Handle zero USBL noise deviation

[yeseorizi]

Summary

Allow USBL transponders to use sigma=0 as a deterministic noise configuration instead of constructing an invalid C++ normal distribution.

Problem

The checked-in USBL tutorial configures both transponders with:

<mu>0</mu>
<sigma>0.0</sigma>

UsblTransponder::sendLocation() always constructed std::normal_distribution with that standard deviation. On the tested libstdc++ build, the distribution enforces stddev > 0, so the first ping aborted Gazebo:

std::normal_distribution<...>::param_type: Assertion '_M_stddev > 0' failed.

The reproduced baseline exited with code 134 and returned no USBL locations.

Change

• For sigma=0, use the configured mean as a deterministic per-axis offset without constructing std::normal_distribution.
• Preserve the existing random sampling path for sigma>0.
• Clamp a negative standard deviation to zero and emit a warning.

No topic names, interrogation routing, coordinate conversion, or positive-sigma sampling behavior is changed.

Validation

The exact candidate source was compiled in the ARM64 DAVE environment with ROS 2 Lyrical and Gazebo Jetty.

The original two-transponder tutorial geometry was then run with mu=0, sigma=0 and common interrogation:

Check	Result
Gazebo remained alive after pings	PASS
Spherical responses	6, IDs 1 and 2
Cartesian responses	6, IDs 1 and 2
Unexpected IDs	0
Maximum Cartesian axis error	6.13e-11 m
Maximum spherical reconstruction error	6.13e-11 m
Assertion / SIGABRT	0

A second run with sigma=-1 also stayed alive, returned the same deterministic geometry, and emitted one warning for each of the two invalid transponders.

Repository checks:

pre-commit run --all-files
# Passed

git diff --check
# Passed

Checklist

• Reproduced the checked-in sigma=0 crash
• Exact candidate source compiled
• Two-transponder runtime path verified
• Non-positive sigma behavior verified
• Repository hooks passed