v0.1.5

Added LSTM models

Library is still under heavy development, so everything's changing all the time

Reward paper

crowd-rewards

One more

Thesis version

Freezing the code as of the (probably) last PhD paper. After this, I might start decoupling the RL code from the thesis-specific elements to make it more widely usable.

Understanding RL Crowds

This version of coltra has been used for most of the experiments in "Understanding Reinforcement Learned Crowds"

MCMES

Version of the code used for MCMES experiments

Fixed optuna sweep

Remove number of episodes from the search

First optuna sweep

optuna-first

COME ON

v0.1.0-smartnav

This is a frozen version of the repository to work with SmartNav and SmartHider.

Note that the API diverged a bit between the two projects. SmartHider includes a legacy mode to be compatible with that. By default, this repo works with non-legacy SmartHider. To support legacy/SmartNav, make sure to appropriately modify the METRICS list in train_smartnav.py.