dynamic_obstacle_avoidance.md

Demo: Dynamic Obstacle Avoidance

---

Preparation

• Install: On top of the packages for Leap Motion Teleoperation, you need to install the following packages:

  • dope: object detection algorithm
  • ros_astra_camera_OLD: ROS package for Astra Orbbec
  • rain_ai_assist: generates obstacle avoidance modulation

• Copy demo_dyanmic_obstacle_avoidance.rviz to /.rviz.

---

Execution

Step (1)-(3): UR5 + ros_control + Moveit

• Follow Step (1)-(3) of the instruction in "How to control the robot by a Xbox Joystick".

• In summary, these steps are as follows (for real robot case):

roslaunch ur_modern_driver ur5_ros_control.launch robot_ip:=172.22.22.2

roslaunch rain_moveit_config moveit_planning_execution.launch

rosservice call /controller_manager/switch_controller "start_controllers:
- 'joint_group_position_controller'
stop_controllers:
- 'pos_based_pos_traj_controller'
strictness: 2"

(4) jog_arm

• You should set some parameters in "leapmotion_to_twist.py", for example, scaling factors.

roslaunch moveit_jog_arm jog_with_leapmotion.launch
rosrun moveit_jog_arm leapmotion_to_twist.py $(option = ROS)$ (TODO: Will be included in jog_with_leapmotion.launch)

• In ROS option, the node will take a leap motion input in ROS coordinate. Otherwise, it assume that an input in Unity coordinate will be used.

• That is, for this demo, just use rosrun moveit_jog_arm leapmotion_to_twist.py ROS

(5) Pose Estimation by DOPE

• Launch a Astra Camera and match its coordinate with the UR's coordinate
  • The parameters 1.45 -0.05 0.135 3.141592 0 0 should be appropriately modified according to your test environment (i.e. the relative positions of the camera and the robot).

roslaunch astra_camera astra.launch
rosrun tf2_ros static_transform_publisher 1.45 -0.05 0.135 3.141592 0 0 world camera_astra_link 

• Run DOPE and match the coordination of pose estimation result with the UR5's coordinate
  • TODO: The pose estimation results from DOPE are not matched with the UR5's cooridnate. At the moment, these are addressed in a hacky way in dope_to_obstacle.py, wihch should be refactored later.

roslaunch dope dope.launch
rosrun rain_ai_assist dope_to_obstacle.py

• Run Rviz
  • rviz_modulation.py: Publishes some Marker topics to visualise additional information.

rosrun rain_ai_assist rviz_modulation.py
rosrun rviz rviz -d ~/.rviz/demo_dyanmic_obstacle_avoidance.rviz

(6) Leap Motion with Dynamic Obstacle Avoidance

• Genenate a fake leapmotion input, which will then be modulated by the dynamic obstacle avoidancce algorithm.
  • TODO: demo_dynamic_obstacle.py should be separated to a node generating a fake leap motion input and another node that addresses dynamic obstacle avoidance.

rosrun rain_ai_assist demo_dynamic_obstacle.py