AWS Deepracer - Machine Learning

[Gizmotronn]

IrisDroidology/Python-Learning

AWS DeepRacer is an autonomous 1/18th scale race car designed to test RL models by racing on a physical track. Using cameras to view the track and a reinforcement model to control throttle and steering, the car shows how a model trained in a simulated environment can be transferred to the real-world. Find out more about AWS here: Amazon Web Services/Deep Racer

To find out our login details for Deep Racer, go to the Gitlab version of this issue.

[Gizmotronn]

So I've started the first part of this: https://console.aws.amazon.com/deepracer/home?region=us-east-1#getStarted

Creating Account Resources

This is currently one of my tasks set by @EXYNOS-999 - see #86/s.o.d. post

[Gizmotronn]

I'm now ready to start learning Reiterative Learning on AWS Deepracer

[Gizmotronn]

Some cool forks and repositories for machine learning:

• https://github.com/IrisDroidology/homemade-machine-learning
• https://github.com/IrisDroidology/awesome-machine-learning

This list will be updated periodically. Right now I'm working through the Introduction to Reinforcement Learning - https://d2k9g1efyej86q.cloudfront.net/

[Gizmotronn]

Check out https://github.com/EXYNOS-999/AWS_JPL_DRL/tree/documentation

[Gizmotronn]

I'm following the datacamp - http://campus.datacamp.com - tutorials and working on setting up a dual boot on my PC with Linux & Windows 10 home for the AWS JPL challenge. After major scares yesterday my dad is going to be getting some advice from the techies at his work (BHP Billiton) and we will set up something this weekend. Check out http://github.com/EXYNOS-999/AWS_JPL_DRL/issues/3

[Gizmotronn]

I'm doing a test 10-minute run on AWS Deepracer today, with the default reward function. You can check it out at IrisDroidology/Python-Learning on Github

[Gizmotronn]

def reward_function(params):
    '''
    Example of rewarding the agent to follow center line
    '''

    # Read input parameters
    track_width = params['track_width']
    distance_from_center = params['distance_from_center']

    # Calculate 3 markers that are at varying distances away from the center line
    marker_1 = 0.1 * track_width
    marker_2 = 0.25 * track_width
    marker_3 = 0.5 * track_width

    # Give higher reward if the car is closer to center line and vice versa
    if distance_from_center <= marker_1:
        reward = 1.0
    elif distance_from_center <= marker_2:
        reward = 0.5
    elif distance_from_center <= marker_3:
        reward = 0.1
    else:
        reward = 1e-3  # likely crashed/ close to off track

    return float(reward)

Test Reward Function

Links:

1. https://github.com/acord-robotics/stellarios/issues/88#issuecomment-570110930
2. https://console.aws.amazon.com/sagemaker/home#/jobs/dr-sm-rltj--20200102033006-acd52245-dbfe-4e86-af80-ad53ff548692

[Gizmotronn]

89

[Gizmotronn]

https://github.com/acord-robotics/stellarios/issues/89#issuecomment-570755208

90

[Gizmotronn]

New activity in Trello in card @ Board

by: --

Attachments:

URL: https://trello.com/b/jRkNkT4t/awsjplosrdrl//https://trello.com/b/jRkNkT4t

Content: