G-FOLD Diversion Test

[Thiel]

JPL released a video of their new landing divert test.

[PakledHostage]

Impressive. Thanks for sharing.

[maccollo]

What are they making this for?

Are they planning to use this for the mars 2020 rover or is it just something they're doing on the side?

[Streetwind]

This isn't for anything specific. This is a general purpose improvement to autopilot functionality. The idea is that all rockets or vehicles designed for powered landings will be capable of doing this in the future.

[tavert]

If you look at the control algorithms and researchers involved here, this has common development heritage with SpaceX's Grasshopper. JPL's more interested in the scientific benefits of improved powered landing accuracy on other planets, SpaceX for now is looking at returning and landing their launch stages.

[K^2]

It's just a choice of parameters, though, isn't it? I very much doubt that something would go horribly wrong with Grasshoper control algorithm if somebody adjusted parameters for gravity and atmosphere. I suppose, some stopping conditions could be hard-coded magic numbers, but it's kind of a bad practice, and I sincerely hope that SpaceX would know better.

[tavert]

For terminal guidance I imagine the implementations could be quite similar. Hence the JPL flight testing looking pretty similar to the Grasshopper testing. What might be quite a bit different would be the incoming initial conditions on re-entry, would probably look much different for a Mars scientific mission than an Earth booster return.

I've been voraciously reading as many papers like this one http://www.lpi.usra.edu/meetings/marsconcepts2012/pdf/4193.pdf and anything else by the same authors trying to find out the implementation details. Other than being formulated as a second-order cone problem, I don't know whether they're using an existing solver or if they wrote their own custom optimization code.

Edited January 23, 2014 by tavert

[rdfox]

I asked on YT about whether this uses a predefined menu of landing site options, or if it uses onboard sensors to select a new landing site entirely autonomously.

Even if it *is* picking from a predefined menu, it's a major step towards that sort of fully-autonomous site-selection option.

[magnemoe]

This isn't for anything specific. This is a general purpose improvement to autopilot functionality. The idea is that all rockets or vehicles designed for powered landings will be capable of doing this in the future.

One obvious use is automated landings who can change landing spots if the current is unsuitable, kind of like Apollo 11 did.

Probably more interesting for Mars landings where you can get off target in reentry and parachute braking phase.