Adcs controllers and main pointing strategy #92

nhz2 · 2019-11-27T20:55:00Z

No description provided.

kylekrol

For the most part looks really good. I left a couple miscellaneous questions. I think my largest concern is some slight divergence from where I think the flight computer implementations of the controllers may have been going (particularly the pointer state's PD buffer).

I just want to make sure we're one the same page about the internals of the MATLAB controllers and CXX controllers so we don't have any issues integrating the CXX functions into MATLAB (which I plan on doing once the full suite of attitude controllers is complete).

Let me know what you think!

@stewartaslan for awareness.

kylekrol · 2019-11-27T21:26:05Z

MATLAB/config.m

@@ -70,6 +70,7 @@
 % Torque on fuel/difference in angular rates in eci Nm/(rad/s).
 const.ATTITUDE_PD_KP = 75e-4; % imported from simulink
 const.ATTITUDE_PD_KD = 32.5e-4; % imported from simulink
+const.ATTITUDE_PD_derivative_buffer_size=1;


Will we need this for the flight algorithm you think? We're already buffering w_body in the estimation step.

I'm not sure, I wanted to have way to point even if quat_body_eci was NaN. Before, the derivative term was just the gyro, but I wanted to see what would happen if I just did finite differences.

Though I guess this basically doing the same thing as estimating w_body

Ah okay. But in terms of the sorta version 1.0 of the GNC algorithms we're good keeping the finite difference estimation in the estimator and no buffer in the controller?

MATLAB/detumbler.m

MATLAB/recursive_least_squares.m

kylekrol · 2019-11-27T21:35:05Z

MATLAB/sensor_reading.m

+target_position_eci= other_satellite_state.dynamics.position_eci;
+sensor_readings.target_position_ecef= utl_rotateframe(quat_ecef_eci,target_position_eci);
+target_velocity_eci= other_satellite_state.dynamics.velocity_eci;
+sensor_readings.target_velocity_ecef= utl_rotateframe(quat_ecef_eci, target_velocity_eci)-cross(rate_ecef,sensor_readings.target_position_ecef);


Does CD GPS actually give us a velocity reading or were we going to calculate that ourselves?

Yeah, we have to calculate that ourselves.

So sensor_readings.target_velocity_ecef will only live in MATLAB?

MATLAB/sensor_reading.m

kylekrol · 2019-11-27T21:37:56Z

MATLAB/update_FC_state.m

+estimator.sat2sun_eci= env_sun_vector(sensor_readings.time);
+estimator.mag_ecef= env_magnetic_field(sensor_readings.time,sensor_readings.position_ecef);
+[estimator.quat_ecef_eci,~]=env_earth_attitude(sensor_readings.time);
+estimator.quat_eci_ecef= utl_quat_conj(estimator.quat_ecef_eci);
+estimator.mag_eci =utl_rotateframe(estimator.quat_eci_ecef,estimator.mag_ecef);
+estimator.sat2sun_body= sensor_readings.sat2sun_body;
+estimator.quat_body_eci=utl_triad(estimator.sat2sun_eci,estimator.mag_eci,sensor_readings.sat2sun_body,sensor_readings.magnetometer_body);
+estimator.angular_rate_body= sensor_readings.gyro_body;
+estimator.angular_momentum_body= const.JB*estimator.angular_rate_body+sensor_readings.wheel_momentum_body;
+estimator.magnetometer_body= sensor_readings.magnetometer_body;
+estimator.magnetic_field_body=estimator.magnetometer_body;
+estimator.time= sensor_readings.time;
+estimator.position_eci= utl_rotateframe(estimator.quat_eci_ecef,sensor_readings.position_ecef);
+estimator.velocity_eci= utl_rotateframe(estimator.quat_eci_ecef,sensor_readings.velocity_ecef+cross(const.earth_rate_ecef,sensor_readings.position_ecef));
+estimator.target_position_eci=utl_rotateframe(estimator.quat_eci_ecef,sensor_readings.target_position_ecef);
+estimator.target_velocity_eci=utl_rotateframe(estimator.quat_eci_ecef,sensor_readings.target_velocity_ecef+cross(const.earth_rate_ecef,sensor_readings.target_position_ecef));
+estimator.tumbling= ~(norm(estimator.angular_momentum_body)<const.MAXWHEELRATE*const.JWHEEL*0.7);
+estimator.low_power= false;
+estimator.eclipse= env_eclipse(estimator.position_eci,estimator.sat2sun_eci);
+estimator.valid_gps=all(isfinite([estimator.position_eci; estimator.velocity_eci; estimator.time]));
+estimator.valid_target_gps=estimator.valid_gps && all(isfinite([estimator.target_position_eci; estimator.target_velocity_eci; estimator.time]));
+
+if estimator.valid_gps && ~estimator.eclipse && ~all(isfinite(estimator.sat2sun_body))
+    estimator.sat2sun_body= [0;0;-1]; % assume the sun is coming from the docking face
+    estimator.quat_body_eci=utl_triad(estimator.sat2sun_eci,estimator.mag_eci,estimator.sat2sun_body,estimator.magnetic_field_body);
+end


Throwing this in an attitude estimator function and gps estimation function may be a good idea?

MATLAB/update_FC_state.m

kylekrol · 2019-11-28T07:25:13Z

MATLAB/update_FC_state.m

+    state.secondary_desired_direction_body= [0;0;1];
+end
+if strcmp(state.main_state,'rendezvous')
+    if estimator.tumbling || estimator.low_power || estimator.eclipse || ~estimator.valid_gps || ~estimator.valid_target_gps


This may be temporary, but if everything is behaving nominally except for having target position, velocity would we want to go back to detumble or hold attitude with the full ADCS system? The latter would probably give us a better shot at getting comms, gps, etc

EDIT: Actually setting it to 'detumble' will let it propagate through to 'get gps', 'get target orbit', etc on the next FC pass correct? If so, I think this is still good.

Hopefully if cd gps lock is lost, the orbit estimator can just propagate the target orbit, that way it won't have to go into detumble.

nhz2 · 2020-01-06T16:04:03Z

To run the simulation in MATLAB, go to the MATLAB directory in psim, then run:

run_tests and make sure there are no failed tests.

main will generate the trajectories

plot_pointing_errors(main_state_trajectory) will show the pointing errors

plot_state_transitions(computer_state_follower_trajectory) will show the state of the follower flight computer.

there are other plotting functions in the plot folder

tanishqaggarwal

Looks great, I just have a few questions (mostly due to my ignorance of sim)

tanishqaggarwal · 2020-01-07T16:31:52Z

MATLAB/adcs/adcs_detumbler.m

+    case 5
+        %save magnetic field reading
+        state.old_magnetic_field= magnetometer_body;
+    case 9
+        %calculate finite difference and bang bang controller
+        deltaB= magnetometer_body-state.old_magnetic_field;
+        state.magrod_moment_cmd= -sign(deltaB)*const.MAXMOMENT;


What's the meaning of 5 and 9?

MATLAB/adcs/adcs_pointer.m

MATLAB/initialize_main_state.m

MATLAB/sensor_reading.m

tanishqaggarwal · 2020-01-08T23:32:11Z

MATLAB/update_FC_state.m

+    %move to next state if initialization time has passed
+    if state.initialization_hold_done || state.on_time>(30*60*1E9)
+        state.initialization_hold_done=true;%save this in eprom
+        state.main_state='detumble';


FYI Flight Software stores on_time instead of hold_done. The rationale is that the satellite can have low battery charge and can die during the post-deployment phase. I don't think it's necessary to copy this into sim though.

MATLAB/update_FC_state.m

tanishqaggarwal · 2020-01-09T07:57:09Z

MATLAB/update_FC_state.m

+    r_hat_cross_sat2sun_body=cross(estimator.sat2sun_body,r_hat_eci);
+    state.primary_current_direction_body= r_hat_body;
+    state.primary_desired_direction_body= [1;0;0];
+    if( norm(r_hat_cross_sat2sun_body) > sin(10*pi/180))


I'll change the alignment threshold in Flight Software to also be 10 degrees

In general is there some way we could standardize these thresholds across Flight Software and sim? It would be great to not have to manually ensure their similitude

kylekrol

I like the organizational improvements! If this is good to go, merging it sooner rather than later would be cool so I can get on the rendezvous changes as well.

tanishqaggarwal

I'm going to approve bc my questions aren't that important and sounds like work is waiting on this PR

stewartaslan

Looks good to me.

* starting adcs_update state machine * draft of adcs_update, main running * Draft of pointing and detumbling controllers * optional finite diff for pd controller * moved adcs into folder * removed config from unit tests * Update initialize_computer_states.m * better initial conditions * Long edge pointing * addressing comments * Some basic unit tests Former-commit-id: cc672c0 Former-commit-id: 695fc2fb42cf48094fdb355e53e9986d8f070488

nhz2 requested review from kylekrol, jer296 and shihaocao November 27, 2019 20:59

kylekrol reviewed Nov 27, 2019

View reviewed changes

nhz2 requested a review from stewartaslan November 28, 2019 01:27

kylekrol reviewed Nov 28, 2019

View reviewed changes

nhz2 added 6 commits January 6, 2020 10:18

starting adcs_update state machine

7c0b7fb

draft of adcs_update, main running

c6c1323

Draft of pointing and detumbling controllers

2f117a7

optional finite diff for pd controller

6fe93bd

moved adcs into folder

2d2647a

removed config from unit tests

178a0bb

nhz2 force-pushed the adcs_update branch from 9f356f2 to 178a0bb Compare January 6, 2020 15:20

Update initialize_computer_states.m

beeac2c

nhz2 marked this pull request as ready for review January 6, 2020 15:55

nhz2 requested review from kylekrol and tanishqaggarwal January 6, 2020 15:55

nhz2 added 2 commits January 6, 2020 10:58

better initial conditions

a722e61

Long edge pointing

f46aa49

tanishqaggarwal reviewed Jan 9, 2020

View reviewed changes

nhz2 added 2 commits January 11, 2020 12:47

addressing comments

a0e662f

Some basic unit tests

1f167cc

nhz2 requested a review from tanishqaggarwal January 15, 2020 17:10

Merge branch 'master' into adcs_update

d46e2da

tanishqaggarwal mentioned this pull request Jan 17, 2020

Magnetometer calibration task pathfinder-for-autonomous-navigation/FlightSoftware#139

Closed

kylekrol approved these changes Jan 18, 2020

View reviewed changes

tanishqaggarwal approved these changes Jan 18, 2020

View reviewed changes

stewartaslan reviewed Jan 18, 2020

View reviewed changes

stewartaslan self-requested a review January 18, 2020 20:04

stewartaslan approved these changes Jan 18, 2020

View reviewed changes

Merge branch 'master' into adcs_update

44e571b

nhz2 merged commit cc672c0 into master Jan 19, 2020

nhz2 deleted the adcs_update branch January 19, 2020 01:26

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Adcs controllers and main pointing strategy #92

Adcs controllers and main pointing strategy #92

nhz2 commented Nov 27, 2019

kylekrol left a comment

kylekrol Nov 27, 2019

nhz2 Nov 27, 2019

nhz2 Nov 27, 2019

kylekrol Nov 28, 2019

kylekrol Nov 27, 2019

nhz2 Nov 27, 2019

kylekrol Nov 27, 2019

kylekrol Nov 27, 2019

kylekrol Nov 28, 2019 •

edited

Loading

nhz2 Nov 29, 2019

nhz2 commented Jan 6, 2020

tanishqaggarwal left a comment

tanishqaggarwal Jan 7, 2020

tanishqaggarwal Jan 8, 2020

tanishqaggarwal Jan 9, 2020

tanishqaggarwal Jan 9, 2020

kylekrol left a comment

tanishqaggarwal left a comment

stewartaslan left a comment

Adcs controllers and main pointing strategy #92

Adcs controllers and main pointing strategy #92

Conversation

nhz2 commented Nov 27, 2019

kylekrol left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

kylekrol Nov 28, 2019 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

nhz2 commented Jan 6, 2020

tanishqaggarwal left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

kylekrol left a comment

Choose a reason for hiding this comment

tanishqaggarwal left a comment

Choose a reason for hiding this comment

stewartaslan left a comment

Choose a reason for hiding this comment

kylekrol Nov 28, 2019 •

edited

Loading