"Pickup objects" task using behavior trees in ROS

This repository encompasses the implementation of a behavior trees node in ROS for a turtlebot3 platform in ROS noetic Gazebo to complete an object collecting task. This work is developed as a component of Lab 02 for the Hands-on Planning course offered by IFROS.

Contributors

• Atharva A. Patwe
• Loc Thanh Pham

Instructor

• Dr. Narcis Palomeras Rovira

Date

March 11, 2024

Introduction

This repository hosts the codebase for a pickup behaviors tree node designed in ROS, crafted as a requisite for the Hands-on Planning course at IFROS. This project utilizes the py_trees library.

Using these behavior tree, we implement the following task.

The robot has to navigate to a list of points and in each one it has to check if there is an object on it. If there is an object, it picks it up and takes it to (-1.5, -1.5) if the object was a beer can or to (1.5, 1.5) if the object was a coke can. Once there, the robot leaves the object and goes to the next point. The task ends once two objects have been collected or all provided points have been explored. The points where the objects can be found are:

• (1.25, 0.5)
• (1.25, -1.25)
• (0.0, -1.25)
• (-0.5, 1.25)
• (-1.25, 0.5)

This package is developed and tested in ROS Noetic on Ubuntu 20.04 machines.

Setting up

Prerequisites

This lab uses the py_trees library. To install the py_trees library do:

pip install py_trees

To implement the pick up objects task package perform following steps:

1. Install all the turtlebot packages from here
2. Clone this repository as a package in the src folder
3. Build the catkin workspace : catkin build or catkin_make
4. Source the setup.bash file: source devel\setup.bash
5. Export turtlebot model: export TURTLEBOT3_MODEL=burger

Task 1: Free map without obstacles

1. Launch the pick up objects task 1 launch file present in the launch folder: roslaunch pick_up_objects_task pick_up_objects_task_1.launch
2. Run the pickup_behaviors_node.py file: rosrun pick_up_objects_task pickup_behaviors_node.py

Task 2: Map with obstacles (with the path planning controller that running RRT algorithm to find path)

1. Clone repository turtlebot_online_path_planning from here as a package in the src folder.
2. Launch the pick up objects task 2 launch file present in the launch folder: roslaunch pick_up_objects_task pick_up_objects_task_2.launch
3. Run the pickup_behaviors_node.py file: rosrun pick_up_objects_task pickup_behaviors_node.py

To Visualise the Tree, Path other markers, add the markers in rviz

Implementation

The code is organized mainly in three files the ROS Node file, manage objects helper file and turtlebot controller helper file

Planning Objective

The main goal of this project is to develop a behavior tree planning a sequence of behaviors need to be done to achieve the picking objects task. The Pick up behaviors node, developed for this purpose.

This behavior tree has following behaviors:

Check number of points less than N_points (length of points list)

It checks the number of points in the list, which were travelled to by the robot. If it is less than N_points, the behavior return True. Meanwhile, if it is equal or greater than N_points, return False.

Check number of collected objects less than N_objects (number of objects on the map)

It checks the number of collected objects. If it is less than N_objects, the behavior return True. Meanwhile, if it is equal or greater than N_objects, return False.

Set Point

This behavior sets the next point location the robot need to travel to based on number of points the robot already went to. The robot goes to each point in the list sequencely. Then, number of travelled points will increase one. Finally, return True.

Move to Point

The robot move to the point x, y location in the point list. This behavior is provided by the node pickup_behavior_node.py from this package. To be more specific, this behavior publishes the x, y location, then compares subcribed position of the robot to the target position as an ending condition. If this distance is less than 0.35, return True, otherwise, return False.

Check Object

This behavior runs a service check_object provided by the manage_objects_node.py. It returns False if no object is close to the robot and True plus the object's name if an object is close to it. If it returns True, number of collected objects will increase 1.

Get Object

This behavior run a service get_object provided by the manage_objects_node.py. It returns False if no object is close to the robot and True if an object is close to it. It also moves this object over the robot.

Move to Destination

The robot move to the destination x, y location, where robot will leave object and depends on object type. This behavior is provided by the node pickup_behavior_node.py from this package. To be more specific, this behavior publishes the x, y location, then compares subcribed position of the robot to the target position as an ending condition. If this distance is less than 0.35, return True, otherwise, return False.

Let Object

This behavior run a service let_object provided by the manage_objects_node.py. It returns False if no object is over the robot and True if an object is over it. It also moves this object to the floor.

Overall Contribution

The overall contributions of this work can be summarized as follows:

• Behavior tree

Limitations and Scope for improvements

1. With optional part, we used RRT algorithm to find path. However, in this lab, a laser is used to scan obstacles on the map leading to the robot thinks objects are obstacles and RRT algorithm did not work. To deal with this problem, in the turtlebot_online_path_planning package, class StateValidityChecker, we added a new method, set() which set neighborhood of the goal is 0, after get the occupancy map. It is not a perfect solution, but it is a reasonable answer in this situation. In our point of view, in order to completely resolve the problem, we need to another sensor besides the laser, which allows us to detect objects and update the map.

Demonstration

A video showing the working of the behavior tree is showed below.

Task 1: Map without obstacles

Task 2: Map with obstacles