Adding a New Nav2 Task Server

A nav2 task server consists of server side logic to complete different types of requests, usually called by the autonomy system or through the Behavior Tree Navigator. In this guide, we will discuss the core components needed to add a new task server to Nav2 (ex. Controller, Behavior, Smoother, Planner Servers). Namely, how to set up your new Lifecycle-Component Node for launch and state management and the communication of semantically meaningful error codes (if necessary).

While this tutorial does not cover how to add the complementary Behavior Tree Node to interact with this new Task Server, that is covered at length in Writing a New Behavior Tree Plugin so this Task Server can be invoked in the BTs in BT Navigator.

If you’ve created a new Task Server that may have general reuse for the community, consider contacting the maintainers to add it to the Nav2 project! Nav2 gets better by contributions by users like you!

Lifecycle Nodes

The Lifecycle node is the first key component of a nav2 task server. Lifecycle nodes were introduced in ROS 2 to systematically manage the bringup and shutdown of the different nodes involved in the robot’s operation. The use of Lifecycle nodes ensures that all nodes are successfully instantiated before they begin their execution and Nav2 shuts down all nodes if there is any unresponsive node.

Lifecycle nodes contain state machine transitions that enable deterministic behavior in ROS 2 servers. The Lifecycle node transitions in Nav2 are handled by the Lifecycle Manager. The Lifecycle Manager transitions the states of the Lifecycle nodes and provides greater control over the state of a system.

The primary states of a Lifecycle node are Unconfigured, Inactive, Active, and Finalized. A Lifecycle node starts in an Unconfigured state after being instantiated. The Lifecycle Manager transitions a node from Unconfigured to Inactive by implementing the Configurating transition. The Configurating transition sets up all configuration parameters and prepares any required setup such as memory allocation and the set up of the static publication and subscription topics. A node in the Inactive state is allowed to reconfigure its parameters and but cannot perform any processing. From the Inactive state, the Lifecyle Manager implements the Activating transition state to transition the node from Inactive to Active, which is the main state. A node in the Active state is allowed to perform any processing operation. In case a node crashes, the Lifecycle Manager shuts down the system to prevent any critical failures. On shutdown, the necessary cleanup operations are performed and the nodes are transitioned to the Finalized state via Deactivating, CleaningUp, and ShuttingDown transition states.

See also

For more information on Lifecycle management, see the article on Managed Nodes.

You may wish to integrate your own nodes into the Nav2 framework or add new lifecycle nodes to your system. As an example, we will add a new notional lifecycle node sensor_driver, and have it be controlled via the Nav2 Lifecycle Manager to ensure sensor feeds are available before activating navigation. You can do so by adding a sensor_driver node in your launch file and adding it to the list of nodes to be activated by the lifecycle_manager before navigation, as shown in the example below.

lifecycle_nodes = ['sensor_driver',
                   'controller_server',
                   'smoother_server',
                   'planner_server',
                   'behavior_server',
                   'bt_navigator',
                   'waypoint_follower']

...

Node(
    package='nav2_sensor_driver',
    executable='sensor_driver',
    name='sensor_driver',
    output='screen',
    parameters=[configured_params],
    remappings=remappings),

Node(
    package='nav2_lifecycle_manager',
    executable='lifecycle_manager',
    name='lifecycle_manager_navigation',
    output='screen',
    parameters=[{'autostart': autostart},
                {'node_names': lifecycle_nodes}]),

In the snippet above, the nodes to be handled by the Lifecycle Manager are set using the node_names parameter. The node_names parameter takes in an ordered list of nodes to bringup through the Lifecycle transition. As shown in the snippet, the node_names parameter takes in lifecycle_nodes which contains the list of nodes to be added to the Lifecycle Manager. The Lifecycle Manager implements bringup transitions (Configuring and Activating) to the nodes one-by-one and in order, while the nodes are processed in reverse order for shutdown transitions. Hence, the sensor_driver is listed first before the other navigation servers so that the sensor data is available before the navigation servers are activated.

Two other parameters of the Lifecycle Manager are autostart and bond_timeout. Set autostart to true if you want to set the transition nodes to the Active state on startup. Otherwise, you will need to manually trigger Lifecycle Manager to transition up the system. The bond_timeout sets the waiting time to decide when to transition down all of the nodes if a node is not responding.

Note

More information on Lifecycle Manager parameters can be found in the Configuration Guide of Lifecycle Manager

Composition

Composition is the second key component nav2 task servers that was introduced to reduce the memory and CPU resources by putting multiple nodes in a single process. In Nav2, Composition can be used to compose all Nav2 nodes in a single process instead of launching them separately. This is useful for deployment on embedded systems where developers need to optimize resource usage.

See also

More information on Composition can be found here.

In the following section, we give an example on how to add a new Nav2 server, which we notionally call the route_server, to our system.

We make use of the launch files to compose different servers into a single process. The process is established by the ComposableNodeContainer container that is populated with composition nodes via ComposableNode. This container can then be launched and used the same as any other Nav2 node.

  1. Add a new ComposableNode() instance in your launch file pointing to the component container of your choice.

    container = ComposableNodeContainer(
        name='my_container',
        namespace='',
        package='rclcpp_components',
        executable='component_container',
        composable_node_descriptions=[
            ComposableNode(
                package='nav2_route_server',
                plugin='nav2_route_server::RouteServer',
                name='nav2_route_server'),
        ],
        output='screen',
    )
    

    See also

    See example in composition demo’s composition_demo.launch.py.

  2. Add the package containing the server to your package.xml file.

    <exec_depend>nav2_route_server</exec_depend>
    

Error codes

Your nav2 task server may also wish to return a ‘error_code’ in its action response (though not required). If there are semantically meaningful and actionable types of failures for your system, this is a systemic way to communicate those failures which may be automatically aggregated into the responses of the navigation system to your application.

It is important to note that error codes from 0-9999 are reserved for internal nav2 servers with each server offset by 100 while external servers start at 10000 and end at 65535. The table below shows the current servers along with the expected error code structure.

Server Name

Reserved

RANGE

NONE=0, UNKNOWN=1

2-99

Controller Server

NONE=0, UNKNOWN=100

101-199

Planner Server (compute_path_to_pose)

NONE=0, UNKNOWN=200

201-299

Planner Server (compute_path_through_poses)

NONE=0, UNKNOWN=300

301-399

Smoother Server

NONE=0, UNKNOWN=500

501-599

Waypoint Follower Server

NONE=0, UNKNOWN=600

601-699

Behavior Server

NONE=0

701-799

Coverage Server

NONE=0, UNKNOWN=800

801-899

Last Nav2 Server

NONE=0, UNKNOWN=9900

9901-9999

First External Server

NONE=0, UNKNOWN=10000

10001-10099

Error codes are attached to the response of the action message. An example can be seen below for the route server. Note that by convention we set the error code field within the message definition to error_code.

# Error codes
# Note: The expected priority order of the errors should match the message order
uint16 NONE=0 # 0 is reserved for NONE
uint16 UNKNOWN=10000 # first error code in the sequence is reserved for UNKNOWN

# User Error codes below
int16 INVILAD_START=10001
int16 NO_VALID_ROUTE=10002

#goal definition
route_msgs/PoseStamped goal
route_msgs/PoseStamped start
string route_id
---
#result definition
nav_msgs/Route route
builtin_interfaces/Duration route_time
uint16 error_code
---

As stated in the message, the priority order of the errors should match the message order, 0 is reserved for NONE and the first error code in the sequence is reserved for UNKNOWN. Since the the route server is a external server, the errors codes start at 10000 and go up to 10099.

In order to propigate your server’s error code to the rest of the system it must be added to the nav2_params.yaml file. The error_code_id_names inside of the BT Navigator define what error codes to look for on the blackboard by the server. The lowest error code of the sequence is then returned - whereas the code enums increase the higher up in the software stack - giving higher priority to lower-level failures.

error_code_id_names:
    - compute_path_error_code_id
    - follow_path_error_code_id
    - route_error_code_id

Conclusion

In this section of the guide, we have discussed Lifecycle Nodes, Composition and Error Codes which are new and important concepts in ROS 2. We also showed how to implement Lifecycle Nodes, Composition and Error Codes to your newly created nodes/servers with Nav2. These three concepts are helpful to efficiently run your system and therefore are encouraged to be used throughout Nav2.