Conv1D → BiGRU → more GRUs → norm → dense → sigmoid need layer sizes / sequence length alongside the nice boxes. also '3 states' but 4 names listed?? Poster architecture, transcribed in order: concatenated input → Conv1D →...
Bluetooth crib: 1=current angle, 2=average, 3=run mechanism missing bits: delimiter? ack? reconnect mid-command? don't let a half-packet turn into a real angle Updated protocol from the app notes: Request Purpose Response / behavior...
mean total 1.8844 s, but 21/50 trials over 2 s keep the tail, not just the average. battery trace gets to ~18 min continuous use... need duty-cycle notes next to that The latency sheet separates logic, mechanical and total response time across...
GRU sheets: 850 train / 150 test, no exact row overlap label 0 includes quick falls AND walking. need collection method + session split before treating that as a clean safe/unsafe dataset The three GRU spreadsheets are actual tabular data, not...
app: angle view, dated fall log, notification action 1 sec in requirements vs 3 sec in command table. also ±50° trigger vs model trigger... pick one version before implementing the scheduler transport requirement: the mobile-app document...
filename literally says SAMPLE confusion matrix, keep that attached fractional counts + June file vs September split sheets. not automatically the same experiment The document is explicitly named “JIC - sample confusion matrix.” Its two...
angle error / 2-minute drift / rate noise — three different things means under 5° but some readings over. also need to match each calibration number to its own run instead of borrowing the nicest one The CSEF angle-error table contains seven...
350 cm lever arm??? check whether that was meant to be 35 one digit and the COM shift grows a tentacle. redraw origin + actual weight positions before reusing this Board Stuff gives the one-dimensional relation x_COM = (m1*x1 + m2*x2)/(m1+m2) and...
gyro prediction + accel correction; check the variance notation before copying the Kalman function Ts = 0.004 from the example, still needs replacing. wondering how much loop-timer jitter alone shifts the estimate inputs: angular velocity from...
ARW 0.5487, BI 8.62 — first below target, second still above need the stationary recording + exact Allan-curve fit region. 'gyro noise' is way too vague a label Terms: angle random walk describes the integrated effect of rate noise; bias...
10 lb each side: didn't turn. useful failure, keep it 5 lb averages look better than the slow individual trials. need load + starting angle + commanded rotation in the same row CSEF motor tests record that 10 lb on each side never turned in...
sensor solder joints kept breaking. taped contact was enough to try readings, not enough for wearing this second level example gave angular rates — save axes / units / offsets / actual loop dt hardware problem: solder joints between the...
LCD + speaker working by day 22, finally feedback without serial monitor Bluefruit link here but classic Bluetooth in the later app notes... check which board belongs to which version recorded work: LCD and speaker were functional by Day 22. The...
scaleMotors / convertDeg — ints eating decimals, static input, degree/radian mess redo from the COM equation. test zero + both signs + acos limits BEFORE motor movement reference functions: scaleMotors(float inclination) and convertDeg(float...
second servo added → sweep turned into vibration / continuous rotation?? separate supplies got the test working. need to measure the power behavior instead of leaving the explanation at 'probably draw/reset' symptom: after adding the second...
lunch-bag straps + folded towels for padding, surprisingly useful supplies 23.302 N shoulders / 21.150 N waist. keep strap lengths with the numbers assembly: repurpose lunch-bag cross-body straps, loop them through the wheel assembly, and attach...
waist load down, thigh stress up. same total weight, new annoyance temporary leg assembly came apart too... moving on to shoulder straps recorded distribution: waist 31.923 N; legs 12.529 N. Together these equal the previous 44.452 N total, so...
waist-only test: 45 min, 44.452 N, awkward arms + waist fatigue weight placement matters a lot more than just total mass. shoulders next? legs first in the notes protocol: wear the initial device while walking for 45 minutes; combine qualitative...
extra cardboard made the boxes too small for the plates lol side layers out, tape around instead. save rope spacing + shaft height before changing anything else assembly notes, Days 9–11: use cardboard containers in place of printed parts....
printer unavailable → use the discarded filament spools rope carries the box, shaft pushes it. possible side issue: crooked spool = periodic wobble = fake body movement to the gyro? fabrication constraint: school printing required supplying...
move the weights → shift COM, motor shouldn't hold the whole load 43.3 kg / 73.07 cm / 50 cm shaft... keep the sketch with this or the angles are going to become nonsense again mechanical idea: use two moving pendulum weights to shift...
