Skip to content

TensorFlow op to calculate a robot-dynamics-based loss and its preconditioned gradient

License

Notifications You must be signed in to change notification settings

DFKI-CPS/tf-dynamics-op

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

1 Commit
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

tf-dynamics-op

This repository contains source code for TensorFlow ops to calculate dynamics-based loss functions to train neural networks to produce efficient and feasible robot trajectories.

Requirements

  • Linux + a C++ compiler (supporting at least C++14)
  • TensorFlow (version 2.7.0 is known to work)
    • must have been compiled using -D_GLIBCXX_USE_CXX11_ABI=1 for compatiblity with pinocchio
  • LAPACK
  • pinocchio (version 2.5.6 is known to work)

Building

Run make in the main directory of this repository. Despite the low number of files, this can take a while.

Usage

dynamics.py shows exemplary usage of the ops wrapped in a function with custom gradient.

About

TensorFlow op to calculate a robot-dynamics-based loss and its preconditioned gradient

Resources

License

Stars

Watchers

Forks