具定位避障導航與順應式雙機械手臂操作之野外用智慧型機器人

機器人研發的初衷，便在於代替人類進行重複或危險的任務，而軍方所需進行對具化學性或輻射性物質的處理，便是一個非常典型的案例。本論文針對如上述類型任務的訴求，以開發出能在野外運行與處理待測物之智慧型無人偵檢與操作平台為目標。該類型平台需具有兩大功能，分別為移動與操作，在功能上各自對應到人類的雙足與雙手。在現階段開發中，暫以色彩鮮明的物件來代表具危險性的待測物。在全自主移動找尋到待測物方面的研發，研發出的偵檢平台具有定位、避障、導航、與搜尋四大功能。定位以GPS與INS等多重感測器在Unscented Kalman Filter的架構下融合以估測出平台本體狀態；避障以Vector Field Histogram為基礎，開發出適用於野外的避障演算法；導航結合定位與避障演算法規劃平台移動速度與方位；搜尋則使用相機即時擷取影像，在HSL色彩空間下利用濾色來得出待測物方位。最後，並整合四大功能，以實驗驗證方式確認偵檢平台能在野外未建構環境中，自主導航與避障，搜尋到100公尺以外的待測物。在待測物件操作方面，以開發雙手協同穩定抓取物體為目標，建構出可安裝於偵檢平台上的雙機械手臂系統，並開發雙手臂協同控制，其分為三個層次：第一為建立手臂空間感，利用物體與雙手的空間關係，以Master-Slave的架構建立開迴路控制；第二為建立Slave端手臂順應性，以六軸力規作為手掌，將力對應至移動，力矩對應至轉動，使手掌能貼平物體表面並保持適當正向力，並與人合作抓取物體；第三為結合空間感與順應性，利用Kalman Filter的概念融合開迴路控制與順應控制，使雙手在沿一定軌跡移動物體但受外力干擾時，能即時修正軌跡，使手掌貼平物體表面並保持適當正向力。上述控制架構並經實驗驗證雙手臂能在移動物體時有效抵抗控制誤差、外力干擾、與物體重量變化等，與開迴路控制相較具有顯著的特性提升。

關鍵字

偵查機器人；定位；避障；導航；雙機械手臂；順應控制；雙手協同

並列摘要

The purpose of developing robots is to replace humans when carrying out frequently repeat or highly dangerous tasks. A very typical example of such tasks in the military is handling chemical or radioactive materials. In response to the demand for these types of tasks, this thesis aims to develop an intelligent unmanned reconnaissance and manipulation platform, which can run in rough terrain and handle objects. The platform’s main functions include moving and manipulation, which corresponds to human legs and hands. In its development stage, dangerous targets are temporarily replaced with colorful objects. In developing autonomous searching, this platform consists of four main functions: localization, obstacle avoidance, navigation and searching. Localization, which is based on unscented Kalman filter, estimates the body state by fusing multiple sensors such as GPS and INS; Obstacle avoidance modifies Vector Field Histogram for outdoor usage; Navigation combines localization and obstacle avoidance calculating moving speed; Searching captures live images, converts them to HSL color space and finds the target by color threshold. Finally, we propose a system structure that integrates four functions and validates them through outdoor experiment in rough terrain. This platform can find the target 100 meters away and avoid obstacles automatically when navigating. In object manipulation, we construct a dual robot arm system and develop the basic control to allow the arm to grasp objects stably. Dual arm manipulation is divided into three stages. First, develop a Master-Slave open loop control using the space relationship between the object and hands. Second, develop the compliance of the slave side. The palm can stick to the object’s surface and keep proper normal force by mapping force to displacement and torque to rotation. Third, combine space relationship and compliance with the concept of Kalman filter. When moving objects, the slave side, under the disturbance, can adjust its trajectory to keep proper normal force. Finally, through experiments, we conclude that the compliant dual arm manipulation can deal with control error, external force disturbance, and weight variation.

並列關鍵字

reconnaissance robot ； localization ； obstacle avoidance ； navigation ； dual robot arm ； compliant control ； dual arm manipulation

參考文獻

[1] Keepon. Available: http://en.wikipedia.org/wiki/Keepon

Google Scholar

[2] Scooba. Available: http://store.irobot.com/family/index.jsp?categoryId=2174932

Google Scholar

[3] PrimeAir. Available: http://www.amazon.com/b?node=8037720011

Google Scholar

[4] Google driverless car. Available: http://en.wikipedia.org/wiki/Google_driverless_car

Google Scholar

[5] E. Guizzo and T. Deyle, "Robotics trends for 2012," IEEE Robotics & Automation Magazine, vol. 19, pp. 119-123, 2012.

Google Scholar

國際替代計量

具定位避障導航與順應式雙機械手臂操作之野外用智慧型機器人

全文下載

主題瀏覽