Description

This repository contains the code of the IV2025 paper called "Dynamic Objective MPC for Motion Planning of Seamless Docking Maneuvers"

O. Schumann, M. Buchholz, and K. Dietmayer, “Dynamic Objective MPC for Motion Planning of Seamless Docking Maneuvers,” in accepted for publication at IEEE 36th Intelligent Vehicles Symposium (IV), 2025

This algorithm is based on a model predictive contouring controller (MPCC) created in acados, which is improved by the methods proposed in this paper. These changes allow for high-precision motion planning along a reference path while being able to reach specific goal poses at the same time without stopping or switching to another controller. Especially in docking scenarios, this is a common use case.

Paper

arXiv IEEE (TBA)

Citation

@INPROCEEDINGS{schumann2025,
  author={Schumann, Oliver and Buchholz, Michael and Dietmayer, Klaus},
  booktitle={accepted at IEEE 36th Intelligent Vehicles Symposium (IV)}, 
  title={{Dynamic Objective MPC for Motion Planning of Seamless Docking Maneuvers}},
  volume={},
  year={2025},
  number={},
}

Videos

Setup

1. Build the docker image

cd docker && ./build.sh && cd ..

2. Run the docker

./run_docker.sh

3. Build and source the module

colcon build
source colcon_build/install/setup.zsh

4. Run

ros2 run corridor_planning simulation.py --ros-args -p baseline:=-1 -p track_switch:=6

Different configurations of the planning algorithm can be run, depending on some parameters:

baseline
-1: dynamic objective MPC
0: separated motion plans
1: switched MPCs

track_switch (tracks used in this paper are bold)
0: racing track
1: track with switch and goal pose
2: partially narrow track with switch and goal pose
3: left right switches
4: huge left turn
5: narrow gap
6: track with switch and goal track with switch and goal pose