Fix/ivan/motion stability - #3731
Open
leshy wants to merge 130 commits into
Open
Conversation
…hub.com:dimensionalOS/dimos into andrew/feat/ray-tracer-performance-improvements
md-babel-py run README.md regenerates it from the modules' In/Out hints.
MotionPlanner and TrajectoryFollower, python and rust, no longer subscribe odometry. Each tick looks the body up on tf -- the planner `world_frame -> base_frame`, the follower `path.frame_id -> base_frame`, so it controls in the frame the plan is expressed in -- and go2_tf's odometry edge plus the mount tree is what it finds there. The deadman rides the edge's stamp on our own clock: an edge whose stamp has not advanced for the module's age limit (max_map_age_s, max_path_age_s) is a missing pose -- the planner plans nothing, the follower zeroes the twist and names `pose`. Stamp-keyed rather than stamp-compared, because the stamp is the robot's clock and ours need not agree with it. OdomBasePose stays for goal_relay and basic_path_follower; TfPose sits beside it. The rust tf_pose.rs is now PoseWatch, the same watch.
…dy off path plan_body was the path message again, republished at viz_publish_hz so the bridge had a topic to hang the body boxes on. The bridge can return more than one entity per topic, so the boxes now draw on world/path/body from the one path the planner already publishes -- the port, the rate cap, the config knob and the motion_viz_hz blueprint dial go, in python and rust alike.
The follower had two tracks named the wrong way round: `blind` decoded the room the planner encoded in the path timestamps, `hinted` measured it again off `local_map`. From the follower's side the hints ARE in the path, so the stamp-reading law is now `hinted` and the only law besides the `seed` baseline. The map-reading law, its pyo3 class, its parity sequence and the no-map fallback (ceilings bent back into clearance) are deleted, and with them everything the follower only needed to see the map: the `local_map` port, `clearance_for`, the obstacle model, `body_dilate_m`, the track machinery and the `-blind` blueprint. `controller="module:factory"` stays as the override. Tuning fields only that law read leave `ControllerConfig` and the rust `Tuning`; `gait_band`/`command_slew` stay on the body. Parity is unchanged: the renamed law is bit for bit the one that ran.
# Conflicts: # docs/platforms/quadruped/go2/index.md # docs/usage/modules.md
…sforms' into ivan/feat/motion # Conflicts: # dimos/navigation/test_tf_pose.py # dimos/navigation/tf_pose.py # dimos/robot/unitree/go2/constants.py # dimos/robot/unitree/go2/zenoh/zenohconnection.py
The bake/scp/systemctl recipe was four commands in a doc and a service file parked in misc/. Now it is dimos/navigation/motion/deploy/deploy.sh <host>, with the unit file beside it. It checks the robot over ssh BEFORE baking, so a robot that is off costs seconds instead of a build. It stops the unit before scp because Linux will not overwrite a running executable. It verifies four `config loaded` lines scoped to the invocation it just started, not a time window that would count the previous restart too. The bake is unconditional: cargo already does the incremental staleness check and the config is re-emitted every run, so an mtime heuristic here would be a worse copy of that, and would go stale the first time someone added an import it did not know about. Also drops two stale claims from the doc: the installed unit feeds no stdin, and the journal says `config loaded`, not `module started`.
motion.md and motion-deployment.md go back in when the deployment story is settled. The four things that pointed at them now point at the script and the module docs instead, so nothing dangles in the meantime.
start_relay.py came in with the navigation-with-transforms merge and then got unstaged again -- the resolution swept `git restore --staged` across every added path to drop untracked scratch, and took a real file with it. Its test was tracked, so CI ran a test importing a module that was not in the tree. That is the lint/md-babel/self-hosted failure, not three separate ones. The rust failure is the merge commit, not the branch: #3699 added a dep to dimos-module and relocked the five workspaces that exist on main. The motion crates only exist here, so their locks went stale the moment CI merged main in, and --locked refused them. Relocked adapter, movement_manager and go2/tf; the other two were already current.
self-hosted-large-tests collects every test module but only the pyo3 job runs maturin, so a module-level load_extension() raised through collection and took the whole run down -- 5093 deselected tests never got to run because of one module the job was not going to execute anyway. test_voxel_map, test_mls_planner and both test_transformers already guard with pytest.importorskip. This one now does the same, so it skips there and still runs everywhere the crate is built.
leshy
requested review from
Dreamsorcerer,
aclauer,
mustafab0,
paul-nechifor and
spomichter
as code owners
August 27, 2026 19:08
Contributor
|
Too many files changed for review (140 files, 100 file limit). Bypass the limit by tagging |
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
2 participants
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.
Contribution path
Problem
Solution
How to Test
AI assistance
Checklist