- 學位論文
基於多重感測器融合方法之智慧型機器人同步環境地圖 建構及移動物體偵測
Multi-Sensor Fusion Based Concurrent Environment Mapping and Moving Object Detection for Intelligent Service Robotics
摘要
智慧型服務機器人的發展是應用於人類社群當中相當重要關鍵的研究議題。隨著多樣化和複雜化的各類型社會服務需求，多樣的環境感知能力及自主導航能力將是智慧型服務機器人必不可少的功能。本論文研究的重點是融合多個感測器的整合技術應用於智慧型服務機器人兼顧理論與實務的發展。首先，我們建構出一個具有立體視覺、雷射測距儀等感測系統的智慧型服務機器人發展平台，設計出不僅可以自主估測出環境結構地圖，且能夠同時辨識及標定出大樓室內常用的示意圖型/標誌位置，如洗手間、滅火器、逃生門等。等同於智慧型服務機器人在執行過一次室內巡邏探索任務後，將可自動建構出包含由雷射測距儀所描繪出的精準室內環境之幾何結構以及由視覺圖型指標所構成的豐富資訊地圖。為了實現這個豐富的室內資訊地圖，過程中我們將多重感測器融合技術運用於機器人自身姿態穩健及環境特徵的估測，並提出一種改進的地圖校正技術，能夠應用於提升環境幾何結構的測繪精度。 然而，針對更進一步的智慧型服務機器人室內應用實務發展技術上，我們更進一步期待一個智慧型服務機器人，不僅能夠進行自我定位與地圖創建，還要能夠在它所服務的建築物內同時檢測出移動中的物體或人。我們基於圖型化(Graph-Based)的最佳估計法則下，延伸出一種新型擴增方法，能夠並行估測出機器人本身的運動方位和移動中物體的軌跡。此外對於環境中，移動物體的偵測及追蹤等關聯問題，也建構在使用多重感測器的融合方法加以劃分與克服。
並列摘要
Intelligent service robot development is an important and critical issue for human community applications. With the diverse and complex service needs, the perception and navigation are essential subjects. This investigation focuses on the synergistic fusion of multiple sensors for an intelligent service robot development. One of the objective of this work is to have an intelligent service robot (ISR) not only can estimate the environment structure autonomously, but also detect the commonly recognized symbols/signs simultaneously such as toilet, fire extinguisher, exit doors, etc. The result is enriched map information constructed. To implement this enriched map, multi-sensor fusion techniques are tactically utilized for robust pose association and sign estimation. For mapping consistency, we have proposed an improved alignment technique to enhance the mapping precision in a single simultaneous localization and mapping (SLAM) process. The experimental results show the approaches proposed can construct enriched indoor map information accurately and conveniently. Furthermore, for ISR extended applications, we expect an ISR that not only performs self localization and mapping but also detects moving objects or people in the building it services. In this work, a new augmented approach of graph-based optimal technique has been derived for concurrent robot postures and moving object trajectory estimation. Furthermore, all the moving object detection issues (perception, estimation and association) of an indoor autonomous mobile robot are considered and successfully implemented via multi-sensor fusion approaches. The proof of concept with experiments has been successfully demonstrated and analyzed.