Real-Time Fast Marching Tree (RT-FMT) for Mobile Robot Motion Planning in Dynamic Environments (ICRA 2023)

Overview

This repository contains the implementation of the Real-Time Fast Marching Tree (RT-FMT) algorithm, as proposed in the paper titled "Real-Time Fast Marching Tree (RT-FMT) for Mobile Robot Motion Planning in Dynamic Environments". The algorithm is designed for robotic motion planning, featuring local and global path generation, multiple-query planning, and dynamic obstacle avoidance.

Simulation Results:

The repository includes simulation code demonstrating the superior performance of RT-FMT compared to RT-RRT* in terms of execution cost and arrival time in most scenarios. The simulations also showcase the benefits of taking a local path before the global path is available, emphasizing the algorithm's capability to reduce arrival time even when faced with the possibility of selecting an inferior path.

                                                            Dynamic example                                                   Maze example

Getting started

• Install and open the unity hub software, then open a new project in the cloned repository folder. If you haven't done so It will ask you to install unity 2018.1.0f2
• Once you open the imported project, you will find all scenes under the Assets/Scenes folder, and the code related to the paper can be found in the Assets/Scripts/RT-FMT folder.
• The scenes contain different environments that can be used to visualize the code in Scenes/Example and to run the environments used in the paper, which are located at Scenes/Experiments

Code Structure

The RT-FMT implementation is organized into the following key components:

All RT-FMT code is located at the Scripts/RT-FMT Directory

• RTFMTPlanner.cs: Contains the core implementation of the RT-FMT algorithm.
• RTFMT_example.cs: Instantiates the planner and controls the sphere in the example scenes.
• RTFMT_exp.cs: Instantiates the planner and controls the sphere in the experiment scenes. This code has extra features to validate the experiment by running it multiple times and logging the results into a .csv file.