This is great! Do you have an idea about when subgroup meetings will start?
1 Like
@martin - are you having a mtg tomorrow? @odylec pinged me thinking a meeting is occuring of this subgroup on Wed Feb 11th?
1 Like
He told me that he dont have connection “AI/ML AWG meeting? Unfortunately, I will be in a location without signal during the meeting.”
1 Like
I think he said he was unable to attend the ‘main’ monthly AWG mtg which took place earlier today
Hi Brian,
I do not expect a telemeeting before March. The brainwriting is currently ongoing. I will remind everyone about it again below.
1 Like
For those who may have missed it, I would like to point out that the first phase of the Brainwriting is currently underway.
Please review the document and add your proposal directly into it. The rules are described in the document, and a link to a real data sample is also included there.
The first phase of the Brainwriting should end on February 14, unless someone requests an extension. After that, either I or @odylec will provide instructions on how we will proceed with the second phase.
Thank you for your attention.
Martin
P.S.: If you have any questions about the Brainwriting, please post them directly here.
1 Like
Thank you for this question.
The radiation field aboard a spacecraft is highly complex, and any detector can measure only a portion of this field, while the remaining part must be reconstructed through calculation. A major issue is the passages through the Van Allen radiation belt, during which the radiation environment changes significantly and the measured data must be processed differently. These regions are typically identified using geomagnetic field models, but the belts are not stable over time.
The objective of this research is therefore to develop a method capable of identifying radiation belt crossings directly from the measured data. This would enable analysis without direct dependence on geomagnetic models.
1 Like
Thanks for the answer, Martin
The measurements will surely provide an equilibrium or mean position of the radiation belt crossings, about which the crossings fluctuate.
1 Like
Thanks for preparing this data, Martin. It’s been the perfect testbed for some benchmarking logic I’ve been developing.
I’ve submitted a proposal to the Brainwriting Doc based on this. I’m currently using your subset to build an ‘Ink-Stinct Challenge’ on Kaggle to evaluate how different LLMs reason through these radiation spikes.
You can see the dataset and the logic I’ve built so far here:
-
Dataset: https://www.kaggle.com/datasets/gastondana/spacedos
-
Referee Logic: https://www.kaggle.com/code/gastondana/squidsaa-ink-stinct-level-1-referee
Once the Kaggle team patches a minor indexing issue I found, the full benchmark will be ready for the group to test shortly after.
Thx!
2 Likes
Yes, that could describe the temporal shift of the center of gravity of the SAA. From a dosimetry perspective, however, it is also crucial to define the boundaries of the SAA, that is, to determine from which point to which point the radiation field must be evaluated using a different methodology.
If we were able to determine these boundaries dynamically, it would also be highly valuable for identifying the onset of SPEs and could likely be applicable to missions into deep space as well.
1 Like
Perfect, we will stay tuned!
1 Like
Here’s my PoV:
- We can analyse Radiation time-series data using a sequence machine learning model to detect points where radiation patterns change significantly.
- These change points can then define the dynamic entry and exit boundaries of the SAA.
- This as a result will enable accurate switching of dosimetry methods and also helps in early detection of radiation events like Solar Particle Events, making it useful for both Earth orbit and deep space missions.
- We can then have heatmap based visuals to define entry exit boundaries
Is my understanding correct. Here’s a sample Dashboard view on sample data to define utility of Research Outcomes.
We can include all the ideas to the word file to compare them
4 Likes
Just a reminder that February 14 is the deadline for submissions to the brainwriting document.
4 Likes
As I received several last-minute questions about how to complete the brainwriting document, I am extending the deadline by two days.
Please try to enter your proposals directly into the shared Google document. If you are unsure about any specific paragraph, simply leave it blank.
6 Likes
Hello Working Group,
We currently have seven proposals for data processing:
- Interactive 3D Radiation Mapping (WebGL / Niivue) – Visualization
- Physics-Informed XGBoost with Geomagnetic Context – ML Model
- Hybrid Two-Stage SAA Detector – ML Model
- Multi-LLM Benchmarking & Automated SAA Separation – Benchmarking
- Genomic-Inspired Anomaly Scoring – Feature Engineering
- LSTM Autoencoder Anomaly Detection – ML Model
- Graph Neural Network for Orbital Flux Topology – ML Model
@odylec will now clean up and unify the formatting of the proposals so that they are easier to work with and so that everyone can add comments and suggestions.
This will be an open commenting phase lasting about 14 days after the AI/ML AWG meeting next week. After that period we will proceed with the evaluation of the proposals.
Stay tuned.
Martin
6 Likes
Very interesting topics, done a small background review on SAA specific space radiation at the International Space University (MSS) and currently in contact with ESA EAC and DLR where radiation simulation / radiobiology on the table. Happy to see if any common synergies. Please send a PM.
1 Like
Thanks for your interest. Since this forum is intended for open exchange of ideas and data, it would be great if the discussion could continue here so others can benefit from it as well.
If there are details you cannot share publicly, feel free to send me a PM.
2 Likes
Thanks Martin. I am aware of the prupose of the forum that I also support. To clarify, my intent was merely expressing interest. And any in-detail conversation on how, in what capacity or whether I may join in would be a sort of dialoge happening between those directly involved with specific workflows at first instance, in my view – that’s what I am comfortabe with. In particular, at the stage of exploring whether a good fit for personal aspirations. Obviously, any publicly relevant outcome from these conversations – and for sure all adequate data items – to be in the public domain (as long as no regulatory obstacle) to support open scientific endeavours.
2 Likes
I’m likely behind in the discussion, but the ‘additional files’ under Session 9 in the 26th WRMISS page have interactive plots that may be useful in visualizing some of the parameters affecting SAA isolation. Download and unzip the html files to open them locally for the plots to display properly, but once open in the web browser, you can use the select tool at the top right of the pages to select any areas in the plots (except the LET plot) to see how the remaining plots are affected by the selection. There is also a radio button in the top left to see how ascending and descending passes affect the various components of the field. The plots are from the HERA instrument on ISS with the associated presentation here.
3 Likes