Edouard Leurent comments

Results 269 comments of


                                            Edouard Leurent

AttributeError: module 'tensorflow' has no attribute 'io'

Ah I see, this is related to you learning algorithm then, this repository only specifies the environment. However, you should be able to reach pretty good performance for the highway-fast...

how to reduce the time for env.step(action)

Hi @LianhaoYin, Here are a few ideas to make it faster: https://github.com/eleurent/highway-env/issues/223 See also: https://github.com/eleurent/highway-env/issues/36, https://github.com/eleurent/highway-env/issues/121

would like to observe the image of each agent keep the center constant when the observed car changes lane

Hi @LianhaoYin The center position of the rendered image is controlled here https://github.com/eleurent/highway-env/blob/3f569cfa766b454d545fcf048be3f7fb46f8a11a/highway_env/envs/common/graphics.py#L149 Right now, it uses the position of the observer vehicle (I just patched it because it used...

Multi-objective optimization

Well, this project only describes the simulation environment, and not the agent / learning process. It has been used in the context of multi-objective optimization, see e.g. https://budgeted-rl.github.io/ However, maintaining...

How to create an environment whihc is the same as the road structure in dataset, such as interaction?

Hi @Joe12138, I agree that this would be an awesome feature to have, but it is missing at the moment. This would require - writing or using an existing .osm...

How to avoid car following collision?

I don't really have a ready-made answer to that question, but there are several things you can try: 1. Use more appropriate network architectures, such as Transformers or CNN 2....

Multiple observations simultaneously

Hi, It is not possible at the moment, but it could be done by implementing a new MultipleObservation to automatically aggregate several observation types into a tuple.

Q: The training uses relative information and is normalized. How to get the distance travelled, lateral displacement, absolute velocity, absolute heading angle of the main vehicle at each time

Hi @limeng-1234 > The training uses relative information and is normalized This can be configured by two KinematicsObservation options: ``` absolute: False normalize: True ``` > How to get the...

def space(self) -> spaces.Space:error

I don't know, it should: https://github.com/openai/gym/blob/master/gym/spaces/space.py#L4 Do you have the latest version of gym installed? Try: `pip install gym --upgrade` and ``` > from gym import spaces > print(spaces.Space()) ```

List of Trained Models

Hi! That is a great idea, It requires a bit of work in running a well-chosen RL library with a stable version, so that results can be reproduced and the...