透過您的圖書館登入
IP:18.189.180.244
  • 期刊
  • OpenAccess

Quantitative Comparison between Artificial Neural Networks and Bilinear Interpolation to Predict a Real Robot's Sonar Sensor Readings

類神經法與雙線性內外插法預測機器人聲納感應值之誤差量化比較研究

摘要


本研究嘗試以類神經網路法及雙線性內外插法,來預測移動式機器人在真實環境中之聲納模型。聲納取樣值是經由一具移動式自主機器人在真實的環境中收集而得。在預測機器人於真實環境中聲納感測值之準確性比較上,本實驗之四次驗證結果顯示,類神經法都比雙線性內外插法之預測結果較好,尤其是在取樣密度較低之區域中更為明顯。本研究之結論指出,在資料模擬或資料預測之使用目的下,使用類神經網路法可以比使用傳統內外插法有更好之結果。不僅如此,使用類神經網路法在收集取樣資料之工作上也會更為便利。

並列摘要


This paper presents examples using bilinear interpolation and Artificial Neural Networks (ANNs) to approximate the underlying sonar model of a mobile robot in a real environment. Sonar samples were collected by a mobile robot from a real environment. The comparison between results of approximation by ANNs and interpolation show that ANNs has a better performance than bilinear interpolation for four different trials, especially for those low density sampling areas. This outcome indicates that the method of using ANNs to predict unknown data for simulation or prediction purposes is more useful than using interpolation not only for the accuracy but also for the convenience of collecting samples.

延伸閱讀