Skip to content

golem_slam: stop Nav2's inflation-radius errors (local 0.3 m, start-up footprint, global 0.22 m) - #35

Merged
maxconwa merged 2 commits into
mainfrom
golem_slam/inflation_and_startup_footprint
Nov 19, 2026
Merged

maxconwa merged 2 commits into
mainfrom
golem_slam/inflation_and_startup_footprint

Conversation

@maxconwa

Copy link
Copy Markdown
Contributor

golem_slam: stop Nav2's "inflation radius smaller than the inscribed radius" errors

Nav2 logged The configured inflation radius (0.200) is smaller than the computed inscribed radius (...) of your footprint on both robots. It came from three places:

Where Cause Fix
Local costmap, steady state (about once a second) dynamic_footprint adds the arms. Arms hanging behind the pelvis push the inscribed radius to about 0.25 m, past the 0.2 m inflation. Local costmap and MPPI ObstaclesCritic at 0.3 m (8683be9, from golem_slam/local_inflation, rebased on main)
Local costmap, a 3 s burst at start-up frame_task_server's start-up arm moves swing the arms behind the pelvis: 0.33–0.42 m on the H1-2 dynamic_footprint publishes the body polygon alone until the first /left_ee_pose, which frame_task_server publishes once its start-up moves are done. The robot can't drive before then. New parameter arms_ready_topic (empty: the arms count from the start).
Global costmap, once at start Nav2 pads the fixed body footprint by its default 0.01 m footprint_padding, so the inscribed radius is 0.21 m Global inflation 0.22 m. That's 2 cm more than PR #15's 0.2 m, so the planner still goes through the same narrow gaps.

Measured on the Mac (MacBook Air M3, navigation stacks at real time, kitchen 9/9/42)

Inflation errors RTF Nav2 goal ahead / back to start
H1-2, main about 1 per second 1.00 succeeded / succeeded
H1-2, 0.3 m local only 13, all at start-up 1.00 succeeded / succeeded
H1-2, this PR 0 (start-up, four named_config moves, both goals) 1.00 succeeded / succeeded
G1, this PR 0 1.00 succeeded / aborted 1.36 m short

The G1's goal back to start has aborted short on every run with this setup (0.62–0.85 m before this PR). This run doesn't show whether the change makes it better or worse.

Testing

  • golem_slam tests: 71 pass, including a new test_the_arms_count_only_after_the_controller_starts_up and the example-config check.
  • Not run: the x86 closed-loop gates (tests/sim/test_sim_stack.sh --h12/--g1, tests/g1/test_sim_nav.sh). Their goal clearances should still hold with 0.3 m local / 0.22 m global inflation, but that hasn't been checked.

🤖 Generated with Claude Code

https://claude.ai/code/session_015EG7pDNJLWR31eQmyKaWhr

Correll-Humanoid and others added 2 commits November 19, 2026 15:06
…bed radius

The fixed footprint's rear edge sits at x = -0.20, so its inscribed radius
equals the 0.2 m inflation exactly. dynamic_footprint adds the arm links'
padded boxes, and arms hanging behind the pelvis push the rear edge back
(~0.246 m in the H1-2 sim), so Nav2's local costmap logged "inflation radius
smaller than the inscribed radius" on every footprint update, about once a
second.

Both robots' local costmaps, and the MPPI ObstaclesCritic that reads them,
now inflate by 0.3 m. The global costmaps stay at 0.2 m, so the planner still
plans through the same narrow kitchen gaps.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
… is up; global inflation past Nav2's padding

dynamic_footprint publishes the body polygon alone until the first message on
arms_ready_topic (default /left_ee_pose, which frame_task_server publishes once
its start-up moves are done): those moves swing the arms behind the pelvis
(an inscribed radius of 0.33-0.42 m on the H1-2), past the local costmap's
0.3 m, while the robot cannot drive yet.

Both robots' global costmaps inflate by 0.22 m: Nav2 pads the fixed body
footprint by its 0.01 m footprint_padding, so its inscribed radius is 0.21 m,
and 0.2 m logged the same error once at start.

On the Mac (H1-2 and G1 navigation stacks at 1.00x): no inflation error over
start-up, named_config moves and Nav2 goals, from 13 (H1-2) with the 0.3 m
local inflation alone.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_015EG7pDNJLWR31eQmyKaWhr
@maxconwa
maxconwa merged commit 22be757 into main Nov 19, 2026
2 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants