Edouard Leurent comments

Results 248 comments of


                                            Edouard Leurent

SAC + HER for "highway-fast-v0" or "highway-v0" environment application

Hi, HER is a replay mechanism which is specific to goal based environments (like parking-v0), where the objective is to reach a goal location. This is why it requires a...

How to use DDPG algorithm control the vehicle turn left in intersection env

Hi, Currently I have 3 action types: - continuous actions, there the agent control the gas pedal (longitudinal) and / or the steering wheel (lateral). If one axis is disabled,...

SAC

The reason is that SAC only handles continuous actions, and highway-env uses discrete actions by default. To configure it to use continuous actions, see [the docs](https://highway-env.readthedocs.io/en/latest/actions/index.html#id1).

Why ep_mean_reward and EP_mean_len in the environment training of the highway- DQN-CNN are NAN

This is not supposed to happen. One possible explanation is that your model maybe output a NaN at some point, and that NaN got propagated in the actions, then in...

Why ep_mean_reward and EP_mean_len in the environment training of the highway- DQN-CNN are NAN

This code (self.ep_info_buffer) is part of StableBaselines3 (the agents), not this project (the environment). Moreover, this result is not an error, it only means that the episode info is stored...

Monitor is not called in dqn_cnn.py

Do you mean in sb3_highway_dqn_cnn.py? There shouldn't be any more call to Monitor, since it has been removed from the latest versions of gym, replaced by separate wrappers to record...

Collision check implementation

Hey, sorry for the late reply. I think that would make sense yes, I just liked having all the vehicle state updates in one single place so they're easier to...

how to get the vehicle to derive collision information

Hi, (I am no professor,) You can check whether the controlled vehicle has collided by checking: `env.vehicle.crashed`

How to get the front and rear vehicles of ego-vehicle ?

Hi, This is a tricky question whenever we are dealing with a road network more complicated than a straight road (topogically). For example, if two lanes are merging into the...

Question about discreteMetaAction

The current `DiscreteMetaAction` has the following behaviour: to control the speed of the vehicle, the agent can choose between the actions `SLOWER `/ `IDLE `/ `FASTER`. This sets the target...