Open tiralonghipol opened 6 years ago
Hi,
It should not be hard to implement for real robot, the package is not very complicated but I think it may take a while to understand the planner including the parameters. For detail introduction you could look at example code or the doxygen online: https://sikang.github.io/motion_primitive_library/
The mpl_ros
package is just the planning part, which gives trajectories
as the final output. To execute the trajectory, you need to write your own
controller which takes into account robot's and desired pos, vel, acc, etc
and calculates the desired force and torques for the low level attitude
controller on flight controller (for example, Pixhawk or AscTec autopilot).
The one we use is from Lee's paper on geometric control:
T. Lee, M. Leoky, and N. H. McClamroch, “Geometric tracking control of a
quadrotor uav on se (3),” in 49th IEEE conference on decision and control
(CDC). IEEE, 2010.
You also need to plug in the mapping part according to the sensors that are
used in the project as the input for mpl
planner. The planner is able to
take whatever map type, the one provided in the mpl_ros
by default is
using built-in voxel_map
for collision checking.
Sikang
On Tue, Jul 31, 2018 at 9:41 AM, Paolo De Petris notifications@github.com wrote:
Hi ! I would really love to implement this package on a real quad Any suggestion? What are the 'missing pieces' ? I could not find any information about input/output parameters ...
Thank you !
— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/sikang/mpl_ros/issues/6, or mute the thread https://github.com/notifications/unsubscribe-auth/ABccl_CgwXGWjLB2VD04sGjvmvqybirlks5uMF6OgaJpZM4VoRrn .
Thanks for the fast reply! I'll try to explain my setup, hope you can help me: until now I fusing the slam_out_pose of hector (2d lidar / no odometry) into the 'vision_fusion' plugin of the px4, inside the flight controller then, for the navigation side, I'm using the ros-stack-navigation (so 2D navigation, and is what I need for now!) Since the /mavros/local_position/pose is not precise, I planned to move out the pose estimator and, as you suggested, to implement a low-level trajectory tracking controller Do you think that https://github.com/ethz-asl/mav_control_rw can be used for this?
Forgive me, I've been working with ros and mavs since short time, hope you can help me
Pol
Hi Pol,
Understand, we also used mavros
in our flight control stack. The eth
package is an option, but its high level position control using MPC might
be an overkill. You just need to convert the reference trajectory from the
planner with [p, v, a, etc.] into desired thrust and ypr, so I think the
easiest way is to apply the PID control as the position controller. I hope
this answer could help.
Sikang
On Tue, Jul 31, 2018 at 12:10 PM, Paolo De Petris notifications@github.com wrote:
Thanks for the fast reply! I'll try to explain my setup, hope you can help me: until now I fusing the slam_out_pose of hector (2d lidar / no odometry) into the 'vision_fusion' plugin of the px4, inside the flight controller then, for the navigation side, I'm using the ros-stack-navigation (so 2D navigation, and is what I need for now!) Since se /mavros/local_position/pose is not precise, I planned to move out the pose estimator and, as you suggested, to implement a low-level trajectory tracking controller Do you think that https://github.com/ethz-asl/mav_control_rw can be used for this?
Forgive me, I've been working with ros and mavs since short time, hope you can help me
Pol
— You are receiving this because you commented. Reply to this email directly, view it on GitHub https://github.com/sikang/mpl_ros/issues/6#issuecomment-409277711, or mute the thread https://github.com/notifications/unsubscribe-auth/ABcclyLLAFkVNqnRscxLI9LtfUp2qW31ks5uMIGSgaJpZM4VoRrn .
Hi ! I would really love to implement this package on a real quad Any suggestion? What are the 'missing pieces' ? I could not find any information about input/output parameters ...
Thank you !