Note: There have been many changes made between versions 1.1.7 and 2.0.0 of the ROS driver. You can view the changes and learn the steps to follow to adapt your code in this section.
- Overview
- Usage
- Topics
- Services
- Compatibility break between v1.1.X and v2.0.X
- About Conan
- Support for multiple arms
- Generation (advanced)
This node allows communication between a ROS node and a Kinova Gen3 or Gen3 lite robot.
The source code is released under a BSD 3-Clause license.
Author: Kinova inc.
Affiliation: Kinova inc.
Maintainer: Kinova inc. [email protected]
This package has been tested under ROS Noetic (Ubuntu 20.04).
For older ROS versions, checkout on corresponding branch :
- melodic-devel for ROS Melodic and Ubuntu 18.04 support.
- kinetic-devel for ROS Kinetic and Ubuntu 16.04 support, but the branch is no longer maintained (the melodic-devel branch might work for this configuration).
The kortex_driver
node is the node responsible for the communication between the ROS network and the Kortex-compatible Kinova robots. It publishes topics that users can subscribe to. It also advertises services that users can call from the command line or from their own code to configure or control the robot arm or its sub-devices (actuators, vision module, interface module).
Arguments:
-
arm : Name of your robot arm model. See the
kortex_description/arms
folder to see the available robot models. The default value is gen3. -
dof : Number of DOFs of your robot. The default value is for Gen3 is 7 and the default value for Gen3 lite is 6. You will have to specify this value only if you have a Gen3 6DOF.
-
vision : Boolean value to indicate if your arm has a Vision Module. The default value is for Gen3 is true and the default value for Gen3 lite is false. You will have to specify this value only if you have a Gen3 6DOF without a Vision Module. This argument only affects the visual representation of the arm in RViz.
-
gripper : Name of your robot arm's tool / gripper. See the
kortex_description/grippers
folder to see the available end effector models (or to add your own). The default value is "". For Gen3, you can also put robotiq_2f_85 or robotiq_2f_140. For Gen3 lite, the default (and only) gripper is gen3_lite_2f. -
robot_name : This is the namespace in which the driver will run. It defaults to my_$(arg arm) (so "my_gen3" for arm="gen3").
-
prefix : This is an optional prefix for all joint and link names in the kortex_description. It is used to allow differentiating between different arms in the same URDF. It defaults to an empty string.
-
ip_address : The IP address of the robot you're connecting to. The default value is 192.168.1.10.
-
username : The username for the robot connection. The default value is admin.
-
password : The password for the robot connection. The default value is admin.
-
use_hard_limits : [Gen3 only] If set to true, the arm's soft speed and acceleration limits are set to the hard limits and the MoveIt configuration uses those limits for the trajectories. If false, the default soft limit values are used. The default value for the parameter is false. Be aware that setting this argument to true will set you arm's speed and acceleration limits to the maximum, so it will move way faster! Be cautious when using it for the first time as it may cause unwanted behaviour.
-
cyclic_data_publish_rate : Publish rate of the base_feedback and joint_state topics, in Hz. The default value is 40 Hz.
-
api_rpc_timeout_ms : The default X-axis position of the robot in Gazebo. The default value is 0.0.
-
api_session_inactivity_timeout_ms : The duration after which the robot will clean the client session if the client hangs up the connection brutally (should not happen with the ROS driver). The default value is 35000 ms and is not normally changed.
-
api_connection_inactivity_timeout_ms : The duration after which a connection is destroyed by the robot if no communication is detected between the client and the robot. The default value is 20000 ms and is not normally changed.
-
start_rviz : If this argument is true, RViz will be launched. The default value is true.
-
start_moveit : If this argument is true, a MoveIt! MoveGroup will be launched for the robot. The default value is true.
-
default_goal_time_tolerance : The default goal time tolerance for the
FollowJointTrajectory
action server, in seconds. This value is used if no default goal time tolerance is specified in the trajectory. The default value is 0.5 seconds. -
default_goal_tolerance : The default goal tolerance for the
FollowJointTrajectory
action server, in radians. This value is used if no default goal tolerance is specified in the trajectory for the joint positions reached at the end of the trajectory. The default value is 0.005 radians.
To launch it with default arguments, run the following command in a terminal :
roslaunch kortex_driver kortex_driver.launch
To launch it with optional arguments, specify the argument name, then ":=", then the value you want. For example, :
roslaunch kortex_driver kortex_driver.launch ip_address:=10.0.100.239 start_rviz:=false robot_name:=terminator
You can also have a look at the roslaunch documentation for more details.
If everything goes well, you will see a "The Kortex driver has been initialized correctly!" message. If you also start MoveIt!, the kortex_driver
output may be flooded in the move_group
output, so pay attention to the warning and error messages! If the node fails to start for any reason, you will get an error message followed by a "process has died" message.
You will read below about the topics and services the driver offers. To read more about how to use those tools, go to the kortex_examples documentation.
You can publish on those topics for joint or Cartesian velocity control of the robot. You can also quickly stop its motion, apply the emergency stop (which will trigger a robot Fault state) and clear the Fault state with the corresponding topics.
Note: the robot's high level commands function at a rate of 40Hz. This will be improved in a future release.
-
/your_robot_name/in/joint_velocity
Publishing joint velocities (radians/second) on this topic will move the arm until it reaches a limit, or until you send a zero-velocity command or a Stop. You can see the message description here.
From the command line, with your robot name being "my_gen3", you can publish a joint velocity to joint 0 to this topic like so:
rostopic pub /my_gen3/in/joint_velocity kortex_driver/Base_JointSpeeds "joint_speeds: - joint_identifier: 0 value: -0.57 duration: 0"
-
/your_robot_name/in/cartesian_velocity
Publishing a Cartesian velocity (meters/second for linear, rad/second for angular) on this topic will move the arm until it reaches a limit, or until you send a zero-velocity command or a Stop. You can see the message description here.
From the command line, with your robot name being "my_gen3", you can publish a twist to this topic like so:
rostopic pub /my_gen3/in/cartesian_velocity kortex_driver/TwistCommand "reference_frame: 0 twist: {linear_x: 0.0, linear_y: 0.0, linear_z: 0.05, angular_x: 0.0, angular_y: 0.0, angular_z: 0.0} duration: 0"
-
/your_robot_name/in/clear_faults
This clears the robot's fault state (if the faults are clearable).
From the command line, with your robot name being "my_gen3", you can publish to this topic like so:
rostopic pub /my_gen3/in/clear_faults std_msgs/Empty "{}"
-
/your_robot_name/in/stop
This stops the robot's motion smoothly.
From the command line, with your robot name being "my_gen3", you can publish to this topic like so:
rostopic pub /my_gen3/in/stop std_msgs/Empty "{}"
-
/your_robot_name/in/emergency_stop
This triggers a robot fault.
From the command line, with your robot name being "my_gen3", you can publish to this topic like so:
rostopic pub /my_gen3/in/emergency_stop std_msgs/Empty "{}"
The robot feedback topics are always published by the kortex_driver
. You don't have to activate them.
-
/your_robot_name/kortex_error
Every Kortex error will be published here. You can see the message description here.
-
/your_robot_name/base_feedback
The feedback from the robot is published on this topic at a rate of cyclic_data_publish_rate. You can see the message description here.
-
/your_robot_name/joint_state
The feedback from the robot is converted to a sensor_msgs/JointState and published on this topic at a rate of cyclic_data_publish_rate.
The notification topics are only published by the kortex_driver
if you activate them by first calling an activation service. Once activated, a notification topic will be activated until the node is shutdown.
Subscribing to all the notifications causes a heavy load on the robot CPU. That is why the notification topics were designed in such a way. The users also typically only use one or two notifications, if at all.
For example, if a user wants to subscribe to the /my_robot_name/network_topic (the message type is NetworkNotification), he will have to:
- Call the /my_robot_name/base/activate_publishing_of_network_notification service to enable the publishing of the topic
- Subscribe to the /my_robot_name/network_topic topic
- Process the notifications when he receives them in his code
Most of the services supported by this node are generated from the C++ Kortex API. You can find the documentation here.
The .srv files are generated in different sub-folders depending on the sub-module they affect. For example, all the RPC calls used to configure the vision module are generated in srv/generated/vision_config
and all the RPC calls common to all devices are generated in srv/generated/device_config
. Here is a list of the packages with a short explanation of the services they have to offer:
-
actuator_config : This package contains the functions used to configure a single actuator. Note: To choose the actuator you want to configure, you have to call the /my_robot_name/actuator_config/set_device_id service and specify the device identifier of the actuator you want to configure. You get the device identifiers of actuators when you launch the node, when you parse the output of the ReadAllDevices service or in the Kinova Kortex Web App.
-
base : This package contains :
- Services to read and update the configuration of the robot
- Services to send high level commands to the robot
- Services to read and update the Product Configuration
- Services to activate the publishing of notifications
- Services to read and update the user-related information Note: The base high level commands are treated every 25 ms inside the robot. High level control cannot be achieved at a rate faster than 40 Hz for now.
-
control_config : This package contains the functions used to configure the control-related features on the robot. This includes :
- Reading and setting the cartesian reference frame
- Reading and setting the gravity vector
- Reading and setting the payload information
- Reading and setting the tool configuration
-
device_config : This package contains the functions used to configure a generic Kortex device. This includes :
- Reading and setting safety configurations
- Reading general information on the specified device (software versions, serial numbers, MAC address, IPv4 settings, etc.) Note: To choose the device you want to configure, you have to call the /my_robot_name/device_config/set_device_id service and specify the device identifier of the device you want to configure. You get the device identifiers when you launch the node, when you parse the output of the ReadAllDevices service or in the WebApp.
-
device_manager : This package contains ReadAllDevices service, which is used to get the list of connected device and various informations on each device.
-
interconnect_config : This package contains the functions used to configure the interface module on the robot. Note: You don't have to call the
SetDeviceID
service before calling the interconnect_config services, because thekortex_driver
node goes through the list of connected devices and automatically sets the correct device ID for the interconnect_config services. -
vision_config : This package contains the functions used to configure the vision module on the robot. Note: You don't have to call the SetDeviceID service before calling the vision_config services, because the kortex_driver node goes through the list of connected devices and automatically sets the correct device ID for the vision_config services.
Many things have been changed in the ros_kortex repository between versions 1.1.7 and 2.0.0 and you will have to modify your code if you don't want it to break.
- The
kortex_actuator_driver
,kortex_vision_config_driver
andkortex_device_manager
packages were removed and only thekortex_driver
package remains (one driver to rule them all). - Since we only have one driver and the ROS message generation does not deal with namespaces, the messages and services that are duplicated are now named differently. For example, the Feedback message exists within the
BaseCyclic
,ActuatorCyclic
,InterconnectCyclic
andGripperCyclic
Protocol Buffers .proto files. In ROS, this is now translated as a "PackageName_" prefix before the message name. So, for the Feedback message, the BaseCyclic_Feedback, ActuatorCyclic_Feedback, InterconnectCyclic_Feedback and GripperCyclic_Feedback ROS messages have been automatically generated. You may encounter build errors (in C++) or runtime errors (in Python) because of this change. You can just go in themsg/generated
folder and look for the problematic message to find its new name to change the occurrences in your code. - The services are now all lowercase_with_underscores instead of UpperCase.
- The services are now advertised in /my_robot_name/my_package_name/desired_service (see the Services section to learn about the packages). You can also visualize it if you start the node and type
rosservice list
in a terminal. - The topics are now all lowercase_with_underscores instead of UpperCase.
- The /my_robot_name/base_feedback/joint_state topic is now advertised as /my_robot_name/joint_state.
- The kortex_driver launch file is now located in the
kortex_driver
package instead of thekortex_bringup
package, which was deleted. Some arguments were added to the file.
From release 2.2.0 onwards, the Kortex API is automatically downloaded from our Artifactory Conan server. The steps to install and setup Conan have been added to the root readme file. Conan downloads the binaries and header files in the Conan cache, by default situated in the ~/.conan/
directory.
Note that we only provide packages for the Release and Debug configurations. CMake will pick the most appropriate one given the CMAKE_BUILD_TYPE
of your catkin workspace.
If you want to learn more about Conan, you can read about it on their website.
If you still want to download the ZIP files for the API, you can find the link in the Kortex repository.
You will have to extract the API in the kortex_api
folder as such:
kortex_api/
┬
├ include/
└ lib/
You will then have to build the catkin workspace and pass it the option to disable Conan so it links with your local API:
catkin_make -DUSE_CONAN=OFF
The kortex_driver launch file is primarily used to define one-armed robots. Having more than one arm requires prefixing the joints and links to protect against ambiguity when refering those. To this end, a kortex_dual_driver launch file has been made to demonstrate running a two-armed robot. The same pattern could be repeated to describe a robot with any number of arms.
Note: This launch file is intended as a starting point showing how to launch a robot with two arms. When using this launch file for a specific robot, it should be modified to fit the robot's description.
The launch file uses the same parameters as the one-armed version, except that most parameters are to be defined individually for each arm (with the left_ and right_ prefixes).
The prefixes used in the joints and links names can be changed using the following parameters:
left_prefix
right_prefix
The following parameters are to be specified for each robot (using left_
and right_
prefixes):
arm
dof
vision
gripper
ip_address
username
password
cyclic_data_publish_rate
api_rpc_timeout_ms
api_session_inactivity_timeout_ms
api_connection_inactivity_timeout_ms
default_goal_time_tolerance
default_goal_tolerance
The folowing parameters are NOT to be prefixed:
robot_name
start_rviz
start_moveit
cyclic_data_publish_rate
These parameters are common to both arms.
Example use : roslaunch kortex_driver kortex_dual_driver.launch robot_name:=terminator left_ip_address:=192.168.1.11 left_arm:=gen3 left_gripper:=robotiq_2f_85 right_ip_address:=192.168.1.12 right_arm:=gen3 right_gripper:=robotiq_2f_85 start_moveit:=false
Some source code as well as most of the .MSG and .SRV files in this package are automatically generated, but the generated files are given on GitHub so that users don't have to generate them. However, if you have a special version of the Kortex API and want to generate those files yourself, it is possible. You will first need to follow the instructions to install Protocol Buffers.
The generation process is based on a custom protoc
plugin. Basically, most of the generation process is in the scripts/ros_kortex_generator.py. Before launching the generation ensure that you have the Python JINJA2 module installed.
This scripts works with protobuf-compiler
(protoc
) version 3.5.1 (see 3.5.X protobuf github). Script was not tested for other versions.
To launch the generation of this package:
- Open a terminal window.
- Browse the /scripts directory of this package
- Ensure that the generate_protobuf_wrapper_files.sh file can be executed. If not then run:
chmod +x generate_protobuf_wrapper_files.sh
- Run the command:
./generate_protobuf_wrapper_files.sh
- The result of the generation should be in the following folders:
/include/kortex_driver/generated
/msg/generated
/src/generated
/srv/generated
The protos directory contains the Protobuf files from which the MSG, SRV and source files are generated. The content of this folder should not be modified.
The templates directory contains all the JINJA2 files needed by the protoc
generator.