首页|基于运动约束的激光雷达与车体坐标系旋转参数标定

基于运动约束的激光雷达与车体坐标系旋转参数标定

扫码查看
激光雷达是智能网联汽车环境感知的重要传感器,多坐标系空间标定是激光雷达精准环境感知的前提条件.针对激光雷达与车体坐标系空间同步面临传感器观测单一的问题,提出基于激光雷达与车辆的平面运动和直线运动约束2步标定方法.为构建运动约束,基于激光里程计获取激光雷达运动位姿信息,通过激光雷达运动轨迹信息和时域上多帧地面平面拟合信息进行平面行驶识别,在满足平面路况下构建平面运动约束标定,进而标定横滚角与俯仰角;基于俯仰角和横滚角对车辆轨迹进行修正,通过激光运动轨迹建立直线行驶判别模型判别车辆运动状态,在满足车辆直线行驶路况下构建直线运动约束,从而标定偏航角.最后,在智能驾驶试验车上开展了激光雷达与车辆坐标系标定的实车试验,通过实车采集的数据验证了提出的空间同步方法的可行性.试验结果表明:提出的激光雷达与车体坐标系标定方法优于基于标定物的方法,在原始数据上可以保证标定后的旋转误差降低至0.61,误差率降低约47.4%.在手动调整的扩充数据上标定后的旋转误差降至1.64,误差率降低约40.6%.相对于基于标定物的方法,其旋转误差均有降低且不需要借助特定的标定物与标定场,降低了对环境的依赖程度.同时通过消融对比试验,证明了该方法的有效性以及鲁棒性.
Calibration of Rotation Parameters for LiDAR and Vehicle Body Coordinate Systems Based on Motion Constraints
Light detection and ranging(LiDAR)sensors are crucial for the environmental perception of intelligent networked vehicles,and multi-coordinate system spatial calibration is a prerequisite for the accurate environmental perception of these sensors.Aiming to address the problems arising from single sensor observations in space synchronization between the LiDAR and vehicle body coordinate systems,this study proposed a two-step calibration method based on planar and linear motion constraints imposed on the LiDAR sensor and vehicle.To construct the motion constraint,the LiDAR motion pose information was obtained based on the laser odometry,and the plane driving recognition was conducted based on the LiDAR motion trajectory information and multi-frame ground plane fitting information in the time domain.The plane motion constraint calibration was constructed under plane road conditions;subsequently,the roll angle and pitch angle were calibrated.Based on the pitch and roll angles,the vehicle trajectory was corrected,a linear driving discrimination model was established using the laser motion trajectory to determine the vehicle motion state,and a linear motion constraint was constructed to meet the straight driving requirements of the vehicle based on the road conditions to calibrate the yaw angle.Finally,a real vehicle test was conducted using an intelligent driving test vehicle,and the feasibility of the proposed method was verified using data collected from the real vehicle.The test results showed that the proposed method was superior to the target-based method.The rotation error was reduced to 0.61 on the original data,and the error rate was reduced to 47.4%by the proposed method.The rotation error after calibration was reduced to 1.64 on manually expanded data,and the error rate was reduced to 40.6%.Compared with the method based on the calibration object,the rotation error was reduced,and no special calibration object or calibration field were needed,which reduced the model's dependence on the environmental data.Finally,the effectiveness and robustness of the proposed method were demonstrated through comparative ablation experiments.

automotive engineeringintelligent connected vehicleenvironmental perceptionli-darlaser odometerextrinsic parameter calibration

谢国涛、王帅杰、高铭、汪东升、秦洪懋、秦晓辉

展开 >

湖南大学机械与运载工程学院,湖南长沙 410082

湖南大学整车先进设计制造技术全国重点实验室,湖南长沙 410082

湖南大学无锡智能控制研究院,江苏无锡 214115

汽车工程 智能网联汽车 环境感知 激光雷达 激光里程计 外参标定

国家重点研发计划国家自然科学基金国家自然科学基金长沙市自然科学基金整车先进设计制造技术全国重点实验室开放基金

2021YFF05011025210245652102461kq220216232115013

2024

中国公路学报
中国公路学会

中国公路学报

CSTPCD北大核心
影响因子:1.607
ISSN:1001-7372
年,卷(期):2024.37(3)
  • 32