應用ZMP於輪爪機器人之穩定爬階

Abstract

本論文旨在研發輪爪機器人穩定爬階的方法，分別從硬體與控制的角度去設計輪爪機器人系統。硬體方面包含機構設計與機電系統，機構設計以前一代輪爪機器人Clawheel II為基礎進行改善，保留能變形輪爪切換的概念，並加入新式的輪爪設計，輪爪結構比前一代更強壯，並且具有緩衝功能，增加機器人階梯地形的適應力。機器人的機電系統則增加多重感測器信號回授與馬達同步控制器，讓機器人能感知自身狀態並做出調整，增強機器人系統爬階時的強韌性。控制方面則分析機器人與階梯幾何關係以及輪爪馬達的扭力需求，並推導出一理想的爬階初始位置，在此爬階位置可以讓機器人前、後輪爪扭力輸出較為均勻，並且具有最大的平衡穩定度。爬階控制方法將機器人爬階分為前進過程與上爬過程，前進過程確保機器人回到理想爬階初始位置，上爬過程基於零力矩點 (Zero Moment Point) 概念來設定機器人參數保持動態平衡，整套方法除了使用軟體模擬來快速測試各種參數，讓輪爪機器人實現穩定爬階的功能；同時進行實體測試，在前進過程實驗中，攀爬高度為15cm深度為29cm的階梯，能成功進行輪爪相位的修正，上爬一階花費時間約11秒，包含上升過程約3秒，修正輪爪相位花費時約8秒。上爬過程實驗攀爬高度為20cm深度為25cm之階梯，能夠穩定上爬，上爬過程約8秒，輪爪轉速為每秒13度。此外，此理論與假設亦可適用於其他的輪足類機器人之爬階運動，因此這一爬階控制方法是一可廣泛應用或持續延伸的通用方法。

This research aims to develop an approach that permits a claw-wheel robot to climb stairs steadily through designing hardware and control strategy. The hardware include the mechanism design and mechatronic system. In order to improve prior generations of the claw-wheel robot, the mechanism design of the Clawheel III retains the claw-wheel transformable concept of the Clawheel II but implements new claw-wheels to improve the structure of the former wheel and provide an impact cushion so as to enhance terrain adaptability of the robot. With further implementation of sensors feedback and motor synchronization controller, the robot can automatically adjust its posture, and thus improve stability when climbing stairs. Moreover, we derive an ideal initial position of stair climbing, in which the robot can have more even torque distribution between the front motors and the rear motors and have the most robust stability by analyzing geometric relations between the robot and stairs and evaluating the required motor torque. In control strategy, the stair climbing process is separated into proceeding and ascending process. In the advancing process, the robot is expected to return to the ideal initial position of stair climbing. In ascending process, the robot will maintain an equilibrium for dynamic stability based on the concept of ZMP (Zero Moment Point). Furthermore, this control strategy is set in simulation to instantaneously test various parameters, and the experiments are conducted subsequently. In the experiments of the advancing process, the robot can successfully correct the bilateral phase difference of claw-wheels while climbing up a stair with 15cm rise height and 29cm depth, and it takes 11 sec/stair in average that includes 3 sec in the ascending process and 8 sec in correcting phase difference. In the experiments of ascending process, the robot can steadily climb up a stair with 20cm height and 25cm depth in 8 sec while each wheel rotates at 13 deg/sec. In addition, this theory and hypothesis also apply to the climbing movement of other wheel-legged robots. As a result, this stair climbing control strategy is a general approach which can be widely applied or persistently extended.