search

Unity-Technologies / ml-agents

The Unity Machine Learning Agents Toolkit (ML-Agents) is an open-source project that enables games and simulations to serve as environments for training intelligent agents using deep reinforcement learning and imitation learning.
https://unity.com/products/machine-learning-agents
Other
16.93k stars 4.14k forks source link

General stacking of vector observations #3676

Closed dlindmark closed 4 years ago

dlindmark commented 4 years ago

Is your feature request related to a problem? Please describe. All vector observations are now stacked the same amount of times, even though the properties between them could be very different. For example could one observation change very rapidly (thus benefiting from more stacked samples) and another might change very slowly (needing much fewer stacks to capture the dynamics).

Describe the solution you'd like An option to specify how many times each scalar in the vector observation is stacked. I have not updated to mlagents 0.15.0 yet, and saw that you made big changes to vector observations. So I have no input just how to include this change.

chriselion commented 4 years ago

Hi, In theory, this should be possible using a StackingSensor wrapped around a VectorSensor. Unfortunately, I think I went overboard cleaning up the API and made StackingSensor internal, so you can't do this at the moment.

I've got this logged as MLA-792 in our internal tracker, and will try to get it cleaned up soon.

chriselion commented 4 years ago

The class was made public. Closing this issue.

github-actions[bot] commented 3 years ago

This thread has been automatically locked since there has not been any recent activity after it was closed. Please open a new issue for related bugs.