Acoustics simulator dispatch¶

Entry point: acoustics.launch.py.

Called from task_generator.launch.py's OpaqueFunction (once per environment, ahead of the human dispatch, only when acoustics is not none) with simulator (the acoustics key), namespace and environment_namespace. Its job is to select and delegate to one acoustics backend. Every launch configuration of the parent stays visible to the backend, so undeclared auditory.* keys reach it by prefix.

SelectAction dispatch¶

Keys are Constants.AcousticsSimulator values:

Key	Action
`none`	empty group, no nodes
`arena`	includes `arena/arena.launch.py`

Selected with acoustics:=<key> at the top level.

arena backend¶

arena/arena.launch.py is a shim like human/arena_humansim/arena_humansim.launch.py. It fails with the arena feature auditory install hint when the arena_auditory package is missing. Otherwise it strips the auditory. prefix from every launch configuration that carries it and includes arena_auditory's launch/arena_auditory.launch.py in isolation with:

Arg	Value
`<key>`	`auditory.<key>` of the parent, forwarded as node param `<key>` to every stack node
`debug.map_source`	passed through unchanged when set
`namespace`	task generator node namespace, the stack nodes live below it
`env.ns`	absolute env namespace
`hearing`	`robot.hearing`, `srp` and `seld` default `array.spec` to `four_mic`

arena_auditory.launch.py declares the launch-facing keys itself, with their descriptions, from arena_auditory/params.py. Empty values keep the node default. The nodes, their parameters and topics are documented in the arena_auditory README.

Node side¶

task_generator/simulators/acoustics/ mirrors the human registry: AcousticsSimulatorRegistry yields a BaseAcousticsSimulator per key.

Member	Purpose
`requires_map_server`	the arena backend sets it, the propagation node builds its acoustic scene from the map topic
`displays()`	env-level RViz displays
`robot_displays(robot)`	per-robot RViz displays, the robot's `_acoustics` manifest entry
`plugins()`	RViz panels and tools (`AdapterPlugin`), carried by the viz manifest
`recorded_topics()`	`RecordedTopic` rows, published latched on `state/recorded_topics` with the hearing rows
`pedestrian_hearing(human)`	a `PedestrianHearing` the backend feeds with `on_heard(agent_id, kind, source_id, audible)`
`emit_sounds(sounds)`	the sounds module's `SoundEmission`s of one tick, a no-op by default

simulators/acoustics/arena/ is the only task_generator code importing arena_auditory or arena_auditory_msgs, always inside methods, so importing task_generator never needs the feature. It publishes ContinuousAudioSourceState on continuous_audio_sources, maps WAV variants to wav_loop, and repeats a sounding source that stopped being listed INACTIVE_REPEATS times as inactive.

Layout¶

launch/acoustics/
|-- acoustics.launch.py    dispatcher
`-- arena/
    `-- arena.launch.py    install check, shim into arena_auditory's arena_auditory.launch.py

The sounds module needs only the sound catalog in arena_simulation_setup. With acoustics:=none sounds from auditory.static_sounds or task.modules:=sounds exist as entities but nothing hears them, the launch warns once.