植基於類神經網路之車型機器人路徑規劃

Abstract

近年來機器人的研究逐漸受到重視，而隨著機器人科技的進步，機器人的發展趨勢從單一機器人獨立完成任務演化成多機器人團隊分工合作完成複雜的任務。而在多機器人處理複雜工作時必須考慮到路徑規劃的問題，但傳統的路徑判斷中大多在已知固定環境作精準的路徑判斷。

本研究旨在利用倒傳遞類神經網路進行路徑規劃事前的學習工作，以機器人自走車配置的超音波感測器來讀取環境中的距離和方向。當感測器取得環境的幾何特徵，即將相關資訊匯入倒傳遞類神經網路進行環境的辨識。本研究共採用7種基本環境類型數據，供自走車機器人作路口判斷。實驗結果證實判斷準確。

Recently, the robot research has been gradually received attention. Along with the trends of robot technology development, more research focus on multi-robot team work than single robot independent work. And path planning of moving multi-robot is complicated for finishing team work. It is quite different from the traditional path planning in the static environment.

The main purpose of this research was to develop a path planning algorithm for a robot moving in an unknown environment. Some ultrasonic sensors were used to pick up environmental data such as distance and direction then input these data to a back-propagation neural network to carry out the path planning. Seven environmental situations were used in this research. The experimental results showed that the purposed algorithm can work precisely for a robot to make decision when encounters complicated crossroad.