feat: ESO gain optimization (2nd-order LESO, argmin GSS, at_ceiling marker)#136
feat: ESO gain optimization (2nd-order LESO, argmin GSS, at_ceiling marker)#136nerdCopter wants to merge 2 commits into
Conversation
|
Note Reviews pausedIt looks like this branch is under active development. To avoid overwhelming you with review comments due to an influx of new commits, CodeRabbit has automatically paused this review. You can configure this behavior by changing the Use the following commands to manage reviews:
Use the checkboxes below for quick actions:
📝 WalkthroughWalkthroughAdds an optional 2nd-order LESO bandwidth optimizer and optional Markdown statistical reporting, new Changes
Estimated code review effort🎯 4 (Complex) | ⏱️ ~45 minutes 🚥 Pre-merge checks | ✅ 5✅ Passed checks (5 passed)
✏️ Tip: You can configure your own custom pre-merge checks in the settings. ✨ Finishing Touches📝 Generate docstrings
🧪 Generate unit tests (beta)
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. Comment |
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
This comment was marked as outdated.
This comment was marked as outdated.
Sorry, something went wrong.
There was a problem hiding this comment.
Actionable comments posted: 1
🧹 Nitpick comments (5)
src/main.rs (1)
1316-1331: Minor:any_validguard is effectively unreachable.Any unknown/empty axis token already triggers
print_usage_and_exitvia theotherarm, so theif any_validcheck at line 1329 never fires for thefalsecase. It's harmless, but you could drop it and just always assignplot_config.eso_axes = axes;for clarity. Not a bug.🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed. In `@src/main.rs` around lines 1316 - 1331, The any_valid boolean guard is redundant because unknown tokens call print_usage_and_exit in the match's other arm; remove the any_valid flag and its conditional and always assign plot_config.eso_axes = axes after parsing. Specifically, delete the any_valid variable and the if any_valid { ... } check, leaving axes to be set directly into plot_config.eso_axes (referencing the axes array, the any_valid variable, and plot_config.eso_axes and the print_usage_and_exit call to locate the code).src/report.rs (2)
176-176: UseAXIS_COUNTinstead ofAXIS_NAMES.len().min(3).
axis_names.rsalready definesAXIS_COUNT = 3with a compile-time assertion that it equalsAXIS_NAMES.len(). The.min(3)embeds a magic number that duplicates that invariant.- for axis_idx in 0..AXIS_NAMES.len().min(3) { + for axis_idx in 0..crate::axis_names::AXIS_COUNT {As per coding guidelines: "Axis indices (0=Roll, 1=Pitch, 2=Yaw) must use the
src/axis_names.rsmodule withAXIS_COUNT,AXIS_NAMES, andaxis_name()function instead of creating redundant axis index constants".🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed. In `@src/report.rs` at line 176, The loop uses a magic-number expression AXIS_NAMES.len().min(3); replace that with the canonical constant AXIS_COUNT (from axis_names.rs) so axis indexing relies on the single authoritative value; update the for loop range to for axis_idx in 0..AXIS_COUNT and ensure any related uses rely on AXIS_NAMES and axis_name() rather than duplicated constants.
78-82: Dead condition and magic number.The outer loop iterates
0..AXIS_NAMES.len().min(3)(=0..3), soaxis < 4is always true and the check is dead. If the intent was to also report Throttle (setpoint index 3), the loop bound needs to change; otherwise drop the guard. The literal4is also a magic number per the repo guideline.- if axis < 4 { - if let Some(sp) = row.setpoint[axis] { - setpoint.push(sp); - } + if let Some(sp) = row.setpoint[axis] { + setpoint.push(sp); }As per coding guidelines: "All constants go in
src/constants.rs— no hardcoded magic numbers in function code".🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed. In `@src/report.rs` around lines 78 - 82, The condition `if axis < 4` in the block using `axis` and `row.setpoint[axis]` is dead because the loop already iterates `0..AXIS_NAMES.len().min(3)`; either extend the loop bound to include Throttle (index 3) or remove the guard entirely. Replace the hardcoded `4` with a named constant (e.g., MAX_REPORTED_AXIS) defined in src/constants.rs and use that constant to control the loop or the check; update the loop range to `0..AXIS_NAMES.len().min(MAX_REPORTED_AXIS)` if you intend to include index 3, otherwise delete the `axis < ...` guard. Ensure references are to `AXIS_NAMES`, the loop variable `axis`, and `row.setpoint` so the change is localized to the existing logic.src/eso.rs (1)
244-253: Minor: move the/ 3.0stability divisor intoconstants.rs.The
sample_rate / 3.0stability bound is a domain-specific constant (LESO discrete-time stability limit) repeated inline. Putting it insrc/constants.rs(e.g.ESO_STABILITY_DIVISOR: f64 = 3.0) keeps the rule in one place and matches how the other ESO knobs are defined.As per coding guidelines: "All constants go in
src/constants.rs— no hardcoded magic numbers in function code".🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed. In `@src/eso.rs` around lines 244 - 253, Replace the hardcoded divisor 3.0 with a named constant defined in src/constants.rs (e.g. ESO_STABILITY_DIVISOR: f64 = 3.0) and use that constant in the omega0_max_stable calculation in the function that computes omega0_max_stable (references: omega0_max_stable, sample_rate, config.omega0_max, config.omega0_min); update the error message computation that multiplies config.omega0_min * 3.0 to use the same constant (config.omega0_min * ESO_STABILITY_DIVISOR) so the stability rule is centralized in constants.rs.src/plot_functions/plot_eso.rs (1)
51-54: Nit:p95_idxname is misleading.The index is computed from
UNIFIED_Y_AXIS_PERCENTILE, not a hardcoded 95th percentile. Rename topctl_idx(or similar) so it doesn't imply a specific value if the constant is later tuned.🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed. In `@src/plot_functions/plot_eso.rs` around lines 51 - 54, The local variable name p95_idx is misleading because the index is derived from UNIFIED_Y_AXIS_PERCENTILE (a tunable constant); rename p95_idx to a neutral name like pctl_idx (or percentile_idx) in the block that computes half_range and update any subsequent uses in the same scope to match (referencing all_abs, UNIFIED_Y_AXIS_PERCENTILE, UNIFIED_Y_AXIS_HEADROOM_SCALE, UNIFIED_Y_AXIS_MIN_SCALE and the half_range calculation) so the code reflects the generic percentile-based computation.
🤖 Prompt for all review comments with AI agents
Verify each finding against the current code and only fix it if needed.
Inline comments:
In `@src/plot_functions/plot_eso.rs`:
- Around line 98-108: The f_hat scaling uses UNIFIED_Y_AXIS_MIN_SCALE instead of
the actual plot half_range, so compute fhat_scale after half_range is known
(i.e., inside draw_stacked_plot or by passing half_range into where fhat_scale
is built) and base it on data.fhat_max_abs and half_range so f_hat fills the
intended fraction of the Y range; replace the magic 0.5 with a new constant
ESO_FHAT_Y_FRACTION placed in src/constants.rs, and update usages of
fhat_scale/fhat_max_abs in plot_eso.rs (references: build_axis_data,
draw_stacked_plot, data.fhat_max_abs, fhat_scale, UNIFIED_Y_AXIS_MIN_SCALE) to
use the new calculation and constant.
---
Nitpick comments:
In `@src/eso.rs`:
- Around line 244-253: Replace the hardcoded divisor 3.0 with a named constant
defined in src/constants.rs (e.g. ESO_STABILITY_DIVISOR: f64 = 3.0) and use that
constant in the omega0_max_stable calculation in the function that computes
omega0_max_stable (references: omega0_max_stable, sample_rate,
config.omega0_max, config.omega0_min); update the error message computation that
multiplies config.omega0_min * 3.0 to use the same constant (config.omega0_min *
ESO_STABILITY_DIVISOR) so the stability rule is centralized in constants.rs.
In `@src/main.rs`:
- Around line 1316-1331: The any_valid boolean guard is redundant because
unknown tokens call print_usage_and_exit in the match's other arm; remove the
any_valid flag and its conditional and always assign plot_config.eso_axes = axes
after parsing. Specifically, delete the any_valid variable and the if any_valid
{ ... } check, leaving axes to be set directly into plot_config.eso_axes
(referencing the axes array, the any_valid variable, and plot_config.eso_axes
and the print_usage_and_exit call to locate the code).
In `@src/plot_functions/plot_eso.rs`:
- Around line 51-54: The local variable name p95_idx is misleading because the
index is derived from UNIFIED_Y_AXIS_PERCENTILE (a tunable constant); rename
p95_idx to a neutral name like pctl_idx (or percentile_idx) in the block that
computes half_range and update any subsequent uses in the same scope to match
(referencing all_abs, UNIFIED_Y_AXIS_PERCENTILE, UNIFIED_Y_AXIS_HEADROOM_SCALE,
UNIFIED_Y_AXIS_MIN_SCALE and the half_range calculation) so the code reflects
the generic percentile-based computation.
In `@src/report.rs`:
- Line 176: The loop uses a magic-number expression AXIS_NAMES.len().min(3);
replace that with the canonical constant AXIS_COUNT (from axis_names.rs) so axis
indexing relies on the single authoritative value; update the for loop range to
for axis_idx in 0..AXIS_COUNT and ensure any related uses rely on AXIS_NAMES and
axis_name() rather than duplicated constants.
- Around line 78-82: The condition `if axis < 4` in the block using `axis` and
`row.setpoint[axis]` is dead because the loop already iterates
`0..AXIS_NAMES.len().min(3)`; either extend the loop bound to include Throttle
(index 3) or remove the guard entirely. Replace the hardcoded `4` with a named
constant (e.g., MAX_REPORTED_AXIS) defined in src/constants.rs and use that
constant to control the loop or the check; update the loop range to
`0..AXIS_NAMES.len().min(MAX_REPORTED_AXIS)` if you intend to include index 3,
otherwise delete the `axis < ...` guard. Ensure references are to `AXIS_NAMES`,
the loop variable `axis`, and `row.setpoint` so the change is localized to the
existing logic.
🪄 Autofix (Beta)
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: Organization UI
Review profile: CHILL
Plan: Pro
Run ID: 209c4fdb-6a38-4c19-8c97-10636f7e228e
⛔ Files ignored due to path filters (1)
Cargo.lockis excluded by!**/*.lock
📒 Files selected for processing (7)
Cargo.tomlsrc/constants.rssrc/eso.rssrc/main.rssrc/plot_functions/mod.rssrc/plot_functions/plot_eso.rssrc/report.rs
✅ Files skipped from review due to trivial changes (3)
- Cargo.toml
- src/plot_functions/mod.rs
- src/constants.rs
This comment was marked as outdated.
This comment was marked as outdated.
Sorry, something went wrong.
This comment was marked as outdated.
This comment was marked as outdated.
Sorry, something went wrong.
This comment was marked as outdated.
This comment was marked as outdated.
Sorry, something went wrong.
|
@coderabbitai, all nitpicks resolved? full review |
This comment was marked as outdated.
This comment was marked as outdated.
This comment was marked as outdated.
This comment was marked as outdated.
This comment was marked as outdated.
This comment was marked as outdated.
|
@Quick-Flash is this useful/working? still interested? |
This comment was marked as outdated.
This comment was marked as outdated.
This comment was marked as outdated.
This comment was marked as outdated.
nerdCopter
changed the title
|
@Quick-Flash , is this ESO thing functioning as desired? [edit: replaced graphs with better legend] 
via |
Add 2nd-order LESO (Linear Extended State Observer) bandwidth optimizer: - src/eso.rs: golden-section search via argmin over omega_0 ∈ [ESO_OMEGA0_MIN, min(sample_rate/ESO_OMEGA0_STABILITY_RATIO, ESO_OMEGA0_MAX)]. N-step-ahead open-loop prediction cost (unimodal: low omega_0 = stale f_hat, high = noise). b0 auto-estimated via OLS on rate derivatives (QuickFlash method); falls back to ESO_DEFAULT_B0 when too few high-excitation samples. at_ceiling: bool set when omega0_opt >= omega0_max_stable - ESO_GSS_TOLERANCE. - src/plot_functions/plot_eso.rs: stacked time-domain ESO output plot per axis (omega_meas blue, omega_hat orange, f_hat green scaled to ESO_FHAT_Y_FRACTION). Legend shows [at ceiling] when optimizer hit its stability-constrained bound. - src/constants.rs: ESO_OMEGA0_MIN/MAX, GSS_TOLERANCE, GSS_MAX_ITER, DEFAULT_B0, N_AHEAD_STEPS, WARMUP_FRACTION, B0_* thresholds, OMEGA0_STABILITY_RATIO, FHAT_Y_FRACTION, COLOR_ESO_*. - src/main.rs: --eso flag (eso_requested bool); --eso-b0 sets b0 with eso_b0_user_override; plot_config.run_eso/eso_b0/eso_b0_user_override applied after arg loop; --eso alone triggers PlotConfig::none() so only ESO output is generated. - Cargo.toml/Cargo.lock: add argmin 0.11 dependency. - OVERVIEW.md: ESO section and TOC entry. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
nerdCopter
force-pushed
the
20260402_Quickflash-ESO
branch
from
eab71ab to
ccaad92
Compare
nerdCopter
changed the title
- Merge origin/master (fddd046, PR #156 always-on markdown report) - Add eso_results: [Option<EsoResult>; AXIS_COUNT] field to FlightReport - Add ESO Gain Optimization section to generate_markdown_report: table of omega_0, beta1, beta2, b0, b0 source, MSE, [at ceiling] note per axis; skipped axes shown as N/A - Move ESO block before report generation so results and PNG link are available when FlightReport is assembled - ESO PNG link added to png_links on successful plot write Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
|
Tip For best results, initiate chat on the files or code changes.
PR (ﾉ◕ヮ◕)ﾉ*:･ﾟ✧ 🐇 |
1 participant
AI Generated pull-request
Summary
Adds offline 2nd-order LESO (Linear Extended State Observer) bandwidth optimization from blackbox flight data, with time-domain output plots and markdown report integration.
CLI Flags
--eso— enables ESO gain optimization (disabled by default)--eso-b0 <value>— overrides control effectivenessb0(default: auto-estimated via OLS on high-excitation samples using the QuickFlash method)New Files
src/eso.rsCore optimizer. Per axis:
omega) and PID sum (P+I+D+F) as measured output and control inpute = omega_meas[k] - omega_hatomega_hat += Ts * (f_hat + b0 * u[k] + beta1 * e)f_hat += Ts * (beta2 * e)beta1 = 2*omega_0,beta2 = omega_0^2[ESO_OMEGA0_MIN, min(sample_rate/3, ESO_OMEGA0_MAX)]omega_0 < sample_rate / 3(enforced automatically)at_ceiling: boolset when the optimizer reaches the stability-constrained upper boundb0auto-estimated via OLS on rate derivatives; falls back toESO_DEFAULT_B0when too few high-excitation samples are availableEsoResultper axis:omega0_opt,beta1,beta2,b0,b0_auto,mse,at_ceilingsrc/plot_functions/plot_eso.rsStacked time-domain output plot per axis:
omega_meas(blue),omega_hat(orange),f_hat(green, scaled toESO_FHAT_Y_FRACTIONof Y half-range)[at ceiling]when optimizer hit its upper bound<root>_ESO_output_stacked.pngModified Files
src/main.rs--eso/--eso-b0flag parsing and validationeso_resultsand the ESO PNG link are available toFlightReportpng_links.push(...)for the ESO plot on successful writeeso_resultspassed intoFlightReportsrc/report.rseso_results: [Option<EsoResult>; AXIS_COUNT]field onFlightReportgenerate_markdown_report(): table withomega_0,beta1,beta2,b0,b0 source,MSE, and[at ceiling]note per axis; section is omitted when ESO was not runsrc/constants.rsESO tuning and plot constants:
ESO_OMEGA0_MIN/MAX,ESO_GSS_TOLERANCE/MAX_ITER,ESO_DEFAULT_B0,ESO_N_AHEAD_STEPS,ESO_WARMUP_FRACTION,ESO_B0_*thresholds,ESO_OMEGA0_STABILITY_RATIO,ESO_FHAT_Y_FRACTION,COLOR_ESO_MEAS/HAT/FHATOVERVIEW.mdESO section and TOC entry
Cargo.toml/Cargo.lockAdds
argmin 0.11dependency