Modular AirStack 6/9: vehicles, fleets, airstack.yaml, generic fleet spawner - #393
Open
andrewjong wants to merge 8 commits into
Open
Modular AirStack 6/9: vehicles, fleets, airstack.yaml, generic fleet spawner#393andrewjong wants to merge 8 commits into
andrewjong wants to merge 8 commits into
Conversation
… Isaac spawner RFC #380 scales 1-2, all OPT-IN (no fleet declared = byte-identical behavior, test-pinned): - airstack.yaml: the one hand-edited entry point (release/fleet/sim/modules/ external stack repos); airstack sync resolves it (module sync + stacks/ .external fetch + fleet validation + effective_sources.yaml) - config/vehicles/quad_default: pass-through vehicle (URDF generation future); sensors carry the real driver+sim pairing per RFC #380 §1; per-serial calibration in gitignored config/local/ - config/fleets/{sim_one_default,sim_three_mixed}: robot = vehicle × stack × unit; sim_three_mixed exercises the SPLIT case (robot_3 on lite_offload_global with hosts: {offboard: gcs}) - tools/fleet/resolve_fleet.py (opt-in via FLEET_CONFIG_FILE; legacy resolver untouched and default; parity with it test-proven) + generate_fleet_compose.py (per-robot/per-ground-host services for heterogeneous fleets; homogeneity detection; split placement gives the ground host ENTRY=offboard; bridge hard-gate asserted) - fleet_spawn.py: generic Isaac spawner driven by the fleet file (spawns, vehicle lidar toggle, scene); example_* scripts remain escape hatches - airstack up --fleet / airstack fleet list|generate / harness --fleet; effective-config gains the resolved robot table; explicit env still wins with the override banner - docs/development/fleets.md; configure-multi-robot rewritten fleet-first Unit suite: 238 passed (+30, zero regressions). Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
…ampaigns to fleet size - ready.sh matched only compose replicas (-robot-); fleet-generated services (airstack-robot_N-1) were invisible, so heterogeneous bring-ups timed out at 'robot containers running' with all containers actually up. Ground-host tenants (gcs-robot_N) are excluded — they never run MAVROS/PX4 - pytest_generate_tests now implements what the --fleet docstring promised: campaigns run at exactly the fleet's robot count instead of the --num-robots matrix (a 3-robot campaign against a 1-robot fleet failed on 'no odometry' for robots the fleet never declared) Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
…nerate bridge router configs robot_3's onboard entry (split stack) loads the bridge-derived DDS-router config at .airstack/generated/dds_router.<stack>.yaml — the dir was never mounted and generation was a manual step. Both compose paths (robot-base and the fleet-generated services) now mount it read-only, and fleet generate / up --fleet materialize router configs for every split stack the fleet references. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
- discover fleet-generated robot services (same pattern fix as ready.sh); ground-host tenants excluded - --live --stack <name> captures only robots whose AIRSTACK_STACK_DIR names that stack — one wiring.md describes one stack, not the fleet union (validated live against the running sim_three_mixed fleet: full_default diff shrank from robot_3's whole graph to a single honest environmental finding — the GCS action_relay client count differs between 1-robot and 3-robot deployments; host-grouped wiring per RFC #380 is the refinement) Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
… names fleet_spawn produced multi-style prim names for a 1-robot fleet, deviating from the validated example_one configuration (/World/base_link + PX4Multirotor) — the fleet parity flight stalled at takeoff. Single-robot fleets are now byte-equivalent to the single script's drone config. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
…keoff px4_ready proves EKF/MAVROS signals, but sending takeoff the instant it passes races the drone_safety_monitor's watchdog after slow sim loads — observed as 'Goal was rejected' and 'failed to arm' (three distinct race signatures on fleet runs; manual takeoff on the settled stack accepted fine). Wait up to 45s for state_estimate_timed_out=false; proceed with a warning if unconfirmed. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Fleet-generated services were invisible to 'down' (only the base compose file was loaded), leaving orphaned robot containers busy-looping at ~100% CPU each (ddsrouters spin without a sim /clock) — which tanked the machine (load 70) and caused every subsequent flight test to time out regardless of dispatch path (legacy control reproduced it). down now includes the generated overlays and passes --remove-orphans as a backstop. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Stack 6/9 · base:
rfc/s5-macvo-flatten-splitHeterogeneous AirStack scales 1–2 (RFC #380): fleets separate who exists from how each one flies.
What's inside
config/fleets/*.yaml): robot = vehicle (config/vehicles/) × stack × unit, plushosts:placement of split-stack offboard halves onto ground hosts.airstack up --fleet <name>validates, derivesNUM_ROBOTS, and selects the generic Isaac fleet spawner (fleet_spawn.py).airstack fleet list|generate: per-robot compose generation for heterogeneous fleets (homogeneous keepsdeploy.replicas); bridge router configs auto-generated per fleet;.airstack/generated/mounted into robot containers.tools/fleet/resolve_fleet.pyviaFLEET_CONFIG_FILE): containers resolve their whole fleet entry; without a fleet, the replica-based path is byte-identical.airstack syncreconcilesairstack.yaml.--fleetoption;doctor --livehandles heterogeneous fleets;airstack downsees generated compose services (fixes orphaned fleet containers); flight harness waits for the state-estimate watchdog before commanding takeoff.Validation
global_plancrossing the DDS router.NUM_ROBOTSmulti-robot still green; single-robot fleets keep the historical single-drone prim names.🤖 Generated with Claude Code