Learning symmetric and low-energy locomotion




  • Paper topic: Animation and Simulation
  • Software type: Code
  • Able to run a replicability test: True
  • Replicability score: 3
  • Paper listed in the Graphics Replicability Stamp Initiative
  • Software language: Python
  • License: MIT
  • Build mechanism: Other script
  • Dependencies: libeigen3-dev libassimp-dev libccd-dev libfcl-dev libboost-regex-dev libboost-system-dev libopenscenegraph-dev libbullet-dev liburdfdom-dev libnlopt-dev libxi-dev libxmu-dev freeglut3-dev libtinyxml2-dev swig tensorflow numpy dart pydart
  • Documentation score {0,1,2}: 1
  • Reviewer: Julie Digne <julie.digne@liris.cnrs.fr>
  • Time spent for the test (build->first run, timeout at 100min): 100min

Source code information


All scripts are provided for reproducing the examples in the paper.
The dependencies can be easily installed using a script, but on ubuntu 19.10 I had to install libnlopt-cxx-dev.
The policy learning went well, but I was not able to produce the results of the paper, due to the failure of the test_policy function with error: Attempted to look up malformed environment ID: b'../data/precomp_data/dog_run1/policy_params.pkl'. (Currently all IDs must be of the form ^(?:[\w:-]+\/)?([\w:.-]+)-v(\d+)$.)
This occured also when testing the provided precomputed policies.

I also tried with an older release of dart-env (v0.7.4 - 2017) which seems to be handling registration but the same problem occured.

If you want to contribute with another review, please follow these instructions.

Please consider to cut/paste/edit the raw JSON data attached to this paper.