Eugene Vinitsky

Negative length value for object

4

Hello, we were examining tf-record 00008-of-01000_364 and we found a vehicle that seems to have a "length": -1.295910358428955. Is it correct to just take the absolute value of this? Do...

Prevents wraparound of lines in the time space diagram for ring and figure eight

bump the high of traffic light grid po env

8

Change the observation space to handle the fact that values go slightly outside of it.

Done fix

1

The done condition is not correct for sims_per_step greater than one. Its possible a vehicle might have left during one of the steps. Additionally, fix the done condition to get...

We should have a flag that can be used to suppress all of the warnings so that the output is legible. It should NOT be the default because warnings that...

Add an implementation of "Learning Multi-agent Communication with Backpropagation" to Flow

10

It would be interesting to port a few basic communication environments/training procedures to Flow. In particular, a popular communications baselines is "Learning Multi-agent Communication with Backpropagation" (https://arxiv.org/abs/1605.07736). As a request...

## Pull request information - **Status**: ? (ready to merge / in development) - **Kind of changes**: ? (bug fix / new feature / documentation...) - **Related PR or issue**:...

Fix visuallizer bug in rllib and add more logging to train.py

If noise parameter isn't multiplied by sim_step, the user has to re-scale it so that it is invariant to sim-step size.