PathParam — Path-Parametric Planning & Control

For the implementation details, please check the paper, and/or the website.

Follow these steps:

1. Create a python environment with python 3.10. For example, with conda:

   conda create --name pathparam python=3.10
   conda activate pathparam
   pip install -r requirements.txt

2. Install acados, as well as its python interface.

3. Update python path to ensure that the package is visible to python.

   export PATHPARAM_PATH=$path_to_pathparam
   export PYTHONPATH=$PYTHONPATH:$PATHPARAM_PATH
   

   Make sure to add this to your .bashrc or .zshrc files.

To compare the Frenet-Serret (FSF) and Parallel Transform(PTF) frames, run the following command:

python examples/moving_frame_comparison.py

In here we conduct the experiments associated with the robotic manipulator example in the paper.

With this set of examples we aim to answer the question Why path-parametric?. We do this by conducting three different experiments:

1. Tracking vs Following: Shows the advantage of an spatial reference over a temporal one. Run it with:

   python examples/robotic_manipulator/example1_tracking_vs_following.py --case f --d

   The options are the following ones:

   • --case: f (following) or t (tracking)
   • --d: Introduces disturbance that mimics the robot blockage

2. Velocity Profile: Shows how path-parametric approaches allow to have a desired velocity profile, a very attractive feature for certain applications. Run it with:

   python examples/robotic_manipulator/example2_velocity_profile.py --case 2

   where --case defines the velocity profile as 1 (constant), 2 (sinusoidal) or 3 (quadratic).

3. Corridors - Control around a reference: Shows how path-parametric references are a more general way to conduct motion control around a trajectory -- by means of corridors/tunnels/funnels--, accounting for deviations within a predetermined volume, and therefore, achieving behaviors that are considerably more flexible. Run it with:

   python examples/robotic_manipulator/example3_corridor.py

For the brachistochrone problem run

python examples/brachistochrone/brachistocrhone.py

For the dubins path as a path-parametric problem run

python examples/dubins/dubins.py

When comparing progress maximization against a standard time minimization approach, we used the well-known contouring control based formulation.

To replicate these results, see the original corrgen repository.