Location
US
Badges
Activity
Challenge Categories
Challenges Entered
Completed
MineRL Labs
Learning From Human-Feedback
Latest submissions
Carnegie Mellon University
Arrival
The first, open autonomous racing challenge.
Latest submissions
See All| graded | 176756 | ||
| graded | 176754 | ||
| failed | 176753 |
| Participant | Rating |
|---|
| Participant | Rating |
|---|
Learn-to-Race: Autonomous Racing Virtual Challenge
Open discussion on some tricks we can use to work towards solving this task!
Almost 4 years agoI wanted to discuss certain ideas on how to approach solving this challenge, anyone wants to chip in they are more than welcome to discuss on top of this -
- Combining imitation learning and reinforcement learning. One of the things I am thinking about is possibly pretraining the model using images from racetracks captured during the run of the simulator. Using this the agent can gauge the upcoming environment and probably stay on track. Multiple ways of feeding this to the agent could be based on encoding the images in such a way that’s easily interpretable for the setup.
- Knowledge distillation from multiple sensor inputs. Because the evaluation procedure only allows for a subset of the sensors that can be used for training an agent, it could be worth exploring best practices for transferring knowledge from the training sensors to the testing subset, e.g., through pre-trained encoder models.
- Another thing I am planning to try during my run with this challenge is comparing the usage of On-policy vs Off-policy agents also possible comparing and contrasting the two with the base Soft Actor-Critic being off policy an On-policy algorithm like PPO could be an interesting approach to try for this challenge.
- Also, we need to keep in mind the second part of the challenge as well which consists of the Safety aspect of the model we are using. What would be interesting is how the points mentioned above could be useful for the safety part of the challenge as well. Also to come up with a specific model that focuses more on the safety rewards would be interesting and some sort of ensemble of the fast model and the safety model could be the solution.
Camera configuration on evaluation server
Almost 4 years agoHi @bernd_heihoff,
You are allowed to use any camera configuration for training purposes.
So you can use the cameras for pertaining and learning the weights for your model.
During the evaluation, it will only be a single camera predecided from our end that is going to be used.
So no need to worry about the evaluation portion those configs are decided from the backend.
Docker Image - Documentation
Almost 4 years agohttps://www.youtube.com/watch?v=W6WdWrB10g4 here’s a walkthrough for setting up the resources for the challenge.
Error in starter code
Almost 4 years agoThanks for pointing this out will be fixed and pushed to the starter_code soon.
Docker Image - Documentation
Almost 4 years agoHey,
We are checking the version provided to you for the docker image.
Meanwhile, to get you started you can run it locally if you have a Linux setup or you can use a virtual machine using AWS.
For that, it’s easier just to run the bash script in the LinuxNoEditor folder. Keep this running in the background.
Also sharing a template for the controller_kwargs = {
“sim_version”: “ArrivalSim-linux-0.7.0-cmu4”,
“quiet”: True,
“user”: “ubuntu”,
“start_container”: False,
“sim_path”: “–path–/ArrivalSim-linux-0.7.0-cmu4/ArrivalSim-linux-0.7.0.182276/LinuxNoEditor”,
}
Camera configuration on evaluation server
Almost 4 years agoHi @bernd_heihoff,
We had a discussion among the organizing committee we will be allowing the use of three cameras for the evaluation phase for now. To add the configuration -
config.py - active_sensors = [
“CameraFrontRGB”,
]
Add names of the other cameras here in the config file.