utexas-bwi / eband_local_planner Goto Github PK

Why my eband find the path like this?? Anyone had the same issue like me and solved it???

I am really need this package for my robot. Unfortunately I plan to upgrade my robot to ROS Melodic Morenia soon. Would you kindly release the compatible version?

What parameter should we use to enable backward motion?

probably an artifact of the band hack

All stacks apart from control_toolbox have catkin equivalents available. Proceed with converting groovy_devel to catkin as control toolbox is only used in holonomic drive control.

ref: utexas-bwi/bwi#13

The flowing code(return goal_reached_;) dosen't work correctly.

bool EBandPlannerROS::isGoalReached(){
  ....
  return goal_reached_;
  ....
}

Why did you comment out next lines?
from here
// nav_msgs::Odometry base_odom;
//...
to here
//return is_goal_reached;

bool EBandPlannerROS::isGoalReached()
{
  // check if plugin initialized
  if(!initialized_)
  {
    ROS_ERROR("This planner has not been initialized, please call initialize() before using this planner");
    return false;
  }

  return goal_reached_;

  // // copy odometry information to local variable
  // nav_msgs::Odometry base_odom;
  // {
  // 	// make sure we do not read new date from topic right at the moment
  // 	boost::mutex::scoped_lock lock(odom_mutex_);
  // 	base_odom = base_odom_;
  // }

  // geometry_msgs::PoseStamped global_pose;
  // costmap_ros_->getRobotPose(global_pose);

  // // analogous to dwa_planner the actual check uses the routine implemented in trajectory_planner (trajectory rollout)
  // bool is_goal_reached = base_local_planner::isGoalReached(
  //     *tf_, global_plan_, *(costmap_ros_->getCostmap()),
  //     costmap_ros_->getGlobalFrameID(), global_pose, base_odom,
  // 		rot_stopped_vel_, trans_stopped_vel_, xy_goal_tolerance_,
  //     yaw_goal_tolerance_
  // );

  // return is_goal_reached;

}

I was trying to build the fork ros-autom/eband_local_planner without doing sudo apt-get install ros-indigo-eband-local-planner. I have an error about the missing file eband_local_planner/EBandPlannerConfig.h.

But if I install the the release version of the original utexas-bwi/eband_local_planner first by $ sudo apt-get install ros-indigo-eband-local-planner first, I can build the ros-autom fork.

I think it might be a bug? But I am not sure about your implementation structure though :)

I really like the ros-autom fork because it has the dynamic reconfigure function.

Thank you!

The hysteresis loop in the current release is broken, but was fixed in 36cd9ee

We need to do a minor/patch release for this package.

1. The holonomic drive trajectory controller makes use of a PID from control_toolbox, which has not been catkinized yet (ros-controls/ros_control#39). To proceed with catkinization and using the hydro-devel branch for navigation, the holonomic part of the trajectory controller has been broken. Once control_toolbox gets catkinized, fix holonomic drive control
2. Additionally, replace the custom PID in the differential controller with the one from control_toolbox

It would be great if the eband_local_planner could be released into Lunar. I've just tested the package on Lunar, and it compiles without problems. So this should just be a matter of running bloom, no changes to the source code required.

I'm trying to use a planner frequency of 1.0 to speed-up local re-planning around many obstacles, and this message displays excessively:

[ INFO] [1427384535.102261903, 99.359000000]: Global plan set to elastic band for optimization

Change it to debug.

Hello guys

I found this tool really interesting, i have a ROS node which is giving x,y obstacle information , in short can any one please explain me how can i give the obstacle information from my ros node to this local planner ??
so as to get the obstacle free path from this planner !!

Thanks

Hi,
The planner often makes move_base abort since the robot footprint is in an obstacle at some point in the global plan. I understand that work needs to be done in convertPlanToBand to repair the band. But in the meantime, would it be best if we break out of the loop as long as there is some part of the band that can be achieved by the robot? This also allows the robot to wait for dynamic obstacles to pass and then continue, rather than aborting.

For example, rather than returning false on distance<=0.0, we could use:
if(band.size()) break;
else return false;

Hi,
Is the robot footprint considered when converting the plan to band? Or is this planner solely for a point robot?

I have a robot that is 1.7mx0.6m and it often hits walls.

Hello, my bubbles won't change size when I used the EBand planner. Sometimes they go through walls for some reason as well. Rarely, when they do change size, they\re smaller than the robot and the robot keeps going. I thought the algorithm would halt the robot if that happened.

I am using ROS Melodic.

I use the eband_local_planner as the base_local_planner for my four mecanum wheels base, and the eband_local_planner_params is as follows:

EBandPlannerROS:
   xy_goal_tolerance: 0.03
   yaw_goal_tolerance: 0.05
   rot_stopped_vel: 0.01
   trans_stopped_vel: 0.01
   marker_lifetime: 0.5
   eband_min_relative_overlap: 0.7
   eband_tiny_bubble_distance: 0.01
   eband_tiny_bubble_expansion: 0.01
   eband_internal_force_gain: 1.0
   eband_external_force_gain: 2.0
   num_iterations_eband_optimization: 3
   eband_equilibrium_approx_max_recursion_depth: 4
   eband_equilibrium_relative_overshoot: 0.75
   eband_significant_force_lower_bound: 0.15
   costmap_weight: 10.0


   max_vel_lin: 0.8
   max_vel_th: 0.6
   min_vel_lin: 0.1
   min_vel_th: 0.0
   min_in_place_vel_th: 0.2
   in_place_trans_vel: 0.0
   k_prop: 4.0
   k_damp: 3.5
   Ctrl_Rate: 10.0
   max_acceleration: 1.0
   virtual_mass: 0.75
   max_translational_acceleration: 1.0
   max_rotational_acceleration: 1.2
   rotation_correction_threshold: 0.5
   differential_drive: false
   bubble_velocity_multiplier: 2.0
   rotation_threshold_multiplier: 1.0
   disallow_hysteresis: false

From the above configure parameters, I have set the max_vel_lin is 0.8 m/s . But in the test, I have found that the robot's max_speed is only 0.43m/s. I have tune some parameters: k_drop, k_damp, eband_tiny_bubble_distance, etc. I have increase and decrease such parameters, but the robot's max_speed is also about 0.43 m/s. I don't know how to tune which parameters to increase the robot's speed. By the way , I use dwa_local_planner the robot's max_speed can meet the requirements.

Hello !
I didn't find anywhere a documentation about the list of available parameters for the eband_local_planner.
It would be nice if anyone can provide it.
Thanks.

during the 1st rotation step in the differential drive trajectory follower, sometimes the next bubble is too close and we start doing an infinite jitter rotation.

This bug is a bit hard to reproduce

The robot seems to be getting stuck while following the path. Giving it an alternative destination works temporarily.

This is a restriction of the approach, as it essentially combines global and local planning.
A possible solution would be to only use 1 costmap instead of 2.

I want to reduce my global plan to the local window, using this function:

eband_local_planner::transformGlobalPlan(*tf_, global_plan_, *costmap_ros_, costmap_ros_->getGlobalFrameID(), transformed_plan_, start_end_counts);

But I have this error:

undefined symbol: _ZN19eband_local_planner19transformGlobalPlanERKN7tf2_ros6BufferERKSt6vectorIN13geometry_msgs12PoseStamped_ISaIvEEESaIS8_EERN10costmap_2d12Costmap2DROSERKNSt7__cxx1112basic_stringIcSt11char_traitsIcESaIcEEERSA_RS4_IiSaIiEE
Thank you in advance for your help

Hello,
the codes here I can't understand. I guess you maybe want to use the function Pose2DToPose?

https://github.com/pal-robotics-forks/eband_local_planner/blob/0873df88df8022a90fda0bdc622eafaf548850dd/src/eband_local_planner.cpp#L1543

https://github.com/pal-robotics-forks/eband_local_planner/blob/0873df88df8022a90fda0bdc622eafaf548850dd/src/eband_local_planner.cpp#L1554

Thanks

rviz throws a bunch of errors while trying to display markers

The elastic band is only maintained within the local costmap window. Whenever the robot moves, new portions of the global plan are visible, and added to the band.

In case new portions of the path cannot be added to the band, eband_local_planner continues to operate, and pretends to successfully complete after reaching the obstacle.

This might prevent overshooting the goal, which seems to happen a lot in open areas.

both locations use their own methods to find out if the robot is within goal tolerance, and often trajectory follower says it's done, whereas navigation doesnt seem to think so.

Consequently, completion of the action is never sent.

Good evening people and thanks for that stable planner.
everything is working fine except the dynamically avoidance of an obstacle

what i do:
Scenario 1)
*Using DWA for global planner and Eband for local planner
Scenario 2)
*Using Eband for both global and local planner

i give a goal to the robot via rviz.Robot start moving smoothly.
i add an obstacle via gazebo,on the global planner green line.
i get warnings:

Calculation of Distance between bubble and nearest obstacle failed. Frame 24 of 203 in collision. Plan invalid
[ WARN] [1500970819.269624973, 141.404900000]: Conversion from plan to elastic band failed. Plan probably not collision free. Plan not set for optimization
[ WARN] [1500970819.670180554, 141.805300000]: Optimization failed - Band invalid - No controls availlable

robot is moving forward like there is no obstacle

any ideas ?
thanks in advance

Should be done with the next release. i.e. release > 0.2.1

As many have probably realized, in order for the robot deal with dynamic obstacles with this planner, it's necessary to duplicate the costmap obstacle layers on both the local and global maps. This becomes a big resource hog when using computationally intensive layers like 3d voxel layers. It would be great if this planner took the global plan and acted on it solely based on its own local cosmap layers, and it didn't require all layers to be on both global and local planners.

Hi,

I'm wondering if you have any plans for official ROS Kinetic support (ie, binaries). I was able to compile this code from source with absolutely 0 problems under Ubuntu 16.04 + ROS Kinetic.

Thanks.