Robot_Manipulator_Control

This is a developing ROS robot manipulator project, involved in using a UR10, 'self-made' dynamixel gripper. The main idea of this package is to enable fast configure on the UR10 motion sequence within in a short period of time. This is done by just edit the motion_config.yaml file, without changing the source code.

Here, step-by-step instructions are listed here to guide user on how to setup the environment to: (1) Run it on Rviz (2) Run it on a real UR10 Robot!!!!! Have fun people!!

Table Of Contents

1. Environment Setup
2. Run Demo Script on Rviz
3. Run On UR10 hardware
4. Understand motion_config.yaml
5. Brief Code Explanation
6. TODO

---

1. Environment Setup

a) Install ROS

b) Install Moveit and ur10_moveit_config

ROS MOVEIT! is used for motion planning. UR10 moveit pkg is here.

sudo apt-get install ros-kinetic-moveit
sudo apt-get install ros-kinetic-ur10-moveit-config

c) Install current ROS Package

catkin_make
roscd ur10_rmf/scripts
chmod +x robot_manipulator_control.py
source ~/xxx/devel/setup.bash

• If Dynamixel Gripper is used: Refer to the README.md in to dynamixel_gripper pkg.

** d) If Interfacing with UR10 hardware

Install Some UR hardware dependencies

sudo apt-get install ros-kinetic-ur-msgs    # to interface with hardware
sudo apt-get install ros-kinetic-controller-manager
sudo apt-get install ros-kinetic-industrial-msgs

Install ur_modern_driver

git clone https://github.com/ros-industrial/ur_modern_driver/
git checkout kinetic-devel`             # if using kenetic
catkin_make --pkg ur_modern_driver      # here apt-get all relevent dependencies of ur_modern_driver

** e) Configure IP for UR10 hardware and PC Network Connection

# FIRST CREATE A LOCAL IP TO COMMUNICATE WITH THE HARDWARE DEPENDING ON IP AND MASK
ping 192.168.88.70 #depends on $UR10_IP

** g) If Dynamic Dynamic cartesian planning is used (optional):

• Urg_node: Optional hardware intergration with hokoyu lidar, package is Here
• Object Pose Estimation: Optional detection of target object's pose respective to lidar, package is Here

---

2. Run Demo Script on Rviz

This is to run UR on Rviz, without a hardware setup.

a) Run Movegroup and Rviz

Robot’s config file

roslaunch ur10_rmf ur10_test.launch
# or default .launch file
roslaunch ur10_moveit_config demo.launch

b) Motion Planning Python Script

To program a series of motion , edit the yaml file at config/motion_config.yaml. Once done, run the script to visualize the motion control.

rosrun ur10_rmf robot_manipulator_control.py

c) Optional Dynamic Planning on target (with hokoyu)

roslaunch urg_node urg_lidar.launch                         # pls configure hokoyu's ip
rosrun object_pose_estimation object_pose_estimation_ros

Notes

• uncomment joint_state_publisher node in ur10_test.launch will enable rviz ur10 to run without UR10 hardware
• Change planning between: STOMP, CHOMP, OMPL
• If select STOMP planner, build package from industrial moveit
• For 'ur_modern_driver' bringup, Arg servoj_time happened to be related to the network between arm and pc, increase to prevent congestion, or maybe some jerking motion

---

3. Run On UR10 hardware

Here, Hokoyu Lidar is used for pose estimation of the target object. Refer to object_pose_estimation ros pkg for reference.

a) Launch MoveGroup, Rviz, Urg_node, Ur10 Hardware BringUp

roslaunch ur10_rmf ur10_hardware.launch

Here, you will be able to see the current ur10 hardware pose on Rviz. If dynamic cartesian planning is not used, user can comment out urg_node, object_pose_estimation pkg.

b) Enable Moveit Execution on hardware

roslaunch ur10_moveit_config ur10_moveit_planning_execution.launch limited:=true

• Steps 1 and 2 combined: roslaunch ur10_rmf ur10_hardware_combined.launch

c) Run and Test Script

rosrun ur10_rmf robot_manipulator_control.py
rostopic pub /ur10/motion_group_id std_msgs/String "INPUT" #INPUT: G1, G2... 

*To know current pose and joint angle, check printout after each motion

---

4. Understand motion_config.yaml

All defination and setting of motion is configure on config/motion_config.yaml. User just need to change the config file to configure each request motion.

• 4 types of motion: cartesian, joint_goal, pose_goal, 2d_dynamic_cartesian
• 2 types of gripper motion: eef_grip_obj, eef_release_obj
• Edit enable_gripper in .yaml file to True to enable usage of gripper
• In the yaml file, the hierachy of each is: motion_group > motion > cartesian_motion.
• Use execute_motion_group_service to check printout of current 6 joints and current eef pose, this helps in configuring the motion yaml
• adding a Coefficeient, e.g. '3', '-2', on each motion_id or cartesian_motion_id is supported in the yaml file.

Working with Pose Estimation

• To have dynamic cartesian planning, use 2d_dynamic_cartesian in yaml file motion type.
• Pose input of the target_pose is via ROSTOPIC /ur10/target_pose with [x, y, theta] info
• RosMsg for pose is from a pose estimation node, e.g: lidar point cloud pose estimation.
• input 2D info of fix_laser_pose, target, tolerance
• Here is my own lidar pose estimation ros node

Working with Gripper

• If wanna run with the gripper, pls refer to the package readme.md to run the launch file: roslaunch dynamixel_gripper gripper_manager.roslaunch
• In yaml file, enable_gripper: False, and use eef_grip_obj, eef_release_obj.
• ur10 control is a higher level control of dynamixel gripper

Error Msg

• enable log_motion_error with True, this will record all unsuccessfull motion, which can be in planning or execution phase
• the file error_log.txt will be generated in the current working directory

---

5. Brief Code Explanation

ManipulatorControl Class

Class ManipulatorControl simplfied the use of typing code to control the robot manipulator. This helps user to create a series of motion just by edit the motion_config.yaml file. 3 useful functions in this class are:

• execute_all_motion_group(),
• execute_motion_group(string motion_group_id) return True/False, True: success, False: fail
• execute_motion(string motion_id) return True/False, True: success, False: fail
• execute_motion_group_service() ros service by getting request of motion group

ArmManipulation Class

This class directly interact with the ROS moveit package.

• go_to_joint_state(joint_goal, time_factor) return bool (success anot)
• go_to_pose_goal(self, pose_list, time_factor ) return bool (success anot)
• plan_cartesian_path(self, motion_list, time_factor) return obj, float (planned trajectory, success fraction, 1.0)
• execute_plan(self, plan) return bool (success anot)

Pub Sub for execute_group_service()

Use ur10.execute_motion_group_service() to start ros service, which request group_id to user. This will activate ros pub sub mentioned below:

• /ur10/motion_group_id: Group ID (string) Sub
• /ur10/manipulator_state: State of arm and gripper (custom msg) Pub
• /ur10/rm_bridge_state: same as above's state, temp solution to feed to ros bridge (float32_array) Pub

Other Pub Sub Being used:

• /gripper/state: gripper status (grip_state msg) Sub
• /gripper/command: command gripper on ros1 (Int32) Pub
• /ur10/target_pose: 2D Pose from pose estimation (Pose2D) msg Sub

---

6. TODO

• update scene obstacle creation
• stop execution feature, maybe with asyncExecute()
• cleanup transformation of dynamic cartesian planning
• robustness
• Adjust dynamic planning tolerance, chg to when hit limit, stop execution
• Handle pub of error_flag