This repository contains code for the RA-L paper "An Informative Planning Framework for Target Tracking and Active Mapping in Dynamic Environments with ASVs".

The paper can be found here. If you found this repository useful, feel free to cite it.

@article{sanjeev2026informative,
  title={An informative planning framework for target tracking and active mapping in dynamic environments with asvs},
  author={Ramkumar Sudha, Sanjeev and Popovi{\'c}, Marija and Coates, Erlend M},
  journal={IEEE Robotics and Automation Letters},
  volume={11},
  number={3},
  pages={2690--2697},
  year={2026},
}

The simulation experiments are performed with the VRX simulator. The software has been developed and tested with the simulator in Ubuntu 22.04 with ROS2 Humble and Gazebo garden. To get started, build a docker image by following the setup instructions. First clone the repository in your workspace:

git clone [email protected]:sanjeevrs2000/ipp_dyntrack.git

Then build the docker image:

docker build -f Dockerfile -t user/ipp_dyntrack:1.0 .

To open a container with the newly built image:

chmod +x run_docker.sh
./run_docker.sh

Note: To use GPU acceleration with NVIDIA GPUs, ensure that you have nvidia-ctk installed. If you do not have a GPU, in docker-compose.yaml comment the runtime: nvidia argument before running docker compose up as it might cause some errors.

The packages dyntrack_planner contains nodes related to implementation of our IPP framework. The package asv_control implements a path following PD controller for the ASV based on line of sight (LOS) guidance. In the vrx_gz package, some files have been modified from the original simulator for the sake of our experiments.

To access a bash terminal inside the container, open a new terminal and run:

docker exec -it ipp_dyntrack_container bash

First, build the packages and then source the workspace in the container:

colcon build --merge-install
source install/setup.bash

To run an instance of the IPP framework inside the docker container, run the launch file:

ros2 launch dyntrack_planner informative_planner.launch.py headless:=True 
# run without headless:=False to view the Gazebo GUI

To visualize with rviz2, open another bash terminal inside the container and run:

ros2 launch dyntrack_planner vis_rviz.launch.py

For further experimentation, you can tune some parameters related to the experiments in src/dyntrack_planner/dyntrack_planner/params.py.

To view the setup for dataset generation, training and testing of the spatiotemporal prediction network, go to the src/spatiotemp_pred_nn directory. To run sample predictions with the network, inside the container:

cd src/spatiotemporal_pred_nn/src
python3 test.py

To run a training instance, run python3 train.py.