Local Motion Planning Benchmark Suite

This repository is meant to allow quick comparison between different local motion planning algorithms. Running and postprocessing is available and we aim to offer a nice interface to implement a wrapper to your own motion planner.

Screenshots

Trajectory planar arm	Trajectory point robot	Simulation panda arm

Evaluation of series

Getting started

This is the guide to quickly get going with the local motion planning benchmark suite.

Pre-requisites

• Linux Ubuntu LTS >= 18.04
• Python >3.6, < 3.10
• pip3
• gnuplot (sudo apt install gnuplot)
• [Optional] poetry
• [Optional] embotech forces pro for mpc
• [Optional] acados_template for mpc

Installation

You first have to download the repository

git clone [email protected]:maxspahn/localPlannerBench.git

Then, you can install the package using pip as:

pip3 install .

Optional: Installation with poetry

If you want to use poetry, you have to install it first. See their webpage for instructions docs. Once poetry is installed, you can install the virtual environment with the following commands. Note that during the first installation poetry update takes up to 300 secs.

poetry update
poetry install

The virtual environment is entered by

poetry shell

Tutorial

Simple

The following is a very simple example case containing a point mass robot and a PD planner.

Run an experiments:

Experiments should be added in separate folder in examples. One very simple example can be found in this folder. Note that you need to active your poetry shell if you have installed the package using poetry by

poetry shell

Or alternatively active your virtual python environment

Then you navigate there by

cd examples/point_robot

Then the experiment is run with the command line interface

runner -c setup/exp.yaml -p setup/pdplanner.yaml --render

Postprocessing:

The experiments can be postprocessed using the provide executable. Again make sure you are in the virtual environment, when using poetry run: (poetry shell)

cd examples/point_robot

The you can run the post processor with arguments as

post_process --exp path/to/experiment -k time2Goal pathLength --plot

Advanced

To showcase the power of localPlannerBench we would also like to show you a more complex example, containing the 7-DoF frankaemika panda robot arm and a custom opensource acados based MPC planner.

Again make sure you are in your virtual python environment.

poetry shell

Install acados_template inside of the virtual environment if you haven't already.

Then you navigate to

cd examples/panda_arm

Then the experiment is run with the command line interface

runner -c setup/exp.yaml -p setup/acados_mpc.yaml --render

The you can run the post processor with arguments as

post_process --exp results --latest -k time2Goal pathLength --plot