Metric-Semantic SLAM with Kimera: A Hands On Tutorial

Introduction

This tutorial aims to guide you through using Kimera's modules for running a real-life demo with the Intel Real-Sense depth camera D435i, as well as executing Kimera on the Euroc dataset and the Go-SEEK simulator. By the end of this tutorial, you will be equipped with the knowledge to implement metric-semantic SLAM (Simultaneous Localization and Mapping) effectively.

Step 1: Setting Up Your Environment

Before diving into the practical applications, you need to set up the necessary software environment.

1. Install Dependencies

   • Ensure you have the following software installed:
     • ROS (Robot Operating System)
     • OpenCV
     • Eigen
     • PCL (Point Cloud Library)

2. Clone Kimera Repositories

   • Open your terminal and run the following commands to clone the Kimera repositories:

     git clone https://github.com/MIT-SPARK/Kimera-VIO.git
     git clone https://github.com/MIT-SPARK/Kimera-VIO-ROS.git
     git clone https://github.com/MIT-SPARK/Kimera-RPGO.git
     git clone https://github.com/MIT-SPARK/Kimera-Semantics.git
     

3. Build the Packages

   • Navigate to your catkin workspace and build the packages:

     cd ~/catkin_ws
     catkin_make
     

4. Source Your Workspace

   • Don't forget to source your workspace:

     source devel/setup.bash
     

Step 2: Running the Real-Life Demo

To utilize the Intel Real-Sense D435i camera, follow these steps:

1. Connect the Camera

   • Make sure the camera is connected to your computer and recognized by the system.

2. Launch the Camera Node

   • In a new terminal, run the following command to launch the camera node:

     roslaunch realsense2_camera rs_camera.launch
     

3. Run Kimera-VIO

   • Open another terminal and execute Kimera-VIO with the camera data:

     roslaunch kimera_vio kimera_vio.launch
     

4. Visualize the Output

   • Use RViz to visualize the SLAM results. Open RViz in a new terminal:

     rosrun rviz rviz
     

   • Add the appropriate displays to visualize the camera feed and SLAM output.

Step 3: Using the Euroc Dataset

To run Kimera on the Euroc dataset, proceed with the following:

1. Download the Euroc Dataset

   • Visit the Euroc dataset website and download the dataset files.

2. Modify Configuration Files

   • Adjust the configuration files in Kimera-VIO to point to your downloaded dataset. Ensure the paths in the configuration files match the dataset's location.

3. Execute the Dataset with Kimera

   • Run Kimera-VIO with the dataset:

     roslaunch kimera_vio kimera_vio.launch dataset_path:=<path_to_your_euroc_dataset>
     

4. Visualize Results

   • As with the real-life demo, use RViz to visualize the output from the Euroc dataset.

Step 4: Running the Go-SEEK Simulator

Interact with the Go-SEEK simulator to test Kimera in a photorealistic environment:

1. Clone Go-SEEK Repository

   • Clone the Go-SEEK challenge repository:

     git clone https://github.com/MIT-TESSE/goseek-challenge.git
     

2. Set Up the Simulator

   • Follow the instructions in the Go-SEEK repository to set up the simulator and ensure it runs properly.

3. Launch the Simulator

   • Start the Go-SEEK simulator as directed in the documentation.

4. Run Kimera with Go-SEEK Data

   • Similar to previous steps, run Kimera-VIO with the simulator's output data.

Conclusion

In this tutorial, you learned how to set up an environment for Kimera, run demonstrations using an Intel Real-Sense camera, utilize the Euroc dataset, and work with the Go-SEEK simulator. Each step builds your capabilities in metric-semantic SLAM, enabling you to apply these techniques in various robotic applications. For more advanced usage, consider exploring the linked papers and additional resources provided in the video description.