[edufford]

Every car the spawner asks for is now [radar, controller, physics],
with the packaged controller FMU as the controller, so it follows the
car ahead in its lane with IDM and wants its spawn speed back on a free
road. The ego and traffic_scale's cars keep constant speed. The ego
does until it shares a lane with traffic, and the scaling run because
it measures the conductor, and an FMU per car is its own experiment.

The world hash moves from eccd08f9a316bbbc to 580e8d37b14c6abd, for
every traffic car's scans and acceleration commands and the poses they
changed. Every traffic car slows somewhat, since IDM brakes for a lead
well before it is close, and the slowest falls from 16.2 to 13.3 m/s.
The ego's poses and the turnover are unchanged: fourteen cars, eight
retired. README's sample output moves with the hash.

The mapping gives the FMU's pose inputs the state the physics is built
from, since at the instant a car joins its controller steps before its
physics has published, and would otherwise steer from the origin. Its
keys are concrete, because a scenario knows its world where a native
constructor does not. Building a car can now fail for a missing package
or an FMU that will not instantiate, so the scenario functions return a
ScenarioError wrapping those beside ConductorError.

Two tests guard following: no two same-lane cars ever overlap, and some
traffic car changes speed by more than 1 m/s. Pointing the radar
binding at a key nobody publishes fails both. A unit test decodes what
one FMU step published as the plant's own AccelCmd and SteerCmd.
Writing it showed a free road is not exactly zero: FREE_ROAD is a
long finite range, so a car at its desired speed is commanded -3e-16.

cargo xtask verify now packages the FMUs before the tests, since the
tests and the demo cannot start without a package, and an edited law
would otherwise pass locally on the copy packaged before it and fail in
CI. It costs 0.7 s when nothing changed. Validation and the comparison
of each package against its laws stay in verify-fmus.

The examples' log drops an FMU's trace of its own calls, about 1,700
lines a run, through one init_logging the six mains share. It drops
a message only when its target is exactly trace and its module path
is the fmu that fmi gives everything an FMU logs. Targets would
match the target as a prefix and cannot see module paths, so it would
silence any target beginning with the word.

Measured and deferred: each FMU car costs about 70 ms to join on
Windows, since fmi 0.8 loads a fresh copy of the DLL per instance,
taking the release demo from 0.07 s to 1.27 s. PLAN.md's deferred list
carries the shared archive that would fix it. The spawner can place a
car right beside one already in the lane, 0.99 m bumper to bumper at
worst with seed 42, since its frontier is where it last spawned rather
than where the cars now are; that is left for the ego-lane change.

---

Draft notes (deleted before merge)

M6-PLAN.md PR E. Three commits, each passing cargo xtask verify on
its own: verify packaging, traffic on the FMU, the log filter.

• Review the diff, especially controller_mapping and the two
  highway tests.
• CI on all four agents agrees on 580e8d37b14c6abd. This is the
  first time the hash runs through the FMU's own compiled laws on
  each platform.
• Decide whether the spawner fix stays with PR G as agreed.

Found on the way and left alone: PoseLogger re-logs a retired car's
last pose every second, on main too, against the rule that per-actor
state is dropped on leave. Worth an issue or a small PR of its own.

Not run: cargo xtask verify-fmus, since nothing here touches the FMU
crate or the laws it links.

🤖 Generated with Claude Code