Motion Planning of a Mobile Robotic Manipulator for Grasping Tasks in Complex Environments
           

- 指導教授 黃漢邦 博士 研究生 王唯任

- Advisor :Dr.Han-Pang Huang Student :Wei-Jen Wang

Lab. of Robotics., Department of Mechanical Engineering National Taiwan University Taiwan

Abstract:

The main objective of this thesis is to develop an autonomous navigation system for a mobile manipulator for grasping tasks, which can operate in a complex environment. A high-dimensional planning algorithm for a robot to grasp or manipulate an object is developed. First, we generate a set of feasible grasps for a given object. We consider a grasp’s stability (force-closure) and local environment clearance to form a scoring function to generate feasible grasps that are the goals in grasp planning.


Next, in order solve the planning problems that have different dimensions in the configuration space and the goal space, we propose a BiRRT-GCS (Bi-directional RRT with Goal Configuration Search) algorithm, which can make the robot move toward the object and use its arm to reach the goal without using inverse kinematics in a known environment. Moreover, the algorithm is more than five times faster than the BiSpace algorithm in complex scenes. We also propose a Multi-Goal BiRRT-GCS planner to increase the probability of success of a grasping task. The simulation results and the real world experiment show that our algorithms can efficiently find a trajectory of the robot to accomplish grasping tasks.





中文摘要:


本文主要的目的在設計與建立一個針對行動式機器人進行抓握任務的自主 導航系統,使其能在複雜的人類生活環境中運作。透過發展高維度的運動規劃演 算法,使機器人可以對物品進行抓取與操作。


在這篇論文中,我們首先針對一個給定的物體去計算出一組合適的抓握位 置。我們考慮了抓握的穩定性(force-closure)以及局部環境的空曠程度,並藉由評 量函式產生一些較合適的抓握位置及方向,作為抓握任務運動規劃的終點目標。 然後我們針對組態空間(configuration space)和目標空間(goal space)維度不同 的運動規劃問題,提出了BiRRT-GCS (Bi-directional RRT with Goal Configuration Search)演算法,此演算法能使行動式機器人在已知的環境中往欲抓取的物體移動,並且使用機械手臂與機械手成功完成任務而不需要逆運動學(inverse kinematics)。BiRRT-GCS 演算法與現有的BiSpace 演算法相比,規劃的時間更短、 更有效率,在複雜的環境中速度可達五倍以上。而我們也將multi-goal 的概念加 進BiRRT-GCS 提出Multi-Goal BiRRT-GCS 演算法,這可使抓握任務的成功機會更高。最後,實驗的部份以模擬與實作的方式來比較及驗證演算法的效果。