/BT-learning-framework

A Framework for Learning Behavior Trees in Collaborative Robotic Applications

Primary LanguagePythonBSD 3-Clause "New" or "Revised" LicenseBSD-3-Clause

Overview

Framework for learning Behavior Trees (BTs) based on Genetic Programming (refer to this paper) and Learning from Demonstration (refer to this paper).
The target tasks are robotic manipulation ones and are simulated in the AGX Dynamics from Algoryx.

Experiments

  • The GUI allows user to add demonstrations and start/stop/resume the evolution of BTs through the genetic programming.
  • The experiments are defined by uncommenting the corresopnding lines in the config file. Morover, adjust the parameters as explained in the paper.
  • For all experiments but the last, set random_bringup to False.
  • The data with which the plots for the paper have been obtained are provided in the log folder.

Disclaimer

To run the experiment is necessary to have a valid license and installation of the simulator from Algoryx. Please contact the authors of the paper for further instructions.