ROS OCCUPANCY GRID DATA TYPE UPDATE

ROS OCCUPANCY GRID DATA TYPE UPDATE

The update probability of the binary Bayesian filter is currently hard-coded and requires a code change to be modified.(Optional) Error detection and handling # (Optional) Performance characterization # (Optional) References/External links # (Optional) Future extensions / Unimplemented parts #

In several places we have modified the external code written in BSD3 license. The type of grid map for estimating UNKNOWN region behind obstacle point clouds Whether to height filter for ~/input/obstacle_pointcloud and ~/input/raw_pointcloud? By default, the height is set to -1~2m. The overall point cloud used to input the obstacle point cloud Also, the unobserved cells are time-decayed like the system noise of the Kalman filter (2). Using the previous occupancy grid map, update the existence probability using a binary Bayesian filter (1). In addition, If the distance between obstacle points is less than or equal to the distance margin, that interval is filled with OCCUPIED because the input may be inaccurate and obstacle points may not be determined as obstacles. Note that the scan_origin should not be base_link if this flag is true because otherwise all the cells behind the obstacle point clouds would be filled with UNKNOWN.įill in the point where the obstacle point is located with occupied. When the parameter grid_map_type is 'OccupancyGridMapProjectiveBlindSpot' and the scan_origin is a sensor frame like velodyne_top for instance, for each obstacle pointcloud, if there are no visible raw pointclouds that are located above the projected ray from the scan_origin to that obstacle pointcloud, the cells between the obstacle pointcloud and the projected point are filled with UNKNOWN. The input may be inaccurate and obstacle points may not be determined as obstacles.The obstacle point cloud is processed and may not match the raw pointcloud.

It is unlikely that a point on the ground will be immediately behind an obstacle.

There are three reasons for setting a distance margin.

Initialize freespace to the farthest point of each bin.īased on the assumption that UNKNOWN is behind the obstacle, the cells that are more than a distance margin from each obstacle point are filled with UNKNOWN The ray trace is done by Bresenham's line algorithm. The ray trace is performed in three steps for each cell. The following figure shows each of the bins from side view. The bin contains the following information for each point In addition, the x,y information in the map coordinate is also stored for ray-tracing on the map coordinate. About Surround Obstacle Checker (Japanese)Ībout Motion Velocity Smoother (Japanese)įirst of all, input obstacle/raw pointcloud are transformed into the polar coordinate centered around scan_origin and divided int circular bins per angle_increment respectively.Īt this time, each point belonging to each bin is stored as range data.

Ros occupancy grid data type