jderobot / bt-studio Goto Github PK

The GUI needs a design revision. There are placeholder buttons yet and the button hovers are badly placed.

There have been a lot of changes from version 0.1, the readme should explain the usage and the installation properly

tree_gardener is the python module that provides the tree creation functionality to client programs like the generator and the executor.

It needs an API documentation, with proper formatting.

There are two different types of dependencies in BehaviorTrees:

• Execution tools, such as tb4_sim, that should be handled via vcs.
• User defined imports in the actions, that should be installable via rosdep. The translator should edit the xml so they can be found.

Once the user created a diagram, the library provides function to serialize its structure into a json file. This should be converted to a xml tree file

Currently, the template only includes a few dependecies. Check how to implement a system so multiple libraries can be included.

Concretely:

• Laser Bump and Go
• Visual Follow Person

The order of the actions is not updated properly and this causes the behavior tree to misbehave when executed.

Here is the video of how to recreate this behavior:

And here is the incorrect bt:

<Root name="Tree Root">
  <BehaviorTree>
    <ReactiveSequence name="ReactiveSequence">
      <ReactiveFallback name="ReactiveFallback">
        <Inverter name="Inverter">
          <CheckObstacle name="CheckObstacle" amplitude="20" obs_dist="1.0" />
        </Inverter>
        <!-- The order should be the opposite -->
        <Turn name="Turn" />
        <Forward name="Forward" />
        <!--  -->
      </ReactiveFallback>
      <Forward name="Forward" speed="0.5" />
    </ReactiveSequence>
  </BehaviorTree>
    <Code>

The diagram editor works fine until the user interacts with other elements of the page, such as the file browser or the draggable separators.

Currently, there is no port representation and no mechanism to input a value in them.

It is time! The project needs a proper usage section!

Currently, It also shows the xml tree.

Currently, the app is generated by writing the XML structure and executing the different generation scripts. The goal of the next version is to develop a web frontend for writing action and tree creation. This opens the door to future integrations with a dockerized ROS 2 environment.

The user should be able to save and load different projects, that is BT diagrams.

Although not part of the project, a complete and straightforward guide to setup a simulated environment may help the users.

Ideally, the translator should have two steps:

• Parsing the actions and the tree xml into a self contained xml file.
• Generate a complete robotic app in a zip. This app is a ROS2 package.

tree_gardener creates executable trees and handles port access in the actions. Currently, It is a python package that the user has to install to run the generated package. It should be a ROS 2 package and included in the generated app.

The node is removed although it is not selected.

When trying to execute the demos the following line:

import tree_gardener

gives an error about package not found.

Nodes not conecting properly to inputs after loading, but correctly after switching project

Hi! I am a MacOS user and everything executes without error(in both terminals). However, I was trying to write&download some code to test in ROS afterward. But cannot download since I keep getting such display and some buttons are not active. What can be the reason?

To do:

• Implementation of remaining BT.cpp decorators.
• Proper code structure. Translator must be ROS agnostic.
• Usage example. Hello world.

This was changed during the templates PR

Currently, the code is structured as a ROS 2 package. As the aim of the project is to provide a web based solution, the code needs a more agnostic structure.

how does creating new actions, for example turn.py create a new node in your frontend.
What is the logic behind your node creation for action nodes.

Currently this is a placeholder. The file list API should be called and properly displayed.

In order to illustrate the capabilities of the offline execution, a detailed guide of how to use the package must be created. It must comprise exact instructions in a md file. A video should also be recorded.

Start the front end with the following code:

cd ..
 cd frontend
yarn install
yarn start

produced the following error:

00h00m00s 0/0: : ERROR: [Errno 2] No such file or directory: 'install'
00h00m00s 0/0: : ERROR: [Errno 2] No such file or directory: 'start'

Also, I think the installation process is still not very streamlined. I would recommend using a container for development. I asked bard how to add the installation process to a Docker container. Maybe this will help you:
Dockerfile:

FROM ros:humble

RUN source /opt/ros/humble/setup.bash

WORKDIR /opt/bt-studio
RUN git clone https://github.com/JdeRobot/bt-studio

RUN sudo apt update && sudo apt install python3-vcstool python3-pip python3-rosdep python3-colcon-common-extensions -y

WORKDIR /opt/bt-studio/backend
RUN pip install virtualenv
RUN virtualenv backend_env
RUN source backend_env/bin/activate
RUN pip install django djangorestframework
https://github.com/Guelakais
CMD python3 manage.py runserver

WORKDIR /opt/bt-studio/frontend
RUN yarn install
RUN yarn start

EXPOSE 3000

And now how to start it:

docker build -t bt-studio .
docker run -it -p 3000:3000 bt-studio

You should find the frontend at http://localhost:3000

When using the dockerized environment, the state of the execution is known. This could be used to change the run icon to a pause button, as in RA, leveraging the new RADI capabilities.

Currently, there are two remaining BT.cpp decorators unsupported:

• Delay
• RunOnce

This would require two main changes in the frontend:

• A VNC client to connect to the gazebo simulation in the RADI.
• A communication manager with the RADI

Currently, the execution of BT Studio uses two separate testing servers: the django for the backend and yarn for the frontend.

It would be a better approach to serve the frontend as another django webapp using webpack to bundle it.

Altough proper terminate() methods are written in the actions, in the Gazebo Simulator (doesn't happen in Webots), when the user finishes the execution, the terminate methods aren't called and the last command keeps executing. This could be dangerous in a real robot.

The generation of a self contained tree is a key function of the system, and as such, it should be properly separated: the package should allow an ros-agnostic step of generation based on the tree xml and the actions, which can later be used to generate an executable tree.

Really nice project!

However I think there may be a few improvements:
1.- The zoom is really anti-intuitive.
2.- The selected node should be highlighted.
3.- A node should not have two parents.
4.- Baobab could be a great name for this project!!

The web frontend lacks some key interface buttons:

• File deletion and creation
• Saving a file
• Removing blocks (and connections between them)