GitLab

Refactoring: delete new_position from TreeObsForRailEnv and use flatland.core.grid.grid4utils.get_new_position instead

local grid observations will not be updated. It was never used for the challenge and is not well suited for the task at hand.

now detecting malfunctioning agents
now detecting slowest fractional speed

fixed formatting

now detecting malfunctioning agents
now detecting slowest fractional speed

and render custom obvsercations

Now turns with the agent, encodes local information which can easily be enhanced by predictions, distance map and other things.

Now turns with the agent, encodes local information which can easily be enhanced by predictions, distance map and other things.

Agents now get warned about conflicts when other agents have reached their target and are in the way.

ALso reinstated the dead-end branching to keep trees smaller

--> This might be a problem when trees become very large due to the lack of switches...

Updated test for generators to check that distance map is not recomputed when loaded correctly
fixed formatting and removed debug prints

Updated test for generators to check that distance map is not recomputed when loaded correctly

Updated test for generators to check that distance map is not recomputed when loaded correctly

removed prints for debugging.
updated notebooks to new rendering code --> Tests should pass
Updated formatting

removed prints for debugging.
updated notebooks to new rendering code --> Tests should pass

Additional package msgpack-numpy needed for ndarray.
This saves tons of time when loading precomputed files.
Updated test to test loading and saving with and without distance maps
minor update to bug in observation dimension
fixed typos

Additional package msgpack-numpy needed for ndarray.
This saves tons of time when loading precomputed files.
Updated test to test loading and saving with and without distance maps
minor update to bug in observation dimension

Additional package msgpack-numpy needed for ndarray.
This saves tons of time when loading precomputed files.
Updated test to test loading and saving with and without distance maps
minor update to bug in observation dimension

Additional package msgpack-numpy needed for ndarray.
This saves tons of time when loading precomputed files.
Updated test to test loading and saving with and without distance maps
minor update to bug in observation dimension

Additional package msgpack-numpy needed for ndarray.
This saves tons of time when loading precomputed files.

It now sets all the info from pickle file before running all the prediction and observations.