Repository navigation
Merge the remaining reviewed side branches (radio checks, frames, bench preflight, pilot options, public-number check) - #9
Merged
Conversation
…w-log faults A repeat arrival more than 1 s after the first copy of a seq is refused: it comes from a second sender session (board A rebooted while board B kept running), not from the radio. Before, such a pair imported cleanly with wrong arrival times, a false return-leg loss and false duplicates. Errors for 32-bit clock wraps, negative stamps, seq wraps and restarts, repeated or out-of-order sender rows, CR-only line endings, serial noise lines and captures that start after the header now say what happened. Adds tests for these and other raw-log quirks.
…er, so a reboot is never read as a wrap
… after a short first session
tools/check_public_numbers.py finds every number in README.md, docs/RESULTS.md and docs/LIMITATIONS.md and looks it up in tools/public_number_sources.json. Performance numbers must point at a registered run output whose bytes match the registry hash, and must round to the run value at the precision the text shows. Seven numbers not traced yet are listed as known so new drift fails the test.
…instead of adding an extra wrap
…pectation on raw p95
… second check value OFDM frames in sim/radio/mac.py now count the 6 us ERP signal extension that 802.11 includes in TXTIME. The published 1 Mbps unicast throughput in sim/radio/params.py exceeds the 350.9 per second ceiling for 250 B with the standard ACK, so it is marked contradicted with the reason.
…istered hashes (line endings, float tolerance, pixels, real changes)
…rked airtime check
… logs with clear preflight messages
…ws through platform_compare, since Windows CPUs can differ in the last bits
…pilot: lead prediction only while the turn-rate estimate passes a hysteresis threshold
…t/s rate, not as a fixed frame time
… registered rows on Windows
…econd rounding fix
# Conflicts: # tools/tests/test_espnow_airtime_budget.py
…issing run_meta.json and swapped echo logs
…op pilot # Conflicts: # tools/tests/test_pilot_full_chain.py
…w links by the round trip
… across a counter rollover kept in their epoch
…e reboot check, describe sample order in the counter unwrap, and document the reboot rule in the echo log format
…o a changed seed or count is never within tolerance
… with the sums rounded a second time
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Brings the remaining reviewed side branches onto main. No registered result folder, no committed output under
sim/runs/, and nothing intools/reproduce_registry.jsonchanged.What changed
Radio
sim/radio/mac.py,params.py: the 2.4 GHz OFDM airtime now includes the 6 us signal extension; the check value 357/s is flagged where it was wrong. Registered outputs that import these files were checked to be unchanged (the hardcoded 357 in the c-exp014 summary is left alone on purpose).tools/espnow_airtime_budget.py: DSSS/CCK frame time rounds up to a whole microsecond, with a hand-worked check (test_espnow_airtime_hand_check.py). At the default 1 Mbit/s rate, 10 drones at 100 Hz would need about 182 % of the channel.tools/import_espnow_log.py: names the bad field in misformatted echo rows and refuses an echoer log that mixes two sender sessions, telling a reboot from a slow link by the round trip. Documented as rule 12 indocs/ESPNOW_ECHO_LOG.md.Bench and frames
tools/preflight_bench_data.py: fails when radio times look like milliseconds in_uscolumns, names a missingrun_meta.json, and gives a clearer message for swapped echo logs.sim/frames.py,sim/clock_map.py,sim/adapters.py: frame and time layer for a real drone adapter (ENU/NED, clocks, units), with counter rollovers and reorders across a rollover handled.tools/fit_camera_bias_field.py: fits a camera bias field from a known-spots log, with unit checks per spot and per log.Closed-loop pilot (all opt-in, default rows unchanged)
--lead-switch: lead prediction only while the turn-rate estimate passes a hysteresis threshold. Its pre-set pass bands failed (11 of 16), so it is not a default.--grid,--grid-report: delay by loss grid. On turns no cell stays under 0.5 m for either goal rule.Checks
tools/check_public_numbers.pyand its test: public numbers must match a registered run output listed intools/public_number_sources.json. This adds a CI check on public text.tools/compare_reproduce_tolerance.py: explains why reproduced outputs differ from registered hashes (line endings, last-digit floats).sim/tests/test_baseline.py,test_a1_heading.py,tools/sens_monitor_noise.py: fresh reruns compared with registered rows throughplatform_compare, since Windows CPUs can differ in the last bits.Testing
test_printed_estimate_matches_measured_formula, which rounded its recomputed estimate to 0.1 s a second time and so allowed a 0.1 s gap near a rounding boundary. It now compares with the exact sum (78ff521).