Skip to content

多传感器融合定位

Overview

multi_sensor_localization

  1. 需要将底盘反馈轮速与IMU的角速度融合,输出base_link坐标系下的TwistWithCovarianceStamped
  2. 使用点云地图与LiDAR点云进行匹配,输出map坐标系下的PoseWithCovarianceStamped
  3. 将融合后的TwistWithCovarianceStamped与点云地图的PoseWithCovarianceStamped作为输入,通过EKF进行滤波,获得高频与平滑的定位信息
  4. 将融合后的PoseWithCovarianceStamped作为点云匹配的初始值

输入数据

topic 类型 描述
/map/pointcloud_map sensor_msgs/msg/PointCloud2 点云地图
/localization/util/downsample/pointcloud sensor_msgs/msg/PointCloud2 降采样实时点云
/sensing/vehicle_velocity_converter/twist_with_covariance geometry_msgs/msg/TwistCovarianceStamped 底盘轮速反馈,提供线速度
/sensing/imu/imu_data sensor_msgs/msg/Imu Imu数据,提供角速度

输出数据

topic 类型 描述
/localization/kinematic_state nav_msgs/msg/Odometry 融合后Odometry数据
/localization/pose_with_covariance/geometry_msgs/msg/PoseWithConvarianceStamped 融合后的pose数据

融合定位demo

localization_demo

参考资料

https://autowarefoundation.github.io/autoware-documentation/main/design/autoware-architecture/localization/