junjungoal / directfutureprediction

Code for the paper "Learning to Act by Predicting the Future", Alexey Dosovitskiy and Vladlen Koltun, ICLR 2017

Code for the paper "Learning to Act by Predicting the Future" by Alexey Dosovitskiy and Vladlen Koltun

If you use this code or the provided environments in your research, please cite the following paper:

@inproceedings{DK2017,
author    = {Alexey Dosovitskiy and Vladlen Koltun},
title     = {Learning to Act by Predicting the Future},
booktitle = {International Conference on Learning Representations (ICLR)},
year      = {2017}
}

Dependencies:

• ViZDoom
• numpy
• tensorflow
• OpenCV python bindings
• (Optionally - cuda and cudnn)

Tested with:

• Ubuntu 14.04
• python 3.4
• tensorflow 1.0
• ViZDoom master branch commit ed25f236ac93fbe7f667d64fe48d733506ce51f4

Running the code:

• Adjust ViZDoom path in doom_simulator.py

• For testing, switch to the pretrained_models branch and run (using D2 as an example):

    cd examples/D2_navigation
    python3 run_exp.py show
  

• For training, run the following (using D2 as an example):

    cd examples/D2_navigation
    python3 run_exp.py train
  

• If you have multiple gpus, make sure that only one is visible with

    export CUDA_VISIBLE_DEVICES=NGPU
  

  where NGPU is the number of GPU you want to use, or "" if you do not want to use a gpu

• For speeding things up you may want to prepend "taskset -c NCORE" before the command, where NCORE is the number of the core to be used, for example:

    taskset -c 1 python3 run_exp.py train
  

Remarks

• For experiments in the paper we used a slightly modified ViZDoom version which provided the first post-mortem measrements. Surprisingly, this turns out to make a noticeable difference for training. For this reason the results with this code and default ViZDoom may slightly differ from the results in the paper.

• Results may vary quite significantly across training runs: in our experiments, up to roughly relative 15%.

• In battle scenarios reward provided by vizdoom is set to be simply the number of frags. For training the baseline approaches we did not use this reward provided by the environment, but rather a weighted average of the three measurements, same as for our approach.

Troubleshooting

Please send bug reports to Alexey Dosovitskiy ( [email protected] )