基於雷射測距儀之車輛定位及地圖建構系統設計

Abstract

本論文所論述的主要貢獻是利用各種感測器的搭配組合與應用，以雷射測距儀為主要感知元件，當車輛在行駛時，對於其周圍環境進行偵測以及障礙物的地圖建構，並且在未知環境中進行定位，將其應用於智慧車(Taiwan iTS-1)的運作上。 實驗方法是在車上配備各種相關感測器，例如雷射測距儀、全球定位系統、微控制器、車輛方向盤轉角感測器以及車速感測器，利用上述之感測器互相搭配運作，以達到全自動的車輛定位、環境辨識以及地圖建構之功能。 系統運作之初，利用全球定位系統給予車輛初始絕對座標值以及初始方向，之後即利用車速感測器以及車輛方向盤轉角感測器所提供的車輛運動資訊，進行估測下一個時刻的車輛狀態，以獲得其運動到達之位置及方位。在此車輛行進之同時，利用雷射測距儀對環境進行偵測，主動將未知之環境障礙物的分佈情形測得並標示於內建之環境地圖上，以達成自動環境偵測之功能。 全球定位系統只作為位置參考訊號的情形下，本系統利用車輛運動模型(vehicle motion model)的演算法進行方位推估(dead reckoning)運算。並同時在車輛行進當中，以觀測模型(observation model)對於周圍環境進行偵測以及將環境障礙物紀錄於內建地圖上。實驗結果即能夠表示此方法的可行性效益。

The main purpose and contribution of this thesis are to combine and apply several sensors when the vehicle is moving. The system can detect the surrounding environment actively and build the maps of obstacles based on a laser range finder, and the localization for the intelligent vehicle, Taiwan iTS-1, in unknown environments. Various kinds of sensors are located in the vehicle, such as the laser range finder, the global positioning system (GPS), MicroAutoBox, the vehicle steering wheel sensor, and the vehicle velocity sensor to achieve the fully automatic vehicle localization, environment distinguish and build the maps. In the first stage of this system, the initial coordinates and initial direction of the vehicle based on the global coordinates system are given in the environmental measurement by GPS. The next stage uses the information of the vehicle from the vehicle velocity sensor and the steering wheel sensor. The vehicle states can be estimated to obtain the position and direction of the vehicle after movements. A laser range finder is used to detect the environmental status actively as a vehicle moving, and the unknown obstacles can be easily obtained in order to build the maps automatically. When the GPS signal is as the position reference signal, the vehicle motion model is used to compute states of the vehicle at the period based on dead reckoning algorithm without GPS measurements. And the proposed system can detect the surrounding environment and the build map when the vehicle moving based on the observation model. The experimental results show the effectiveness of this approach.