增強型蒙地卡羅定位法及其在單板電腦之實現
Abstract
機器人定位是行動機器人導航的核心技術，為了讓機器人順利執行任務，首要目標即是即時獲得機器人位置。要得到機器人位置則必須考慮三大問題：全域定位、位置追蹤、以及機器人綁架。機器人定位發展已久，最廣為人知且最普遍採用的演算法為蒙地卡羅定位法(Monte Carlo Localization, MCL)，MCL是利用粒子濾波器作為主要架構的機器人定位法，透過粒子濾波器逐漸淘汰不良粒子，進而估測出機器人所在位置。儘管MCL被廣泛的採用，其無法從機器人綁架問題中恢復及運算大量粒子資訊造成計算負擔過大，成為MCL最大的缺點。為了解決此一問題，自調適蒙地卡羅演算法(Self-Adaptive Monte Carlo Localization, SAMCL)加入了相似能量區域(Similar Energy region, SER)及預先存取(Pre-caching)機制，解決了機器人綁架重新搜尋的問題，並提高即時運算速度，但MCL與SAMCL兩者仍存在共同的缺點：粒子淘汰過程中容易受超強粒子影響而陷入區域最佳解，粒子數量在位置追蹤及全域搜尋上的不平衡也會造成搜尋效果不好或追蹤過程計算負擔過大，再者，由於機器人的方向估測不易，導致即便粒子位置正確，但因為方向錯誤而被淘汰的可能性。因此，本論文提出一種改良式蒙地卡羅定位法，稱之為「具強健方向估測之蒙地卡羅定位法」(Improved Monte Carlo Localization with Robust Orientation Estimation, IMCLROE)，加入了方向估測機制來避免粒子位置正確卻被淘汰的問題，以及粒子數量平衡機制來平衡搜尋與追蹤的粒子數量，並且利用競爭選取法避免陷入區域最佳解，以提高粒子分布的多樣性。實驗證明IMCLROE有效解決了MCL及SAMCL在搜尋追蹤及綁架上的不足。
This paper proposes an improved Monte Carlo Localization algorithm with robust orientation estimation (IMCLROE) by incorporating an orientation estimate and weight calculation mechanism to determine an optimal orientation for particles and a particles size balancing mechanism to regulate the number of particles for position tracking and global localization. Based on previously established sensory information, the proposed IMCLROE can improve the computational efficiency of robot localization. Position tracking accuracy and global localization successful rate are also significantly improved while maintaining a minimal population of particles. Simulation and experimental results have confirmed the effectiveness of the proposed approach.
This paper proposes an improved Monte Carlo Localization algorithm with robust orientation estimation (IMCLROE) by incorporating an orientation estimate and weight calculation mechanism to determine an optimal orientation for particles and a particles size balancing mechanism to regulate the number of particles for position tracking and global localization. Based on previously established sensory information, the proposed IMCLROE can improve the computational efficiency of robot localization. Position tracking accuracy and global localization successful rate are also significantly improved while maintaining a minimal population of particles. Simulation and experimental results have confirmed the effectiveness of the proposed approach.
Description
Keywords
蒙地卡羅定位, 方向估測機制, 粒子濾波器, 機器人定位, 機器人追蹤, Monte Carlo Localization, Orientation estimation, Particle filter, Robot localization, position tracking