A Review of Motion Planning for Urban Autonomous Driving

Abstract

The rapid development of robotics technologies has aided the growth of a variety of industries. This includes the mobile industry, which relied on drivers even until now. The rising car accident rate has grabbed huge attention for autonomous driving. However, there are a number of problems with autonomous driving that still need further investigation. Motion planning would be one of them, and previous work has already been done on the topic of motion planning as well. This review has investigated multiple motion planning algorithms, which include Dijkstra's Algorithm, Astar, Rapidly-exploring Random Tree (RRT), and Artificial Potential Field (APF), that could be used for motion planning, specifically in the area of urban driving. Through identifying the characteristics of urban roads and analyzing the benefits and drawbacks of each of the algorithms, APF would be the suggested algorithm to use in motion planning for urban driving. Although previous work already proposed algorithms that could be used for motion planning, only a few of them were researched under the circumstances of urban driving. Most of the findings drew their conclusions without consideration of the uncertainty of real driving. Therefore, only few of them could be actually applied in the real-life experiment since human drivers still exist. After compared four different algorithms and their advantages and disadvantage, further research is needed for motion planning in real-world experiments, such as urban driving, rather than simulation on computers.