acxz / mppi

A simple and extensible Octave/Matlab library for Model Predictive Path Integral control scheme.

MPPI

A sampling based MPC derived from free energy principles implemented in Octave/Matlab.

Dependencies

• Octave or Matlab
  • Octave Download Page
  • Matlab Store Page

Usage

mppi.m is just a function that can be used given the required inputs. In the examples/ directory some example code is provided that uses mppi.m. To run an example code just run example_main.m from Octave or Matlab.

References

G. Williams, P. Drews, B. Goldfain, J. M. Rehg and E. A. Theodorou, "Information-Theoretic Model Predictive Control: Theory and Applications to Autonomous Driving," in IEEE Transactions on Robotics, vol. 34, no. 6, pp. 1603-1622, Dec. 2018.

Citation

To cite acxz/mppi in publications use:

acxz. (2019, November 13). acxz/mppi: MPPI in Octave/Matlab (Version v0.9.0). Zenodo. http://doi.org/10.5281/zenodo.3539762

Here is a BibTex entry for convenience:

@software{acxz_2019_3539762,
  author       = {acxz},
  title        = {acxz/mppi: MPPI in Octave/Matlab},
  month        = nov,
  year         = 2019,
  publisher    = {Zenodo},
  version      = {v0.9.0},
  doi          = {10.5281/zenodo.3539762},
  url          = {https://doi.org/10.5281/zenodo.3539762}
}

Click here or the above DOI badge for more information.

mppi's Issues

Revisit Clamping

Clamping needs to be revisited.

• make sure to use clamped value in traj cost computation
• use clamped error when calculating du
• add feature to create so many unique samples, since clamping will reduce many samples to the same value and limit exploration

Investigate trajectory cost computation

https://github.com/acxz/mppi/blob/master/mppi.m#L128

Seems to be missing a u - v term see Info-Theoretic Model Predictive Control Theory and Appl to Auto Driv https://arxiv.org/pdf/1707.02342.pdf

Edit: but in linked paper it is just v, hmmm

Separate control law calculation and system simulation functionality.

Save and plot trajectory cost values

Add two link IP example

Better plotting (labels, etc.)

Add better documentation

Like input/output descriptions for each parameter of the function.

Allow for the ability for users to supply their own plotting func to mppi

Implement helpful input argument checking bounds/values