$DBO$ Trajectory Planning and $HAHP$ Decision-Making for Autonomous Vehicle Driving on Urban Environment

Dequan Zeng,Zhiqiang Fu,Lu Xiong,Zhuoping Yu,Hongtu Zhou,Peizhi Zhang

doi:10.1109/access.2019.2953510

Dequan Zeng, Zhiqiang Fu + Show 4 more

Open Access

https://doi.org/10.1109/access.2019.2953510

Copy DOI

Abstract

A novel driving behaviour oriented ( DBO ) trajectory planner and hierarchical analytic hierarchy process ( HAHP ) decision maker are presented for intelligent vehicle. Since driving on structural road should satisfy actuator constraints and improve comfortableness as soon as possible, which strictly obeys traffic rules other than making traffic mess, it is rather than purely pursuing the shortest route/time. By analysis traffic rules, the DBO framework is employed to produce trajectories. To make trajectory drivable, cubic B-spline and clothoid curve are modeled to keep continuous curvature, and cubic polynomial curve is to schedule velocity profile satisfying stability and comfort. To pick out the best trajectory, HAHP decision maker is developed to evaluate the candidates. The first layer selects optimal paths considering smoothness and economy, and the second layer selects best trajectory taking smoothness, comfortableness and economy in account. Moreover, DBO rapidly exploring random tree ( RRT ) replanner is embedded to ensure algorithm completeness. Finally, several typical scenarios are designed to verify the real-time and reliability of the algorithm. The results illustrate that the algorithm has highly real-time and stability evaluated by Statistical Process Control method as the probability for the peak time less than 0.1s is 100% except three obstacles avoidance scenario is 59.31% in 1000 cycles. Since the planned trajectory is smooth enough and satisfy the constraints of the actuator, the mean lateral tracking error is less than 0.2m with 0.5m peak error, and the mean speed error less than 0.5km/h with 1.5km/h peak error for all scenarios.

Highlights

Lots of attentions have been payed to trajectory planning and decision making as two key parts of the core technology for autonomous vehicle [1], [2], which is white hope for alleviating the increasingly severe traffic pressure, significantly improving road traffic safety and reducing emissions [3], [4] in recently.As a classical implementation, Dijkstra [5] algorithm, based on breadth first search strategy, searches for best trajectory by sorting length
Aimed at improving performance of intelligent vehicle driving on urban environment, driving behaviour oriented (DBO) trajectory planner and hierarchy process (HAHP) decision maker are proposed in this paper
There is a 4.8m length and 1.2m width obstacle at 20m longitudinal distance from the starting point of the current lane, an obstacle 4m length and 1.6m width at 5m longitudinal distance from the starting point of the right lane, and an obstacle 4.5m length and 1.6m width at 10m longitudinal distance from the starting point of the left lane. Both planned and real pose are no intersection with obstructions and solid lane lines depicted in Fig.32(c), which means the best trajectory picked out by HAHP decision maker could successfully avoid obstacles

Summary

INTRODUCTION

Lots of attentions have been payed to trajectory planning and decision making as two key parts of the core technology for autonomous vehicle [1], [2], which is white hope for alleviating the increasingly severe traffic pressure, significantly improving road traffic safety and reducing emissions [3], [4] in recently. As the experimental vehicle with the maximum turning curvature 0.25/m (the red lines in Fig.3(b)), the result illustrated the generated path is smooth enough and continuous to vehicle track.

DBO-RRT PATH REPLANNER

EXPERIMENTAL RESULTS AND ANALYSIS

CONCLUSION