MPSV

The Motion Planner for Surface Vehicles solves motion planning problems for fully actuated surface vehicles with non-linear dynamics in an environment with non-convex non-moving obstacles. The method is based on the RRT* algorithm. The problem is decomposed into two sequential problems in which first a feasible path is searched for and then a collision-free trajectory is calculated based on the path found. For references, see below.

Header-only library with MATLAB/Simulink support

The MPSV solver is implemented as header-only library (C++) and requires a compiler that supports the C++20 language features. There exists a MATLAB/Simulink toolbox that allows you to run the planner in Simulink. To use the toolbox, open the matlab/MPSV.prj project file or add it as reference project to an existing MATLAB project. Run mpsv.BuildDrivers() once to build all driver blocks of the simulink library. All blocks are supported for C++ code generation.

OpenMP

MPSV requires OpenMP for parallel computing. It has been tested on Windows via MSYS2/MinGW64 and Ubuntu. The package usually comes with GCC. On Ubuntu it can be installed via sudo apt install libomp-dev, for example. It may be necessary to set the environment variables OMP_PLACES or OMP_PROC_BIND.

Overview

The directory structure of this repository is as follows.

File / Directory	Description
apps	contains example applications and a standalone application (C++)
documentation	contains an HTML-based documentation
matlab	contains the MATLAB/Simulink project for the MPSV toolbox
source	contains the header-only library (C++)
LICENSE	license information
Makefile	used to build example applications and the standalone application
README.md	this file

Planning Levels

The API of MPSV consists of several planning levels that are shown in the following image.

PathPlanner

• find path from initial pose to final pose
• RRT* with fixed number of nodes and warm start capabilities

MotionPlanner

• find trajectory from initial state/input to final pose around a given path
• RRT* with fixed number of nodes
• internal control system based on feedback linearization and full-state feedback to generate dynamically feasible trajectory

SequentialPlanner

• combine path and motion planner in sequential manner
• path from path planner is used as guidance for the motion planner

OnlinePlanner

• continuously solve sequential planning problems
• solves motion planning problems for a future timepoint
• merges solution trajectory seamless with a previous solution

AsyncOnlinePlanner

• run the online planner in a separate thread
• results are obtained via polling or callback

When using the header-only library the desired planning level can be used directly in a custom C++ application (see examples). However, when using MATLAB/Simulink, only the PathPlanner and the AsyncOnlinePlanner are currently implemented.

Examples (C++)

Take a look to example applications in the apps directory on how to use the different planning levels of MPSV.

Building a single example

To build a specific application example, run

make app=NAME

where NAME has to be replaced by the name of the application, e.g. the folder name of the application to be built.

Building all examples

To build all example applications, run

make apps

Examples (MATLAB/Simulink)

Example models for Simulink are located at matlab/examples/. Make sure that you build already build the driver blocks of the Simulink library with mpsv.BuildDrivers().

Standalone Application

The standalone application can be used to outsource the planning problem to an external program. The data transfer between the standalone application and a user program is done via UDP. The application is located at apps/mpsv and is compiled with

make app=mpsv

Further details are given here.

Publication

Robert Damerius and Torsten Jeinsch, Real-Time Path Planning for Fully Actuated Autonomous Surface Vehicles, 2022 30th Mediterranean Conference on Control and Automation (MED), 2022, doi: 10.1109/MED54222.2022.9837178

@inproceedings{Damerius2022,
    author    = {Damerius, Robert and Jeinsch, Torsten},
    title     = {{Real-Time Path Planning for Fully Actuated Autonomous Surface Vehicles}}, 
    booktitle = {2022 30th Mediterranean Conference on Control and Automation (MED)}, 
    volume    = {},
    number    = {},
    pages     = {508-513},
    year      = {2022},
    doi       = {10.1109/MED54222.2022.9837178}
}

Robert Damerius, Johannes R. Marx and Torsten Jeinsch, Fast Trajectory Generation on a Path using Feedback Linearization, 22nd IFAC World Congress, 2023, doi: 10.1016/j.ifacol.2023.10.796

@article{Damerius2023,
    author    = {Robert Damerius and Johannes R. Marx and Torsten Jeinsch},
    title     = {{Fast Trajectory Generation on a Path using Feedback Linearization}},
    note      = {22nd IFAC World Congress},
    journal   = {IFAC-PapersOnLine},
    volume    = {56},
    number    = {2},
    pages     = {10990-10995},
    issn      = {2405-8963},
    year      = {2023},
    doi       = {10.1016/j.ifacol.2023.10.796}
}