基於A*與障礙物區域分析之車輛路徑規劃與避障整合系統

An Integrated Vehicle Path Planning and Obstacle Avoidance System Using A* and Obstacle Region Analysis

黃冠智(Kuan-Chih Huang)

指導教授 ： 連豊力

國立臺灣大學/電機資訊學院/電機工程學研究所/碩士(2019年)

https://doi.org/10.6342/NTU201903836

摘要

本篇論文提出一個自動駕駛車輛系統的完整架構，主要貢獻在車輛定位以及 路徑規劃這兩個部分。首先，一個預先建立好的 3D 點雲地圖會被用來與收到的點 雲利用最近點疊代法(Iterative Closest Point)做配對，找到車輛在地圖中的位置，本 論文提出一個點雲前處理方法，利用移除多餘的點雲像是地面上的點，以及落在本 身車子上的點，這些點皆會讓演算法匹配帶來錯誤，經過移除來讓定位更精準以及 執行時間更短。對於一個給定的終點位置，利用 Hybrid A*演算法去規劃一條全局 路徑，而這條路徑會是自駕車的參考路線，在車輛的行徑過程中，蒐集車輛附近的 點雲資料，建立一個二維的佔據圖(occupancy map)，本論文提出一個障礙物區域延 伸的方法對於障礙物區域做處理，主要因為點雲只來自光達反射障礙物之表面，此 會造成低估障礙物的區域。接著，利用占據圖判斷是否有障礙物在參考路線上，若 沒有則沿著原本的參考路線行駛，若有，則依據障礙物的位置以及參考路線來產生一條新的局部路線。而根據所產生的安全路徑，可以利用 Pure Pursuit 演算法產生 控制命令給車輛，此外，此論文也提出一種根據定位結果控制車輛速度的方法，此 有助於使定位結果收斂的更好。最後，通過模擬實驗以及實際實驗，對於所提出的 系統架構以及方法進行測試與驗證。

關鍵字

自動駕駛系統 ； 車輛定位 ； 路徑規劃與避障 ； A* ； 障礙物區域分析

並列摘要

In this thesis, a system structure for autonomous vehicle path planning and navigation system is proposed. The main contributions are in vehicle localization and path planning. First of all, a prebuild 3D point cloud map is utilized to be matched with the obtained point cloud by the Iterative Closest Point (ICP) algorithm, and then find the vehicle position inside the map.
The proposed point cloud processing method removes the redundant points such as ground point and the host vehicle points to improve localization performance and also reduce the execution time of ICP algorithm. The following is to plan a global path from the current pose to a given goal point by the Hybrid A* algorithm. The planned path can be regarded as a reference path for vehicle following. When the vehicle drives, the collected point clouds are projected into a two-dimensional occupancy map, and the obstacle region analysis method by extending the obstacle region is proposed to deal with the underestimated region of the obstacle on the local cost map.
Then, the sensor determines whether obstacles are around the global reference path. If none exists, the vehicle may follow the original reference path. Otherwise, the proposed system generates a local path based on the location of the obstacle and reference path. According to the final path, the pure pursuit algorithm generates the control command to control the vehicle. Also, the thesis proposes a novel velocity control method based on the localization performance, and this can help localization converge well.
Finally, simulations and experiments are provided to verify the proposed system, the proposed control and path planning strategy.

並列關鍵字

Autonomous driving system ； vehicle localization ； path planning and obstacle avoidance ； A＊ ； algorithm ； obstacle region analysis

參考文獻

[1: World Health Organization (WHO) 2018] Word Health Organization (WHO), “Global status report on road safety 2018:Supporting a decade of action,” Geneva, Switzerland, 2018.

[2: Rajasekhar & Jaswal 2015] M. V. Rajasekhar and A. K. Jaswal, “Autonomous vehicles: The future of automobiles,” in Proceedings of 2015 IEEE International Transportation Electrification Conference(ITEC), pp. 4-21, Chennai, India, Aug. 27-29, 2015.

[3: Jo et al. 2014] K. Jo, J. Kim, D. Kim, C. Jang and M. Sunwoo, “Development of Autonomous Car- Part I: Distributed System Architecture and Development Process,” IEEE Transactions on Industrial Electronics, vol. 61, no. 12, pp. 7131-7140, Dec., 2014.

[4: Kuutti el al. 2018] S. Kuutti, S. Fallah, K. Katsaros, M. Dianati, F. Mccullough and A. Mouzakitis, “A Survey of the State-of-the-Art Localization Techniques and Their Potentials for Autonomous Vehicle Applications,” IEEE Internet of Things Journal, vol. 5, no. 2, pp. 829-846, Apr. 2018.

[5: Paden et al. 2016] B. Paden, M. Cap, S. Z. Yong, D. Yershov and E. Frazzoli, “A Survey of Motion Planning and Control Techniques for Self-Driving Urban Vehicles,” IEEE Transactions on Intelligent Vehicles, vol. 1, no. 1, pp. 33-55, Mar. 2016.