具有高計算效率之視覺型即時定位與建圖演算法

Abstract

FastSLAM是目前解決即時定位與建圖的熱門方法。雖然FastSLAM2.0的執行速度已經比EKF-SLAM快，但是當地標越來越多的時候，FastSLAM2.0也會因為需要多次比對量測資訊與已存在粒子中之地標，造成執行速度過慢，無法達成即時處理的目標。因此，本論文提出一種新的SLAM架構，稱之為「具有高計算效率之及時定位與建圖演算法(CESLAM)」，捨棄一開始在FastSLAM2.0中利用環境資訊更新粒子位置的階段，改成跟FastSLAM1.0一樣，先用里程計資訊更新粒子，在更新完粒子資訊後，選擇跟量測資訊似然性最高的已存地標更新粒子狀態後，再更新地標位置。模擬結果顯示，我們所提出的演算法可克服多次比對而造成執行速度過慢的問題，同時也提升了定位與建圖的準確度。實驗方面，我們使用Pioneer 3-DX機器人作為移動平台，搭配Kinect感測器進行以SURF為基礎的視覺型即時定位與建圖(V-CESLAM)，實驗結果證明，本方法可以即時地讓機器人在經過大幅的移動及旋轉後，依舊能定位出自己所在的位置，並成功建立出機器人周圍的環境地圖。FastSLAM is a popular method to solve the problem of simultaneous localization and mapping. However, when the number of landmarks present in real environments increases, there are excessive comparisons of the measurement with all the existing landmarks in each particle. As a result, the execution speed would be too slow to achieve the objective of real-time navigation.

As an attempt to solve this problem, this thesis presents an enhanced architecture for FastSLAM called computationally efficient SLAM (CESLAM), where odometer information is considered for updating the robot’s pose in particles. When a measurement has a maximum likelihood with the known landmark in the particle, the particle state is updated before updating the landmark estimates. Simulation results show that the proposed algorithm can overcome the problem of the time-consuming process due to unnecessary comparisons and improve the accuracy of localization and mapping. To practically evaluate the performance of the proposed method, we use Pioneer 3-DX robot with Kinect sensor to conduct the experiment of the V-CESLAM based on SURF. Experimental results have confirmed that our method can successfully locate the robot and build the map with satisfactory accuracy after a series of movements of the robot.