Troubleshoooting

If you encounter computer slowing down are freeze, restart the process hppcorbaserver.

If the /happypose/detections topic don't appear anymore in the rostopic list, use the command docker compose restart in the happypose_ros/panda_happypose_bridge folder.

Between each multiview images acquisition you need to run ros2 topic pub /clear_buffer std_msg/msg/Empty "{}"

Instructions to run the demo

1. connect to the ip adress of the robot using firefox. On akasan, this address is 172.17.1.3.

2. release the brakes,

3. activate FCI,

4. in terminal, change de $ROS_MASTER_URI and $ROS_IP to ROS_MASTER_URI=http://140.93.16.70:11311 and ROS_IP=140.93.16.70

You can also do : `export ROS_MASTER_URI=http://140.93.16.70:11311 export ROS_IP=140.93.16.70`

5. in terminal 1, in this directory

   cd ~/devel/src/agimus-demos/franka/manipulation
   roslaunch robot.launch arm_id:=panda2 robot_ip:=172.17.1.3

6. to start agimus / SoT

   roslaunch demo.launch
   

7. Start hppcorbaserver

   hppcorbaserver
   

8. Start the GUI

   gepetto-gui
   

9. In the docker on Miyanoura, start the camera :

   ros2 launch realsense2_camera rs_launch.py
   

Multiview

Multiview

10. In the docker, run the happypose multiview command :

ros2 launch happypose_examples multi_view_demo.launch.py use_rviz:=true device:=cuda:0 dataset_name:=tless

11. Start the image relay in the docker :

ros2 run image_relay image_relay

12. Start gathering the images. To do so move the robot do the desired position, where all objects are visible, and run this command :

ros2 topic pub /save_image std_msgs/msg/Int8 "data: #" --once

Replace the # with the number of your shot (from 0 to a max of 9). Your last shot should be :

ros2 topic pub /save_image std_msgs/msg/Int8 "data: 9" --once

13. Publish cam view to cam topic for happypose multiview to process the images.

ros2 topic pub /publish_latched std_msgs/msg/Empty "{}"

14. Happypose will publish the poses in the topic /happypose/detections. In another terminal you can run python -i script_hpp.py and get all the poses with btf.run_pipeline().

Singleview

Singleview

10. In the docker, run the happypose singleview command :

ros2 launch happypose_examples single_view_demo.launch.py use_rviz:=true device:=cuda:0 dataset_name:=tless

11. Happypose will publish the poses in the topic /happypose/detections. In another terminal you can run python -i script_hpp.py and get all the poses with btf.run_pipeline().

Playing the generated path on the robot

To execute a path you have 2 options.

First one with graphic interface to select the path you want to play :

rosrun agimus rqt_path_execution

The second one if you want to play the latest path added to the ProblemSolver (only available in script_hpp) :

move_robot()