Aerie/PlanDev integration with F' and analysis tools

[barugonman]

Overview

Looking to build up a 'reference mission' style demonstration of an end-to-end robotic system including Aerie/PlanDev, the F' (F Prime) robotic control system, and downlink analysis tools built largely on M20 and Clipper.

The choice of F' as the robotic command-thru-telemetry system was based off of recommendations and early prototyping of the F' system and interfaces.

The Clipper developed tools included the concept of 'close the u', where planning data is used in the evaluation of downlinked telemetry to assess health and performance of a robotic system such as a spacecraft. Some work was started on Clipper to complete integration of Aerie with Close the U tools State Manager and the State Data Store, however that was shelved and not completed.

Will update this discussion as we make progress through the prototype.

Goals

Implement a demonstration starting with Aerie activities and simulations, with related command expansions that are delivered as sequences to F'. F' produces the telemetry to be ingested in the State Data Store.

Additionally, Aerie simulation data is exported as state timelines into the State Data Store, and state definitions in the State Manager. This data is also saved to the State Data Store, for query in tools like dtat-library and EAS Channel Viewer.

Links

F':
https://github.com/nasa/fprime

Software Analysis Tools:
https://github.com/orgs/NASA-JPL-Spacecraft-Analysis/repositories

[mrmuszynski]

@barugonman Have you thought much about what your reference mission might be? Maybe you're using the word robotic differently than I do, but I think we might want to start simpler (I tend to think as robotic systems more like a rover with parts that are meant to move throughout the mission lifecycle as opposted to a more static vehicle like a satellite that just deploys systems and then points using GNC).

I put more thoughts here: nasa/fprime#4321, but I've been thinking about how I might contribute to this effort right now, and that it might be valuable for me to start defining a semi-realistic reference mission. Essentially, I'd be going back ot my grad school roots in mission design classes and putting together a dummy mission. But whereas in grad school we did everything in powerpoint, here I'd want to define enough fidelity that we could write some FPrime code to simulate the behaviors we're looking for.

It's been a bit since I really thought about this stuff, but my initial thought is to take some real LEO spacecraft for which I can find TLEs and some higher fidelity ephemeris information. Then I could create

"Predict"

Do a simple two-body propagation of the TLE and use that predicted ephemeris to build some dummy planning products in Aerie

"Actual"

Use the higher fidelity ephemeris to simulate an "actual" flight behavior.

There would be some disagreement between the two ephemerides, but that's by design. It could drive showing off some predict vs actual kind of stuff, and maybe even be used as the seed to kick off some kind of simulated fault protection F-Prime and show how flight rule checking could have precluded that.

I'd use ephemeris data from some real mission, but other than that I'd just fake whatever behavior made sense to demonstrate what makes sense to demonstrate.

[barugonman]

Thanks @mrmuszynski for the inputs! I'll note that outside of some domains like data management and telecom I don't have the engineering expertise to build up a whole reference mission myself and was looking to hunt around to see what had already been done first. But I was planning to think about 'robotics' in the most basic sense, and build up a set of 'reference demos' that could build up into a whole reference mission.

Specifically for predicts vs actuals, I was thinking about a situation where you have a basic power switch of some sort, with channels measuring data values that are impacted by updating the switch. The values change according to the settings of the switch, however we will code it such that the predicted simulation values are in some visual way different than the 'actuals' out of fprime. For example the rate of change of the predict could be a bit different than the actual. Also a predict might just be poor for some other reason (too low, too high, etc.) and one of the goals of such a demo will be to show how we can 'simply' fix the Aerie predict to match the actual (or maybe get it to 'good enough').

Otherwise if I had to pick a more spacecraft specific thing to demo it would be telecom given my background there.

[barugonman]

Note I'm keeping most of the design content in the Fprime discussion, some basic architecture diagrams are now up here:

nasa/fprime#4321