BRN - NASA

[Space Cadet]

Hi SC,

When you hot, you hot:

Blue Ridge Envisioneering, Inc. (BRE) proposes the development of a systematic and methodical approach to deploying Deep Neural Network (DNN) architectures on neuromorphic hardware and evaluating their performance relative to a traditional GPU-based deployment. BRE will develop and document a process for benchmarking a DNN’ s performance on a standard GPU, converting it to run on near-commercially available neuromorphic hardware, training and evaluating model accuracy for a range of available bit quantizations, characterizing the trade between power consumption and the various bit quantizations, and characterizing the trade between throughput/latency and the various bit quantizations. This process will be demonstrated on a Deep Convolutional Neural Network trained to classify objects in SAR imagery from the Air Force Research Laboratory’s MSTAR open source dataset. The BrainChip Akida Event Domain Neural Processor development environment will be utilized for demonstration as it provides a simulated execution environment for running converted models under the discrete, low quantization constraints of neuromorphic hardware.

This 2020 proposal was intended to use MetaTF to demonstrate how Akida performed at various bit samples (1, 2, or 4-bit weights and activations) trading speed and power for accuracy.

This was about the time that BRN developed the multi-bit Akida, and would have had the simulation software (MetaTF) ready in advance.

Given the target audience, the performance data would be classified.

PPPPS: Oh sh!t, the helicopters agai...

I don't know why but I keep looking through the curtains. Keep getting this strange feeling someone is watching.

SC

 

[Diogenese]

I don't know why but I keep looking through the curtains. Keep getting this strange feeling someone is watching.

SC

He's behind you!

 

[stuart888]

With the Fact Finder Finding (FFF) Nasa's newest division: Silicon Space Technology Corporation, there is some detail about them. Contracts Awarded and the money involved too. It appears to be a summary of activities specific to the Silicon Space Technology Corporation. There are a lot of things going on for sure. Notice below it says 16 people, plus it has the link to VoragoTech.com.

https://www.sbir.gov/sbc/silicon-space-technology-corp

 

[Vanman1100]

Wonder if nasa pointed prophesee in our direction?

 

[Deleted member 118]

Another use case for Akida

NASA SBIR 2022-I Solicitation | S17.04-1526 - Improving Data Assimilation in Numerical Weather Prediction of Winter Cyclones using Artificial Intelligence and Machine Learning | Proposal Summary

 

[Deleted member 118]

Artificial Intelligence (AI) | SBIR.gov

www.sbir.gov

Application of Artificial Intelligence for Science Modeling and Instrumentation | SBIR.gov

www.sbir.gov

Automated Artificial Intelligence (AI) Distress Alerts and Monitoring | SBIR.gov

www.sbir.gov

NASA SBIR 2022-I Solicitation | A1.06-2440 - Automated Machine Learning Prediction of Motor Degradation (AMP MD) for eVTOLs | Proposal Summary

Navy - 22.2 SBIR - AI/ML for Additive Manufacturing Defect Detection

DIGITAL ENGINEERING - Artificial Intelligence /Machine Learning Applications to STANDARD Missile Maintenance Data | SBIR.gov

www.sbir.gov

Agile Machine Learning in Dynamic Environments for Complex Event Processing | SBIR.gov

www.sbir.gov

 

[Deleted member 118]

Application of Machine Learning, Artificial Intelligence, and Data Science techniques to improve NL Data Management and Application Services | SBIR.gov

www.sbir.gov

Release Date:
April 20, 2022
Open Date:
May 18, 2022
Application Due Date:
June 15, 2022
Close Date:
June 15, 2022
Description:

OUSD (R&E) MODERNIZATION PRIORITY: Artificial Intelligence/ Machine Learning; Cybersecurity; Nuclear



TECHNOLOGY AREA(S): Information Systems; Nuclear



OBJECTIVE: To develop proof of concept and demonstrate feasibility for a customized application development platform and database management integrated with intelligent, self-learning cybersecurity control monitoring system prototype. As envisioned, the capability will use industry leading technologies, such as Machine Learning (ML), Artificial Intelligence (AI) and Data Science (DS) capabilities, to recognize normal usage patterns for Nuclear Logistics management and reporting application, then apply data analytics and reporting to identify, report and suspend anomalous activities that may represent a cybersecurity threat or operating environment deviation from normal. NL goal is to develop customized applications and database management system using no-code or low-code application environments and commercial of the shelf database software. While these tools incorporate proven components implementing secure coding best practices, we expect incorporating AI/ML/DS technologies and analytics with the low-code environment will enable developing more intelligent monitoring capabilities directly within the applications. The amount of data inherent within custom Nuclear Logistics applications is relatively small by contemporary standards, but usage data from module and database field access is orders of magnitude greater and significantly more dynamic. Automatically incorporating big data analytics and reporting capabilities into low-code developed applications to warrant Surety and Accountability of the nation’s deterrence stockpile. AI/ML/DS capabilities will be integrated into application usage and database access patterns to learn “normal” expected behavior and quickly identify, alert and block anomalous actions.



DESCRIPTION: Nuclear Logistics Technical Division (NLT) is directly responsible for providing application and database services to the DoD customers as part of their mission to maintain the national nuclear stockpile databases. NLT requires next generation systems that will provide the capabilities to securely maintain the integrity of mission essential applications and databases by applying cybersecurity protections through the application AI/ML/DS technologies. To support future requirements and nuclear system modernization, NLT is interested in developing a next generation rapid application development platform and database management prototype integrated with intelligent cybersecurity protection measures. NLT’s goal is to prototype and demonstrate an enhanced NL Application development platform and Database Service using AI/ML/DS and low-code generated custom applications to support robust and secure user applications for NL needs. The application development platform will be configured to use COTS no-code or low-code platforms, and deliver the prototype system using innovative which can effective make use of AI/ML/DS techniques for both user application data and system collected usage and access patterns. As the Restricted Data aspects of the NLT information systems necessitate a cryptographically isolated environment, this prototype system will leverage AI/ML/DS capabilities within a relatively closed environment to support enhanced analytics, monitoring and reporting for Nuclear Logistics and enhance nuclear logistics management capabilities and system modernization as the foundation for future applications beyond current capabilities. This proposal allows development teams to become part of a paradigm that introduces an end-user application design that creates a fundamental connection between the user, mission essential tasks, business operations procedures and the software that directly impacts department specific productivity. This innovation also facilitates a custom application design from idea to an operational reality within a few weeks, perhaps even days.



PHASE I: Design a concept for enhancing low-code application development platforms to automatically incorporate capabilities for reviewing all logged system, module and database access events, identifying patterns representing “normal” system usage over a period of time, which will account for monthly, quarterly and annual data entry and reporting activities, as well as ad hoc activities, such as emergency operations exercises. Using this growing collection of event data, the low-code application development platform enhancements will continually monitor the cyber-health of the application environment using AI/ML/DS services and alert system administrators as to anomalous conditions. If possible, the AI/ML/DS services will also be able to identify and alert operational anomalies, such as network, storage or processing issues before actual system failures occur. In addition to identifying concepts and methods for accomplishing these intelligent system monitoring capabilities, analyze additional processing, storage and network loads to support these capabilities. The analysis should include identifying typical cybersecurity monitoring services and controls that may be able to be replaced by the intelligent system monitoring capabilities.



PHASE II: Design, develop, demonstrate and validate a proof of concept technology demonstration for one or more representative modules incorporating the AI/ML/DS services for enhanced intelligent monitoring in parallel with developing the new module(s). Provide a detailed project plan for developing such a capability, and benchmarking against a similar complexity low-code developed module to quantitatively and qualitatively identify specific operational improvements. Conduct formal acceptance testing with user community stakeholders on the module to confirm its correct operation, then monitor the ability to enhanced capability to collect system events, identify normal usage patterns, and alert when repeated anomalous events occur outside the normal range. Once fielded to a select group of operational users, test the system effectiveness for intelligently adapting to changing usage events by introducing specific anomalies. Document the comparison of the representative baseline module against the enhancement intelligent monitoring module, including benefits, weaknesses and opportunities for improvement. Prior to completion of Phase II, perform typical software maintenance and functional enhancements, as would be typical of user reported problem reports and new / improved functionality change requests, and analyze how the intelligent monitoring capabilities respond to a modified software codebase.



PHASE III DUAL USE APPLICATIONS: Working with NL and the low-code application development platform software vendor, jointly develop functional requirements, specifications, high level designs for incorporating the AI/ML/DS techniques, services and capabilities. Document lessons learned from the Phase II proof of concept capability and identify suggested improvements. Identify changes necessary to generalize the AI/ML/DS service enhancements for the COTS low-code application development platform, including approximate estimates of complexity to accomplish. For NL, identify necessary actions to incorporate the intelligent monitoring capabilities into existing low-code based applications. Also identify actions necessary to extend the AI/ML/DS based intelligent monitoring capabilities from being able to alert system administrators regarding anomalous conditions to enhancements necessary for the intelligent cyber monitoring capabilities to proactively take actions to automatically shut down identified threats. It is assumed identified network, storage and system identified anomalies will need to be manually addressed by system administrators at the end of Phase III. If there are opportunities for commercializing corrective actions in response to system failure modes identified, these should be documented and communicated to NL and the COTS low-code application development platform vendor as well.

 

[Deleted member 118]

Do we have any link with this company, or do they use there own AI?

GALT Awarded Phase II SBIR Contract by the USAF

 

[Space Cadet]

I think this is at the centre of everything Nasa. IMO this is what all the other things revolve around. Nasa are going to build a gateway to the moon and Mars and a base on the moon starting with Artemis sometime in next two years. Everything will be done autonomously and I believe we are in the thick of it. All printed in space and modular in design. I posted an article in last couple of days of mission on May 25 I think, where 2 cubesats autonomously joined in space. It was a test case imo. It would not surprise me at all if we were onboard. Dot joining I know but the more I read the more I think we will surely be ubiquitous as we will be in everything in Space as well.

https://www.sbir.gov/node/1836297

SC

 

[stuart888]

Since Rob Telson stated clearly: "With Nasa helping them get into Orbit", Brainchip has passed up Jeff Bezos Blue Origin and Sir Richard's Virgin Galactic! Brainchip Akida is Orbital not just Suborbital! Nasa is using Brainchip for the farthest-reaching space means a lot.

We may need to charge more for applications that go farther. The fee would go way up for a 10-year cubeSat interstellar voyage? This nasa relationship is huge for us. 1000 eyes is showing it all the time with the detailed research.

 

[stuart888]

Suborbital space does not require the same level of Radiation Harding. I cannot get over that Rob said into Orbit, and since the Radiation Harding project has been going on for a nicely long time. To me, Akida has been to Space and worked in Radiation Space Flight. Instead of him saying in Orbit, he could have said helping Nasa test Akida in Radiation Space Flight conditions.

Are we allowed to ask the Brainchip Staff: What did Rob mean with Orbit? Has Akida ever been used in real orbital radiation conditions? Not test radiation in the lab.

The Space Review: Human factors in commercial suborbital flight: Radiation exposure and weightlessness

 

[jwoody2000]

not sure where to post this query, and it may have been thought of already. Nasa using fast rockets seemed the go......
with senors ...and the tracking of "All of the intercontinental ballistic missiles in the world’s nuclear arsenals are hypersonic, reaching about 15,000 mph (24,140 kph), or about 4 miles (6.4 km) per second at their maximum velocity."
having seen on the ABC it is problem in that air defences can not detect this to defend against as by the time they are aware its too late, and if carrying a war head it will cause the other side to react in kind..hence nuc war.
Can BRN IP track at these speeds?
Question of if it can would one want the world to know you can defend against ballistic missiles, the missiles said to be already used against the urkarians.
if the defence tracking IP is kept as a secret incase (the west with something up their sleeves,)... a progress to nuclear war or even expanded urkaraine where putin (or russia) will use many more ballistic missile (non nuclear vision) in much more volume regardless....
to that the way China is using cyber war/hacking info..will they already have this defence themselves ...or if not, not move to tawain invasion until they have the IT and IP... (and not tell the world either), this point about china having our IP already, stolen or otherwise has been raise in several threads in years past.... with our patents not register in all areas in the world.... some might even recall akida band in Africa...but that wasn't us.

 

[Rise from the ashes]

not sure where to post this query, and it may have been thought of already. Nasa using fast rockets seemed the go......
with senors ...and the tracking of "All of the intercontinental ballistic missiles in the world’s nuclear arsenals are hypersonic, reaching about 15,000 mph (24,140 kph), or about 4 miles (6.4 km) per second at their maximum velocity."
having seen on the ABC it is problem in that air defences can not detect this to defend against as by the time they are aware its too late, and if carrying a war head it will cause the other side to react in kind..hence nuc war.
Can BRN IP track at these speeds?
Question of if it can would one want the world to know you can defend against ballistic missiles, the missiles said to be already used against the urkarians.
if the defence tracking IP is kept as a secret incase (the west with something up their sleeves,)... a progress to nuclear war or even expanded urkaraine where putin (or russia) will use many more ballistic missile (non nuclear vision) in much more volume regardless....
to that the way China is using cyber war/hacking info..will they already have this defence themselves ...or if not, not move to tawain invasion until they have the IT and IP... (and not tell the world either), this point about china having our IP already, stolen or otherwise has been raise in several threads in years past.... with our patents not register in all areas in the world.... some might even recall akida band in Africa...but that wasn't us.

Hypersonic

https://www.navalnews.com/naval-news/2021/06/u-s-missile-defense-agency-new-concept-to-counter-hypersonic-threats/ U.S. Missile Defense Agency New Concept to Counter Hypersonic Threats The U.S. Missile Defense Agency is developing a concept to protect the US, its deployed forces, and allies...

thestockexchange.com.au

 

[Rise from the ashes]

also we have sbir with quantum ventura and further digging see them working on hypersonic detection if i remember correctly.

Quantum Ventura

www.quantumventura.com

 

[jwoody2000]

amazing research. the ABC reporter needs to log in here to be up to date.
great to see, as i quickly scroll through that some are close to developing that defence of ballistic or hypersonic... as there is a difference in deployment in both. there was some mention about cloud, dust, sun or moon light reflection ... as barriers to detection and that still leaves me with queries about both deployment and detection in extreme conditions. deployment would be easily with ground plotting radar (mapped landscape identifiers- great for Akida adaption)...or it is called something else...and this in conjunction with GPS to find point pointed targets. ..but sorting out heavy clouds, and or a cover of electrical storm seems would be harded; there is said to be radar (or other tracking) of Submarines ...so that would sort of make the use of Subs redundant would it not for the deployment in better ocean positions.
im sure the query answer will be within your library of infor, and the better minds behind those that work in that field.

i hope that what ever IP is used in defence that there are other backup or two working on a statelite as well, and not those that make those choice don't get caught up in ego of ...just having slected "one" ...as the one might be us or someone that has influence at ministral level...easy to convince someone that 'ours' (whoever ours is ) is the one and only and best. ...and hand over the money with a money back refund if it doesn't work.

 

[Rise from the ashes]

 

[Deleted member 118]

So this SBR was suppose to finish in jan this year

Spiking Neural Network Modeler (SpiNNMo) | SBIR.gov

www.sbir.gov

But I just came across a new job for the company originally offered the contract.

Scientist/Senior Scientist- Neuromorphic Computing - Monrovia, CA 91016 - Indeed.com

ChromoLogic, LLC

www.indeed.com

So maybe this research has gone into stealth mode as they ain’t be rewarded anything since

CHROMOLOGIC LLC | SBIR.gov

www.sbir.gov

I wonder if we are working with them

Deep Neural Net and Neuromorphic Processors for In-Space Autonomy and Cognition (SBIR) | SBIR.gov

www.sbir.gov

 

[equanimous]

neural network-based flight: Topics by Science.gov

 

[Diogenese]

This article describes some of the work NASA is doing with CubeSats, using SNNs to manage the communications between a flock of CubeSats.


CCAAW_21_final.pdf (nasa.gov)

https://ntrs.nasa.gov/api/citations/20210014036/downloads/CCAAW_21_final.pdf

Applying the Cognitive Space Gateway to Swarm Topologies
Rachel Dudukovich, Adam Gannon, and Robert Short ID
Communications and Intelligent Systems Division, NASA Glenn Research Center

Ricardo Lent, Senior Member, IEEE, University of Houston, Houston, Texas, USA

The paper describes the results of a computer simulation using a SNN to learn the optimal communication paths when one of the inter-satellite links (1067 - 1061) is a bit dodgy.


Introduction

This paper outlines current efforts to apply the Cognitive Space Gateway (CSG) [1] to future NASA Cube Satellite (CubeSat) and/or SmallSat missions. The CSG is a routing algorithm developed for intermittently connected or delay-tolerant networks (DTN) that uses a spiking neural network to make optimal routing decisions in a multi-hop network. It is envisioned that technologies such as the CSG will enable efficient operation of increasingly complex networking scenarios.

...

Time-series Observation

The CSG method uses path latency estimations calculated from single[1] hop performance metrics that allow modifying the synapse weights of the SNN and achieve continual learning. This feature is demonstrated by the lower bundle latency that is achieved with the CSG approach compared to the shortest ETA attempt. Because the CSG uses exploration, which is a necessary mechanism for learning, a certain percentage of the bundles may experience larger delays than the rest. The delay peaks that can be observed with the CSG occur as a result of the use of the high-packet loss link or due to the intra-flow network congestion. In the latter case, the delay is caused by the wait time for other bundles that already occupy the buffers waiting to be transmitted. It can be observed that the peaks tend to decrease over time, which is an indication of the learning effectiveness of the system to achieve low end-to-end bundle latency.


Disclaimer: Any similarities between the spiky orange boxes and any living orange spikey boxes is purely coincidental.

The blue lines in Fig 5a, starting high and decreasing, show how the SNN learned the optimal path compared with an alternative method of path selection ETA.

It is worth noting that multiple 2-hop paths are possible between the source and sink with the selected topology, but only one of them was affected by the larger packet loss ratio in the experiments. No attempts were made to balance the traffic among the 2-hop paths in the reference case, whereas a load[1]balancing effect over paths of any size results as a side-effect of the CSG. For the same experiment, Fig. 5 (b) depicts the path length taken by each consecutive bundle. With CSG, the path affected by the higher loss was used about 5% of the time and those instances have been highlighted in the figure using green diamond-shaped markers. The low usage count for this path suggests that the CSG learned to avoid that route.

As this was a computer simulation, it probably didn't use Akida 1000 SoC, but
...
there is always MetaTF.

 

[Deleted member 118]

https://techport.nasa.gov/datasheet/113116

What’s direct NN @Diogenese

https://www.softinway.com/wp-content/uploads/2022/06/Axial-Compressor-Map-Generation-Leveraging-Autonomous-Turbo-Expo-2022-Proceedings.pdf