Working Notes: a commonplace notebook for recording & exploring ideas.
Home. Site Map. Subscribe. More at expLog.

Distributed Systems / Space

Flight control for visors

Simulation Testing <- quick/thorough

Visors: satellite space mission 24/10

distributed space telescope -- 2 sats in formation

high resolution image of the corona

ambitious & low cost -- proof of concept

40m apart spacecraft

1.5cm by 1.5cm tolerance

for 10s

RTM frame -- relative motion

R = up away from earth

T = direction of motion

N = normal direction

GNC subsystem

stateful software component

motion estimation

gnc / software / linux / flight computer - {actuators, sensors, crosslink}, gps, ground radios

relies on gpus for high precision navigation for cm accurate and

API

exposes callbacks to update the state of the world in GNC

which is interesting, would have expected it to be inside out

outputs are: maneuvers, saving telemetry, ticks

Tick abstraction for delayed work

Fault detection

request current system time and time

~10,000 lines of C++ O_O

ECOS: numerical optimization

SOFA: coordinate frame defs, etc. } 50k lines

threading model?

~5 people

set a reminder to check in if they succeed :)

https://slab.stanford.edu/projects/visors

simulation: replace all the inputs to the software

simulation testing framework

BUT THIS DOESN'T TEST THE SYNCHRONIZATION

runs against the actual software

model as many sources of noise in the simulation

network crosslink -- datagram

makes it easy because it's deterministic to do print statement

deterministic system and deterministic simulation

make it a pure function of inputs

like a large state machine

except for memory management

multi-threading?

simulation maintains an event loop

takeaways -- different set of tradeoffs against

talks

Kunal