For this challenge, we invite you to become "virtual contributors" to the Asteroid Grand Challenge and develop a hypothetical method, concept note or simple prototype that demonstrates how Machine Learning could be used to help us avoid the same fate as the dinosaurs.
Model using NASA NEO data to produce a 3D model modified as the Atari 2600 Game “Asteroids” which Google Deepmind uses to target/ confirm NEOs
Business Requirements Document (BRD) NASA NEO BRD Concept Note
NEO-AI
April 2016
Version 1.1
NASA
Acknowledgements:
1.Ben Noble (PA Consulting- SpaceApps London DataCamp 2016)
2.Shaun Moss (Author of The International Mars Research Station)
3.Dr Demis Hassabis (Google Deepmind AI)
4.Ed Rex (Jukedeck AI)
5.
Date |
Version Number |
Document Changes |
23/04/2016 |
0.1 |
Initial Draft |
Role |
Name |
Title |
Signature |
Date |
Project Sponsor |
NASA |
|||
Business Owner |
NASA |
|||
Project Manager |
Brett Gallie |
|||
System Architect |
Brett Gallie |
|||
Development Lead |
Brett Gallie |
|||
User Experience Lead |
Brett Gallie |
|||
Quality Lead |
Brett Gallie |
|||
Content Lead |
Brett Gallie |
To develop a hypothetical method, concept note or simple prototype that demonstrates how Machine Learning could be used to help us avoid the same fate as the dinosaurs.
This Challenge is part of the 2016 NASA SpaceApps Challenge.
Near Earth Objects can have potentially catastrophic(E.L.E) impacts or huge economic (ie.insurance) impact
Near Earth Objects can also be harvested for resources in the future by asteroid mining Companies this could be harnessed as a funding resource for cash strapped research projects and related NEO finding affiliates.
Existing list of NEO’s have been verified
Jupiter / our moon will take care of 90% of the NEOs
Our data may only go as far back as the 1900’s. The Ancient Chinese kept records dating back for 3,000 years and should be accessed for historical impacts or observations that may be useful.
http://idp.bl.uk/4DCGI/education/astronomy/history...
Unidentified NEOs reaching earth – ie. We missed the observation
An existing NEO’s projected trajectory is altered by Solar Flare or other space phenomina bringing it into collision with earth
The Key Issue is to ensure our 3D model of NEOs is as accurate as possible for the Deepmind AI to target the NEOs.
Data from verified NEO objects are handled by
Repository of all located NEOs and NEAs: http://www.minorplanetcenter.net
1.Data from the Minor Planet Center to be fed into a 3D simulation model
2.The simulation model will resemble a modified Atari 2600 game called “Asteroids” and the 3D model would resemble existing apps. Eg: GoSkyWatch App
3.Additional data will be fed in from an App similar to Seti- at- Home which amateur astronomers NEO observations are electronically sent through by the app for confirmation
4.A modified version of Google Deepmind’s AI machine learning will be used to target NEO’s.
A list of Google Deepmind’s approaches can be found at the following site:
https://deepmind.com/publications.html
The requirements in this document are prioritized as follows:
Value |
Rating |
Description |
1 |
Critical |
This requirement is critical to the success of the project. The project will not be possible without this requirement. |
2 |
High |
This requirement is high priority, but the project can be implemented at a bare minimum without this requirement. |
3 |
Medium |
This requirement is somewhat important, as it provides some value but the project can proceed without it. |
4 |
Low |
This is a low priority requirement, or a “nice to have” feature, if time and cost allow it. |
5 |
Future |
This requirement is out of scope for this project, and has been included here for a possible future release. |
Req# |
Priority |
Description |
Rationale |
Use Case Reference |
Impacted Stakeholders |
General / Base Functionality |
|||||
FR-G-001 |
1 |
3D Simulation model created of the NEO data using Big Data crunching applications (eg.Hadoop) |
3D Simulation model will provide background for Atari 2600 “Asteroid” game for Google’s Deepmind to process |
Development teams Infrastructure engineers |
|
FR-G-002 |
1 |
Google Deepmind’s AI will “target” NEO’s generating a database of “hits” indicating the NEO has a high probability of earth impact |
Deepmind’s AI uses games to refine its search and targeting method with a high level of accuracy |
||
FR-G-003 |
|||||
FR-G-004 |
|||||
FR-G-005 |
|||||
Security Requirements |
|||||
FR-S-001 |
1 |
NEO report findings should be encrypted and sent to NASA |
|||
Reporting Requirements |
|||||
FR-R-001 |
2 |
“Hits” detected by the AI should generate a detailed report which astronomers should be able to verify |
|||
Usability Requirements |
|||||
FR-U-001 |
1 |
Data generating the 3D model should be transferred into a format that can be used by the AI |
|||
Audit Requirements |
|||||
FR-A-001 |
1 |
A full audit report of AI activity needs to be generated after each session |
ID |
Requirement |
NFR-001 |
Multiple nodes of the AI program can run the simulation at the same time and overnight as the data is updated with new observations |
NFR-002 |
|
NFR-003 |
|
NFR-004 |
|
NFR-005 |
NEO – Near Earth Object
ELE- Extinction level event
Google Deepmind - Google DeepMind is a British artificial intelligence company founded in September 2010 as DeepMind Technologies. It was renamed when it was acquired by Google in 2014.
Near Earth Object Machine Learning
Asteroid Mining
The abovementioned is available as a word document with images
Opensource Pygame example of Atari 2600 "Asteroids"
JukeDeck's AI to provide a soundtrack for the video
Stupeflix to customise the video
Resources on GitHub provided by NASA