test: add unit test suite for CUBE algorithm core by rolker · Pull Request #19 · rolker/cube_bathymetry

rolker · 2026-02-25T19:08:19Z

Summary

Adds gtest-based unit test suite covering all core CUBE algorithm components
test_hypothesis (11 tests): Constructor, null hypothesis, monitor reset, Kalman filter update equations, outlier/intervention detection, Bayes factor drift detection, input sample variance tracking, convergence
test_node (12 tests): Hypothesis add/choose/best, update with outlier intervention, median queue pre-filter, queue flush, depth extraction, outlier truncation, convergence
test_parameters (9 tests): All IHO survey orders (exclusive, special, order1a/b, order2), invalid order exception, distance scale, variance scale, context search range, default parameter values
test_grid (6 tests): Construction, bounds, initial NaN values, single/multiple sounding insertion, out-of-bounds rejection, batch insertion
test_map_sheet (8 tests): Construction, empty state, sounding-triggered grid creation, on-demand grid creation, grid index computation (positive and negative coords), bounds expansion, multi-grid tiling

Also includes a guard against divide-by-zero in Node::truncate() for queues with fewer than 3 elements.

Merge order: This PR is rebased on top of PR #20. Merge #20 first, then this PR will apply cleanly.

Closes #17

Test plan

All 5 test executables build cleanly
All 46 unit tests pass (100%)
Copilot review comments addressed
CI passes

Authored-By: Claude Code Agent
Model: Claude Opus 4.6

Copilot

Pull request overview

Adds a GoogleTest (ament_cmake_gtest) unit test suite for the core cube_bathymetry CUBE algorithm types, and includes two numerical fixes in Node that were cherry-picked from PR #18.

Changes:

Add 5 new gtest executables covering Hypothesis, Node, Parameters, Grid, and MapSheet.
Wire the tests into the build (CMakeLists.txt) and add the required ROS2 test dependency (package.xml).
Fix CUBE math in Node: use Parameters::variance_scale for variance dilution and correct the truncate quotient denominator (n*n-1).

Reviewed changes

Copilot reviewed 8 out of 8 changed files in this pull request and generated 4 comments.

Show a summary per file

File	Description
cube_bathymetry/test/test_hypothesis.cpp	Adds unit tests for hypothesis update/monitoring behavior.
cube_bathymetry/test/test_node.cpp	Adds unit tests for node hypothesis management, queueing, truncation, and extraction.
cube_bathymetry/test/test_parameters.cpp	Adds unit tests for IHO order handling and parameter derivations (distance/variance scale, context ranges).
cube_bathymetry/test/test_grid.cpp	Adds unit tests for grid construction, bounds, and sounding insertion.
cube_bathymetry/test/test_map_sheet.cpp	Adds unit tests for MapSheet tiling, grid creation, and grid index computations.
cube_bathymetry/src/node.cpp	Applies the variance dilution + truncate formula fixes.
cube_bathymetry/package.xml	Adds `ament_cmake_gtest` as a test dependency.
cube_bathymetry/CMakeLists.txt	Registers and links all gtest targets under `BUILD_TESTING`.

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

Copilot · 2026-02-25T19:12:47Z

+
+#include <gtest/gtest.h>
+#include "cube_bathymetry/map_sheet.h"
+#include <cmath>


This test file uses std::vector but doesn’t include <vector>, relying on transitive includes from project headers. Add an explicit #include <vector> to make the test self-contained and reduce build fragility.

Suggested change

#include <cmath>

#include <cmath>

#include <vector>

Copilot · 2026-02-25T19:12:47Z

  ssd -= mean*mean/(n+1);
  mean /= (n+1);
-  float ssd_k = n*ssd/(n*n+1);
+  float ssd_k = n*ssd/(n*n-1);

  /* Run the list computing quotients; outliers are removed from the queue.


Node::truncate() can divide by zero / produce invalid math when the queue has too few elements. ssd_k = n*ssd/(n*n-1) is undefined for n==1 (queue size 2), and q later uses (n-1) as well. Since queueFlush() calls truncate() whenever the queue is non-empty, add an early return/alternate handling for small queues (e.g., queue_.size() < 3) before computing ssd_k/q.

Copilot · 2026-02-25T19:12:47Z

+  // The chosen one has the most samples (1 each), so either could be chosen
+  // but we should have at least one hypothesis


This test is named UpdateWithOutlierCreatesNewHypothesis, but it doesn’t assert any condition that would fail if no new hypothesis was created (it only checks chooseHypothesis() is non-null). Consider adding an assertion that distinguishes the 1-hypothesis update case from the intervention/new-hypothesis case (e.g., verifying the chosen estimate is not pulled toward the midpoint, or exposing hypothesis count via a test-only accessor).

Suggested change

// The chosen one has the most samples (1 each), so either could be chosen

// but we should have at least one hypothesis

// If no new hypothesis was created, a single hypothesis updated with both

// observations (10 and 50) would be pulled toward the midpoint (~30).

// Verify the chosen estimate is not near this midpoint, which distinguishes

// the intervention / new-hypothesis behavior from a simple single-hypothesis

// update.

EXPECT_FALSE(std::fabs(chosen->current_estimate - 30.0) < 5.0);

// The chosen one has the most samples (1 each), so either 10 or 50 could be

// chosen, but we should not see an estimate near the midpoint.

Copilot · 2026-02-25T19:12:48Z

+
+#include <gtest/gtest.h>
+#include "cube_bathymetry/grid.h"
+#include <cmath>


This test file uses std::vector but doesn’t include <vector>, relying on transitive includes from project headers. Add an explicit #include <vector> to make the test self-contained and reduce build fragility.

Suggested change

#include <cmath>

#include <cmath>

#include <vector>

The 99% confidence interval on a unit normal distribution is 2.576 (matching the original CUBE's 2.56). The previous value of 2.95 over-estimated the bound, making outlier detection less aggressive and expanding the sounding radius of influence. Closes #10 Co-Authored-By: Claude Code Agent <roland+claude-code@ccom.unh.edu>

The original CUBE code (with TRACK_IP_SAMP_VAR defined, which is the default) reports uncertainty using the input sample variance rather than the posterior estimate variance. Using posterior variance made reported uncertainty overly optimistic. Also fixes queue flush to use the original's alternating left/right walk from the median rather than repeated middle extraction, matching the intended processing order. Closes #13, Closes #14 Co-Authored-By: Claude Code Agent <roland+claude-code@ccom.unh.edu>

Add gtest-based unit tests covering all core CUBE algorithm components: - test_hypothesis: Kalman filter update, Bayes factor monitoring, convergence - test_node: hypothesis management, median queue, outlier truncation - test_parameters: IHO order limits, grid resolution, default values - test_grid: construction, sounding insertion, depth extraction - test_map_sheet: grid creation, tiling, bounds computation Closes #17 Co-Authored-By: Claude Code Agent <roland+claude-code@ccom.unh.edu>

- Add missing #include <vector> in test_grid.cpp and test_map_sheet.cpp - Guard Node::truncate() against divide-by-zero with queue size < 3 - Strengthen UpdateWithOutlierCreatesNewHypothesis assertion - Update ExtractDepthAndUncertaintyWithData test for input_sample_variance Co-Authored-By: Claude Code Agent <roland+claude-code@ccom.unh.edu>

Copilot

Pull request overview

Copilot reviewed 9 out of 9 changed files in this pull request and generated no new comments.

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

Copilot AI review requested due to automatic review settings February 25, 2026 19:08

Copilot started reviewing on behalf of rolker February 25, 2026 19:08 View session

Copilot AI reviewed Feb 25, 2026

View reviewed changes

rolker and others added 2 commits February 25, 2026 14:53

rolker mentioned this pull request Feb 25, 2026

fix: correct CONF_99PC, uncertainty extraction, and queue flush order #20

Merged

3 tasks

rolker and others added 2 commits February 25, 2026 15:50

rolker force-pushed the feature/issue-17 branch from 82e1a95 to 71e4224 Compare February 25, 2026 21:02

rolker requested a review from Copilot February 25, 2026 21:05

Copilot started reviewing on behalf of rolker February 25, 2026 21:06 View session

Copilot AI reviewed Feb 25, 2026

View reviewed changes

rolker merged commit f14724b into jazzy Feb 25, 2026
4 checks passed

rolker deleted the feature/issue-17 branch February 25, 2026 21:18

rolker mentioned this pull request Feb 26, 2026

Unit tests for core algorithms (error_model, grid, hypothesis, geo_grid) #4

Open

rolker mentioned this pull request Apr 22, 2026

Adapt detections_to_pointcloud to mru_transform TwistWithCovarianceStamped API #28

Closed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

test: add unit test suite for CUBE algorithm core#19

test: add unit test suite for CUBE algorithm core#19
rolker merged 4 commits intojazzyfrom
feature/issue-17

rolker commented Feb 25, 2026 •

edited

Loading

Uh oh!

Copilot AI left a comment

Uh oh!

Copilot AI Feb 25, 2026

Uh oh!

Copilot AI Feb 25, 2026

Uh oh!

Copilot AI Feb 25, 2026

Uh oh!

Copilot AI Feb 25, 2026

Uh oh!

Copilot AI left a comment

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

		// The chosen one has the most samples (1 each), so either could be chosen
		// but we should have at least one hypothesis

-  // The chosen one has the most samples (1 each), so either could be chosen
-  // but we should have at least one hypothesis
+  // If no new hypothesis was created, a single hypothesis updated with both
+  // observations (10 and 50) would be pulled toward the midpoint (~30).
+  // Verify the chosen estimate is not near this midpoint, which distinguishes
+  // the intervention / new-hypothesis behavior from a simple single-hypothesis
+  // update.
+  EXPECT_FALSE(std::fabs(chosen->current_estimate - 30.0) < 5.0);
+  // The chosen one has the most samples (1 each), so either 10 or 50 could be
+  // chosen, but we should not see an estimate near the midpoint.

Conversation

rolker commented Feb 25, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Summary

Test plan

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Reviewed changes

Uh oh!

Copilot AI Feb 25, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Feb 25, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Feb 25, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Feb 25, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

rolker commented Feb 25, 2026 •

edited

Loading