摘要
由于目前已有车辆自动驾驶风险预测方法无法提取车辆自动驾驶行为特征参数,导致目标风险检测能力下降,风险预测结果偏差较大。提出新的复杂智能交通环境车辆自动驾驶风险预测方法,将三维点云投影到栅格地图中,采用连通域分析完成风险目标聚类,利用最小包络面积矩阵建模,选取候选目标,将该目标与上一帧目标数据关联,完成风险目标检测。提取车辆自动驾驶行为特征参数,引入数据驱动技术,构建复杂智能交通环境车辆自动驾驶风险预测模型。实验测试结果表明,所提方法下自动驾驶车辆行驶速度与最佳速度一致,且风险预测结果符合真实情况。
The traditional risk prediction methods of vehicle automatic driving lack the extraction results of characteristic parameters of vehicle automatic driving behavior,leading to the defects of traditional methods,such as low detection ability and large deviation of prediction results.This paper reports a novel risk prediction method for vehicle autopilot in a complex intelligent transportation environment.First of all,the three-dimensional point cloud was projected into the grid map.Based on the connected domain,the risk target clustering was investigated in detail.Secondly,the minimum envelope area matrix was used to model and select candidate targets.The selected target and the target data of the previous frame were combined to complete the risk target detection.Then,the characteristic parameters of vehicle automatic driving behavior were extracted,and the data-driven technology was introduced.Eventually,the risk prediction model of vehicle automatic driving in a complex intelligent transportation environment was founded.The results show that the driving speed of autonomous vehicles predicted by this method is consistent with the optimal speed,and the risk prediction result is in accordance with the real situation.
作者
陈宁
姜冉
张轩
CHEN Ning;JIANG Ran;ZHANG Xuan(School of Mechanical and Energy Engineering,Zhejiang Science and Technology University,Hangzhou Zhejiang 310023,China;School of Electronic Engineering,Beijing University of Posts and Telecommunications,Beijing 100876,China)
出处
《计算机仿真》
北大核心
2022年第5期117-120,129,共5页
Computer Simulation
基金
国家自然科学基金(61273240)
国家重点研发计划“政府间国际科技创新合作”重点专项项目(2019YFE0126100)。
关键词
复杂智能交通环境
车辆
自动驾驶
风险预测
目标检测
Complex intelligent traffic environment
Vehicles
autonomous driving
Risk prediction
Target detection
