smaller than an adult palm and can independently perform the tasks without remote computing and communicating. The computation speed is very fast due to simplifying the amount of neuron used by machine learning algorithm. In the optimization of gait control, only one or two motors were used. This design not only can efficiently reduce power consumption for increasing the endurance, but also effectively adjust the gait to pass the angle test, distance test, and maze test. The small-sized bionic robot with four single-joint feet developed in this thesis can easily integrate commercial module and our gait control to effectively achieve the goals of robot miniaturization and reducing power consumption.