This is an experimental code for reproducing below paper's result using chainer.

A. Tamar, Y. Wu, G. Thomas, S. Levine, P. Abbeel, Value Iteration Networks,
Neural Information Processing Systems (NIPS) 2016.

This is preparation code for value iteration networks. This script generats grid word and training data of shortest path. After running this script, map_data.pkl is generated at current directory.

python script_make_data.py

We assume that you already have run preparation script and had the map_data.pkl. This script trains VIN network and generates trained weight in a result directory.

python train.py --gpu 0

This code is testing script of VIN.

python predict.py --model [trained weight]