Online approach to near time-optimal task-space trajectory planning

This repo brings the comparison of the proposed Cartesian Space trajectory planning approach against toppra. The method is based on Trapezoidal Acceleration Profile (TAP) planning which is implemented using ruckig

The method is described more in detail within the preprint.

Interactive simulation

Now available in the form of the Hugging Face space: https://askuric-capacity-aware-trajectory-planning.hf.space/

Installation

The code is implemented in Python and we strongly suggest to use anaconda to install the necessary libraries.

You can install the dependencies using:

conda env create -f env.yaml

And then activate the environment

conda activate planning_env

Pynocchio

This code uses a custom wrapper over pinocchio library called pynocchio which can be found here.

Launch the code

To access the code launch the jupyter lab

jupyter lab

and launch the notebook:

• comparison_with_toppra_rtb.ipynb - using robotics-toolbox-python for the robot simulation
• comparison_with_toppra_pin.ipynb - using pinocchio for the robot simulation

You can also run the comparison script directly using:

python comparison_with_toppra_pin.py

And then follow the instructions within the script and see the plots and robot animation.

Kazam_screencast_00053.mp4

For example comparing the two approaches position, velocity, acceleration and jerk profiles on the given trajectory: and the error between the two approaches:

You will also be able to see the side-by-side animation of the two robots following the same trajectory where the left robot is using our approach and the right robot is using toppra approach.

Kazam_screencast_00056.mp4