Planner and stereo vision camera

edit

In the planner algorithm I use, my robot creates a new path against sudden obstacles according to the lidar data and reaches the target via the new path. But even though the stereo vision camera sees the obstacle, it does not update on the road and continues to go over the obstacle. What might be the cause?

[EDIT]

I am using the Neo Local Planner algorithm on the robot. I use stereo vision camera and Lidar as hardware.

1 . I give the target point with "'2D Nav Goal" from the RVIZ environment to make my robot go to the red dot in picture 1.

Image 1: https://pasteboard.co/a8zsIfn7WwYi.jpg

2 . In Picture 2, my robot forms its path with the planner and starts its path in that direction.

Image 2: https://pasteboard.co/xnXHTFwfjx19.jpg

3 . You can see the obstacles that occur when my robot sees the obstacles created by the point cloud data it receives from the camera in the picture 3-4-5, respectively. However, my robot continues on its way without making any turns against these obstacles and crashes into the obstacle.

Image 3: https://pasteboard.co/nsJsfjb9mwLc.jpg

Image 4: https://pasteboard.co/ESvIFsMMwqLI.jpg

Image 5: https://pasteboard.co/gpDn0yiMcMwt.jpg

My parameters are as follows:

common_costmap_params.yaml

obstacle_range : 7.0

raytrace_range : 7.5

max_obstacle_height : 20

footprint: [[-0.375,-0.375],[-0.375,0.375],[0.375,0.375],[0.375,-0.375]]

footprint_padding: 0.01

resolution: 0.05

inflation_layer:

inflation_radius: 0.4

cost_scaling_factor: 25.0

obstacle_layer:

map_type: costmap track_unknown_space: true
observation_persistence: 0.2
observation_persistence_duration: 0.2 min_obstacle_height: 0.1
max_obstacle_heighht: 2.0
observation_sources: laser_scan_sensor point_cloud_voxel_cam1
laser_scan_sensor: {sensor_frame:laser, data_type: LaserScan, topic:scan, marking: true, clearing: true}
point_cloud_voxel_cam1: {sensor_frame: camera1_link, data_type: PointCloud2, topic: /voxel_grid1/output, marking:true, clearing: true}

rgbd_obstacle_layer_cam1:

enabled: true voxel_decay: 15.0
decay_model: 0
voxel_size: 0.05
track_unknown_space: true
max_obstacle_height: 1.75
unknown_threshold: 15.0
mark_threshold: 0
update_footprint_enabled: true
combination_method: 1
obstacle_range: 5.0 origin_z: 0.0
publish_voxel_map: true
transform_tolerance: 0.2
mapping_mode: false
map_save_duration: 60.0
observation_sources: rgbd1_mark rgbd1_clear

rgbd1_mark:

data_type: PointCloud2
topic: /voxel_grid1/output
marking: true
clearing: false
min_obstacle_height: 0.3
max_obstacle_height: 1.0 
expected_update_rate: 0.0
observation_persistence: 0.0 
inf_is_valid: false 
filter: "passthrough" 
voxel_min_points: 0
clear_after_reading: true

rgbd1_clear:

data_type: PointCloud2
topic: /voxel_grid1/output
marking: false
clearing: true
min_z: 0.3
max_z: 2.0
vertical_fov_angle: 0.8745 
horizontal_fov_angle: 1.5 
decay_acceleration: 5.0 
model_type:  0

local_costmap_params.yaml

local_costmap:

global_frame: odom
robot_base_frame: base_footprint

update_frequency: 18.0
publish_frequency: 11.0
transform_tolerance: 0.2

static_map: true
rolling_window: true

width: 15 height: 15 origin_x: 0.0
origin_y: 0.0

plugins:

- {name: obstacle_layer,  type: "costmap_2d::ObstacleLayer"}
- {name: rgbd_obstacle_layer_cam1,  type: "spatio_temporal_voxel_layer/SpatioTemporalVoxelLayer"}
- {name: inflation_layer, type: "costmap_2d::InflationLayer"}

global_costmap_params.yaml

global_costmap:

global_frame: map
robot_base_frame: base_footprint

update_frequency: 15.0
publish_frequency: 12.0
transform_tolerance: 0.5

static_map: true

width: 17

height: 17

plugins:

- {name: static_layer,    type: "costmap_2d::StaticLayer"}
- {name: obstacle_layer,  type: "costmap_2d::ObstacleLayer"}
- {name: inflation_layer, type: "costmap_2d::InflationLayer"}

neo_planner.yaml

NeoLocalPlanner:

# The x acceleration limit of the robot in meters/sec^2

acc_lim_x : 0.3

# The rotational acceleration ...