ISAR

ISAR - Integration and Supervisory control of Autonomous Robots - is a tool for integrating robot applications into operator systems. Through the ISAR API you can send commands to a robot to do missions and collect results from the missions.

Getting started

Steps:

• Install
• Integrate a robot
• Run the ISAR server
• Run a robot mission

Install

For local development, please fork the repository. Then, clone and install in the repository root folder:

git clone https://github.com/<path_to_parent>/isar
cd isar
pip install -e .[dev]

For zsh you might have to type ".[dev]"

Verify that you can run the tests:

pytest .

The repository contains a configuration file for installing pre-commit hooks. Currently, black and a mirror of mypy are configured hooks. Install with:

pre-commit install

Verify that pre-commit runs:

pre-commit

pre-commit will now run the installed hooks before code is commited to git. To turn pre-commit off, run:

pre-commit uninstall

Robot integration

To connect the state machine to a robot in a separate repository, it is required that the separate repository implements the robot interface. A mocked robot can be found in this repository. Install the repo, i.e:

pip install isar-robot

NB: isar-robot has not been published to PyPi for some time, and needs to be downloaded directly from git to work.

Then, ensure the ISAR_ROBOT_PACKAGE variable in settings.env is set to the name of the package you installed. isar_robot is set by default. See the section for configuration for overwriting configuration.

If you have the robot repository locally, you can simply install through

pip install -e /path/to/robot/repo/

Running ISAR with a robot simulator

A simulator based on the open source robot Turtlebot3 has been implemented for use with ISAR and may be found here. Follow the installation instructions for the simulator and install isar-turtlebot in the same manner as given in the robot integration section. Overwrite the following configuration variables:

ISAR_ROBOT_PACKAGE = isar_turtlebot
ISAR_DEFAULT_MAP = turtleworld

Run ISAR server

To run ISAR:

python main.py

Note, running the full system requires that an implementation of a robot has been installed. See this section for installing a mocked robot or a Turtlebot3 simulator.

Running a robot mission

Once the application has been started the swagger site may be accessed at

http://localhost:3000/docs

Execute the /schedule/start-mission endpoint with mission_id=1 to run a mission.

In this folder there are predefined default missions, for example the mission corresponding to mission_id=1. A new mission may be added by adding a new json-file with a mission description. Note, the mission IDs must be unique.

Running with docker-compose

ISAR may be started with an instance of the isar-robot package by

docker-compose up --build

Provided that the simulator from isar-turtlebot is running ISAR may be started with the turtlebot by

docker-compose -f docker-compose-turtlebot.yml up --build

Configuration

The system consists of many configuration variables which may alter the functionality. As an example, it is possible to change mission planners or add multiple storage handlers as described in the mission planner and storage sections.

There are two methods of specifying configuration.

1. Override the default value by setting an environment variable.

   Every configuration variable is defined in settings.py, and they may all be overwritten by specifying the variables in the environment instead. Note that the configuration variable must be prefixed with ISAR_ when specified in the environment. So for the ROBOT_PACKAGEconfiguration variable:

   export ISAR_ROBOT_PACKAGE=isar_turtlebot

   This means ISAR will connect to isar_turtlebot robot package.

2. Adding environment variables through settings.env.

   By adding environment variables with the prefix ISAR_ to the settings.env file the configuration variables will be overwritten by the values in this file.

Running tests

After following the steps in Development, you can run the tests:

pytest .

To create an interface test in your robot repository, use the function interface_test from robot_interface. The argument should be an interface object from your robot specific implementation. See isar-robot for example.

Integration tests

Integration tests can be found here and have been created with a simulator in mind. The integration tests will not run as part of pytest . or as part of the CI/CD pipeline. To run the integration tests please follow the instructions in this section for setting up the isar-turtlebot implementation with simulator and run the following command once the simulation has been launched.

pytest tests/integration

Note that these tests will run towards the actual simulation (you may monitor it through Gazebo and RVIZ) and it will take a long time.

Documentation

To build the project documentation, run the following commands:

cd docs
make docs

The documentation can now be viewed at docs/build/html/index.html.

Contributing

We welcome all kinds of contributions, including code, bug reports, issues, feature requests, and documentation. The preferred way of submitting a contribution is to either make an issue on GitHub or by forking the project on GitHub and making a pull requests.

Components

The system consists of two main components.

1. State machine
2. FastAPI

State machine

The state machine handles interaction with the robots API and monitors the execution of missions. It also enables interacting with the robot before, during and after missions.

The state machine is based on the transitions package for Python. An visualization of the state machine can be seen below:

In general the states

States.Off,
States.Initialize,
States.Initiate,
States.Stop,
States.Monitor,
States.Paused,

indicates that the state machine is already running. For running a mission the state machine need to be in the state

States.Idle

FastAPI

The FastAPI establishes an interface to the state machine for the user. As the API and state machine are separate threads, they communicate through python queues. FastAPI runs on an ASGI-server, specifically uvicorn. The FastAPI-framework is split into routers where the endpoint operations are defined.

Mission planner

The mission planner that is currently in use is a local mission planner, where missions are specified in a json file. You can create your own mission planner by implementing the mission planner interface and adding your planner to the selection here. Note that you must add your module as an option in the dictionary.

Storage

The storage modules that are used is defined by the ISAR_STORAGE configuration variable. This can be changed by overriding the configuration through an environment variable. It accepts a json encoded list and will use each element in the list to retrieve the corresponding handler. The current options are

ISAR_STORAGE = '["local", "blob", "slimm"]'

Note that the blob and slimm options require special configuration to authenticate to these endpoints.

Implement your own storage module

You can create your own storage module by implementing the storage interface and adding your storage module to the selection here. Note that you must add your module as an option in the dictionary.

Task selection

The tasks of a mission are selected based on a task selector module, defined by the TASK_SELECTOR configuration variable. The default task selector is sequential. When using the default module, tasks are executed in sequential order defined by the current input mission.

Implement you own task selector module

Custom task selector modules may be added by implementing additional versions of the task selector interface.

For every custom module, the interface function next_task() must be implemented. All interface implementations by default have access to the list of tasks in the current mission through the member self.tasks, however additional variables may be supplied by adding arguments to next_task(). To comply with the interface definition, the function should return the next task upon every call, and raise the TaskSelectorStop exception when all tasks in the current mission have been completed:

class CustomTaskSelector(TaskSelectorInterface):
    ...
    def next_task(...) -> Task:

        # Add code here
        ...

        # Raise `TaskSelectorStop` when all tasks have been completed
        ...

Optionally, the initialize() function may be extended by supplementing the parameter list or function body:

class CustomTaskSelector(TaskSelectorInterface):
    ...
    def initialize(self, tasks: List[Task], ...) -> None:
        super.initialize(tasks=tasks)

        # Add supplementary code here
        ...

A custom task selector may be made available during module selection by adding it to the series of options in the dictionary of injector modules. It can then be activated by overriding the task selector configuration variable:

# Add custom task selector module to `modules.py`

class CustomTaskSelectorModule(Module):
    @provider
    @singleton
    def provide_task_selector(self) -> TaskSelectorInterface:
        return CustomTaskSelector()

...

# Make it available to select during injector instantiation

modules: dict[str, tuple[Module, Union[str, bool]]] = {
    ...
    "task_selector": (
        {
            "sequential": SequentialTaskSelectorModule,
            "custom": CustomTaskSelectorModule
        }
        ...
    )
    ...
}

API authentication

The API has an option to include user authentication. This can be enabled by setting the environment variable

ISAR_AUTHENTICATION_ENABLED = true

By default, the local storage module is used and API authentication is disabled. If using Azure Blob Storage a set of environment variables must be available which gives access to an app registration that may use the storage account. Enabling API authentication also requires the same environment variables. The required variables are

AZURE_CLIENT_ID
AZURE_TENANT_ID
AZURE_CLIENT_SECRET

MQTT communication

ISAR is able to publish parts of its internal state to topics on an MQTT broker whenever they change. This is by default turned off but may be activated by setting the environment variable

ISAR_MQTT_ENABLED = true

The connection to the broker will be determined by the following configuration values in settings.py

ISAR_MQTT_USERNAME
ISAR_MQTT_HOST
ISAR_MQTT_PORT

The default values of these are overwritten by the environment in settings.env.

To specify broker password, add the following environment variable to a .env file in the root of the repository:

ISAR_MQTT_PASSWORD

If not specified the password will default to an empty string.