Dynamic Neural Potential Field: Online Trajectory Optimization in Presence of Moving Obstacles

Official Implementation of Dyn-NPField

Paper | Video | Models | Dataset

OVERVIEW

We address local trajectory planning for a mobile robot in the presence of static and dynamic obstacles. The trajectory is computed as a numerical solution to a Model Predictive Control (MPC) problem, with collision avoidance incorporated by adding obstacle repulsive potential to the MPC cost function. Our approach estimates this repulsive potential using a neural model. We explore three strategies for handling dynamic obstacles: treating them as a sequence of static environments, predicting a full sequence of repulsive potentials at once, and predicting future potentials step by step in an autoregressive mode.

GETTING STARTED

Using Docker

Prerequisites:

1. Docker: Ensure Docker is installed on your machine. You can download it from Docker's official site.
2. NVIDIA GPU: Since this Dockerfile is based on NVIDIA CUDA, having an NVIDIA GPU and ensuring NVIDIA Docker Toolkit is installed are necessary for GPU acceleration.

Steps:

1. Clone the Repository: Begin by cloning the repository containing the Dockerfile to your local machine.

   git clone https://github.com/CognitiveAISystems/Dynamic-Neural-Potential-Field
   cd Dynamic-Neural-Potential-Field

2. Build the Docker Image: Use the following command to build the Docker image from the Dockerfile in your repository. Replace <your-image-name> with a name of your choice for the Docker image.

    docker build -t dyn_npfield .

3. Run the Docker Container: After the image is built, run a container from it. Replace <your-image-name> with the name you used in the previous step.

   docker run -it --gpus all --name dyn_npfield -p 80:80 dyn_npfield

   The --gpus all flag enables GPU support in the container, and -p 80:80 forwards port 80 from the container to the host, enabling any web service running inside the container to be accessible on the host machine.

4. Accessing the Container: You can access the running container via:

    docker exec -it dyn_npfield /bin/bash

   This will open a bash shell inside the container where you can interact with the software and run commands.

Troubleshooting:

• Ensure CUDA-compatible drivers are installed on your host machine.
• Check for any Docker-related errors during image building or container running by examining the Docker logs.

Without Docker

Prerequisites:

• Ubuntu 22.04: Ensure you are running on an Ubuntu 22.04 environment.
• NVIDIA GPU: A compatible NVIDIA GPU is necessary since the software relies on CUDA libraries.
• Python, Git, and Build Tools: Install Python, Git, CMake, GCC, and other necessary tools.

Steps:

1. Install CUDA and NVIDIA Drivers: Install CUDA 12.4.1 and corresponding NVIDIA drivers if they are not already installed.

   sudo apt-get install nvidia-cuda-toolkit

2. Clone the Repository: Clone the repository that contains your project.

   git clone https://github.com/CognitiveAISystems/Dynamic-Neural-Potential-Field
   cd Dynamic-Neural-Potential-Field

3. Set Up Python Environment: It's recommended to use a virtual environment.

   python3 -m venv env
   source env/bin/activate
   pip install --upgrade pip

4. Third Party Libraries: Compile and install the following libraries and their dependencies (PyTorch, CMake etc):

   • acados
   • l4casadi

5. Run the Application: Navigate to NPField directory, choose the required version (d1, d2 or d3) and run:

   python test_solver.py
   

Troubleshooting:

• Ensure all dependencies and their versions are correctly installed.
• Verify CUDA and GPU drivers are properly installed and recognized by your system.

CITATION

If you use this framework please cite the following two papers:

Dyn-NPField:

@misc{staroverov2024dynamicneuralpotentialfield,
      title={Dynamic Neural Potential Field: Online Trajectory Optimization in Presence of Moving Obstacles}, 
      author={Aleksey Staroverov and Muhammad Alhaddad and Aditya Narendra and Konstantin Mironov and Aleksandr Panov},
      year={2024},
      eprint={2410.06819},
      archivePrefix={arXiv},
      primaryClass={cs.RO},
      url={https://arxiv.org/abs/2410.06819}, 
}

NPField:

@misc{alhaddad2023neuralpotentialfieldobstacleaware,
      title={Neural Potential Field for Obstacle-Aware Local Motion Planning}, 
      author={Muhammad Alhaddad and Konstantin Mironov and Aleksey Staroverov and Aleksandr Panov},
      year={2023},
      eprint={2310.16362},
      archivePrefix={arXiv},
      primaryClass={cs.RO},
      url={https://arxiv.org/abs/2310.16362}, 
}