即時定位建圖及導航技術應用於設施內智慧載具

Abstract

本研究致力於開發雷射探測環境的智慧載具系統，智慧載具必須有在移動環境中自我定位的能力，因此搭載了各種感測器來進行環境的探測。在機器人的領域中，即時定位建圖是一種探測未知環境的技術，定位是藉由里程計的資訊先對機器人位置做預測，再經由擴展卡爾曼濾波器，利用慣性量測單元計算出機器人的移動慣性，修正因地面不平坦所造成的里程計誤差，最後由粒子濾波器從距離感測器對環境的量測結果，估測出機器人的位置。建圖時是利用雷射測距儀來計算障礙物方位及距離，並在佔有網格地圖上根據likelihood field model來建立障礙物分佈的可能性。藉由佔有網格地圖提供的障礙物資訊，系統便能透過A*路徑規劃演算法，找出定位點在環境內移動至目標點的最佳路徑。車體在移動時編碼器會提供的馬達轉動圈數來計算履帶移動的距離，但常因為路面的摩擦力不均勻或履帶與地面打滑，造成移動時會產生不同程度的誤差。本研究透過線性迴歸的方式做直線移動距離的誤差修正，並能使修正後的相對誤差落在0.5 %以下。此外，本研究自行設計PID控制器來進行履帶移動速度的控制，車體移動的速度指令主要由線速度及角速度組成，速度指令的最大值為0.15 m/s。根據實驗的測試，速度指令為0.1 m/s時，左右履帶移動速度的最大超越量皆為0 %且安定時間皆為5秒，左履帶移動速度的穩態誤差為0.00188 m/s，而右履帶移動速度的穩態誤差為0.00216 m/s。

This study is dedicated to develop an intelligent vehicle system using laser detection to the environment. Intelligent vehicle should have the ability of self-localization in moving environment. Therefore, it is equipped with a number of sensors to detect the environment. In the field of robotics, SLAM (Simultaneous localization and mapping) is a technology for exploring the unknown environment. Odometer is used to predict the position of robot, then by the extended Kalman filter the robot's moving inertia which is measured by the inertial measurement unit to correct the error of the odometer which is caused by uneven surface of the ground. Finally, the particle filter predicts robot’s position by using the distance measurements of the environment. When building a map, a laser rangefinder is used to measure the direction and the distance of the obstacle, then occupancy grid map will record the probability of the obstacle distribution according to the likelihood field model. Based on the obstacle information recorded on the occupancy grid map, the system can use A star path planning algorithm to find the best path from the localization point to the target point. During the vehicle’s moving, the encoder will provide the number of motor rotations to calculate the moving distance. However, the robot will have different degrees of error due to uneven friction on the road or slippage of the track on the ground. In this study, the moving distance error is corrected by linear regression, and the relative error rate can be kept below 0.5%. Moreover, in this study, the PID controller is designed to control the track speed. The speed control commands are mainly composed of linear speed and angular speed. The maximum track speed command is 0.15 m/s. According to the experiment, when the track speed command is 0.1 m/s, both of the left track speed’s and the right track speed’s maximum overshoot are 0 % and their settling time are identically 5 seconds. The steady-state errors of the left track speed is 0.00188 m/s, and that of the right track speed is 0.00216 m/s.