Move_base only send null velocities (robot it's stuck)
Hi, I have a problem using the navigation stack. Everything is set up but I noticed that move_base only send velocities that have all zeroes as values for some time and then abort the plan because it says the robot appears stuck. This is the error I get:
[ WARN] [1642850369.065890614, 1625.003000000]: Clearing both costmaps to unstuck robot (1.84m).
[ WARN] [1642850369.468946798, 1625.153000000]: Rotate recovery behavior started.
[ERROR] [1642850369.469160023, 1625.153000000]: Rotate recovery can't rotate in place because there is a potential collision. Cost: -1.00
[ERROR] [1642850369.978050469, 1625.302000000]: Aborting because a valid control could not be found. Even after executing all recovery behaviors
The problem appear with every goal I send, the only movement I get is when the robot execute the escape route and apply a negative velocity.
These are my parameters files:
Base local planner:
TrajectoryPlannerROS:
max_vel_x: 1.0
min_vel_x: 0.5
max_vel_theta: 1.0
min_in_place_vel_theta: 0.4
acc_lim_theta: 2
acc_lim_x: 1
escape_vel: -0.305
holonomic_robot: false
xy_goal_tolerance: 0.20
yaw_goal_tolerance: 0.20
sim_time: 3.0
sim_granularity: 0.07
vx_samples: 3
vtheta_samples: 3
vtheta_samples: 20
goal_distance_bias: 0.3
path_distance_bias: 0.5
occdist_scale: 0.01
heading_lookahead: 0.325
dwa: true
oscillation_reset_dist: 0.05
meter_scoring: true
Costmap common params
obstacle_range: 2.5
raytrace_range: 5.0
footprint: [[0.21, 0.22], [0.21, -0.22], [-0.21, -0.22], [-0.21, 0.22]]
#robot_radius: ir_of_robot
inflation_radius: 0.55
transform_tolerance: 0.5
controller_patience: 2.0
NavfnROS:
allow_unknown: true
recovery_behaviors: [
{name: conservative_clear, type: clear_costmap_recovery/ClearCostmapRecovery},
{name: aggressive_clear, type: clear_costmap_recovery/ClearCostmapRecovery}
]
conservative_clear:
reset_distance: 3.00
aggressive_clear:
reset_distance: 1.84
Global costmap
global_costmap:
global_frame: map
robot_base_frame: base_footprint
update_frequency: 0.5
publish_frequency: 0.5
static_map: true
Local costmap params
local_costmap:
global_frame: map
robot_base_frame: base_footprint
update_frequency: 2.0
publish_frequency: 2.0
static_map: false
rolling_window: true
width: 3.0
height: 3.0
resolution: 0.05
transform_tolerance: 0.5
observation_sources: point_cloud_sensor
point_cloud_sensor: {
sensor_frame: zed2_camera_center,
data_type: PointCloud2,
topic: pointCloud_topic,
expected_update_rate: 0.5,
marking: true,
clearing: true,
min_obstacle_height: -99999.0,
max_obstacle_height: 50,
}
I am using the simulated environment provided by leo https://www.leorover.tech/integrations/erc-gazebo Rtabmap is used to get the an occupancyGrid map from the camera images and it looks quite right by looking at it from rviz.
Looking at the maps the robot doesn't look stuck, in the following image the global costmap is represented as circles around the obstacles, while the local costmap is the cone in front of the robot.
The image shows that the robot is supposed to be inside the obstacle. Therefore, we believe that it is stuck.
I think
point_cloud_sensoris the relevant sensor. Why don't you commentobservation_sources: point_cloud_sensorfor once and try to move it?Thank you for answering @miura! I commented that part and now the robot moves, it doesn't move perfectly but it's a start.
I think that my error is to pass the entire point cloud generated by the stereo camera to the local costmap but in this way it could consider also the ground as obstacle.
A package that could help me is obstacle_detection from rtabmap that generate obstacles from the point cloud, I'll try to use that package and use an approach similar to this page rtabmap/tutorial/StereoOutdoorNavigation.
I'm glad to hear that you've made progress.