Import fix#2783
Closed
jeff-hykin wants to merge 40 commits into
Closed
Conversation
jeff-hykin
commented
Jul 7, 2026
jeff-hykin
commented
Member
…OS/dimos into andrew/feat/navigation-dev-2
…tion-dev-2' of github.com:dimensionalOS/dimos into andrew/feat/navigation-dev-2
…OS/dimos into andrew/feat/navigation-dev-2
…OS/dimos into andrew/feat/navigation-dev-2
jeff-hykin
requested review from
leshy,
mustafab0,
paul-nechifor and
spomichter
as code owners
Contributor
Greptile SummaryThis PR changes the ray-tracing map and 3D navigation stack. The main changes are:
Confidence Score: 4/5The changed mapping and planner paths can produce missing map data or disconnected routes under realistic robot timing and terrain.
dimos/mapping/ray_tracing/rust/src/main.rs, dimos/navigation/nav_3d/mls_planner/rust/src/main.rs, dimos/navigation/nav_3d/mls_planner/rust/src/mls_planner.rs Important Files Changed
Reviews (1): Last reviewed commit: "fixup import" | Re-trigger Greptile |
![greptile-apps[bot]](https://avatars.githubusercontent.com/in/867647?s=60&v=4){kind=small}
Comment on lines
+64
to
71
| let Some((translation, rotation)) = nearest_pose(&self.poses, time_secs(&msg.header.stamp)) | ||
| else { | ||
| warn_throttled!( | ||
| Duration::from_secs(1), | ||
| "No odometry within tolerance of the cloud stamp, dropped a cloud.", | ||
| ); | ||
| return; | ||
| }; |
Contributor
There was a problem hiding this comment.
When lidar and odometry header stamps differ by more than 100ms, this branch returns before update_map runs. At common 10Hz odometry rates that leaves almost no room for skew or batching, so a normal timing offset can make the voxel map stop updating while the robot is moving.
Comment on lines
+174
to
+175
| let points = emit_points(&self.map, voxel_size, Some(cyl), support_min, &live); | ||
| let local = points_to_cloud(&points, out_frame_id, stamp); |
Contributor
There was a problem hiding this comment.
The local map now applies support_min to healthy voxels while the global map does not. With the default support_min=4, a real but isolated occupied voxel, such as a pole or sparse obstacle in the planner cylinder, is omitted from local_map even though it remains in the global map, so local planning can ignore it.
Comment on lines
+237
to
+251
| let z_max = bounds.pose.orientation.z as f32; | ||
| let sensor_z = self | ||
| .latest_start | ||
| .lock() | ||
| .expect("start mutex") | ||
| .map_or(z_max, |(_, _, z)| z); | ||
| let bounds = RegionBounds::capped( | ||
| bounds.pose.position.x as f32, | ||
| bounds.pose.position.y as f32, | ||
| bounds.pose.orientation.x as f32, | ||
| bounds.pose.orientation.y as f32, | ||
| z_max, | ||
| sensor_z, | ||
| self.config.max_overhead_m, | ||
| ); |
Contributor
There was a problem hiding this comment.
This native path feeds the ground-projected start z into RegionBounds::capped, while the Python transformer passes the raw pose z. With max_overhead_m=2.0, a robot at ground level can cap the region near ground_z + 2m and drop valid upper surfaces from the local update, making stairs or elevated traversable areas disappear from the planner graph.
|
|
||
| /// Max traversable vertical step in cells. | ||
| pub fn step_cells(&self) -> i32 { | ||
| (self.step_threshold_m / self.voxel_size).floor() as i32 |
Contributor
There was a problem hiding this comment.
step_cells() now floors the metric threshold, so non-integral ratios become stricter than the configured step_threshold_m. For example, 0.15m with 0.08m voxels allows only one cell of climb, causing routes over steps below the configured metric limit to be treated as disconnected.
Comment on lines
305
to
310
|
|
||
| /// Spare a clearing miss only when a grazing ray skims a recently hit planar | ||
| /// surface. Stale voxels or those with no normal are left to the health checks. | ||
| fn should_spare( | ||
| c: &Voxel, | ||
| ray_unit: Vector3<f32>, | ||
| graze_cos: f32, | ||
| frame: u32, | ||
| recency_window: u32, | ||
| ) -> bool { | ||
| /// Spare a clearing miss when a grazing ray skims a planar surface. | ||
| fn should_spare(c: &Voxel, ray_unit: Vector3<f32>, graze_cos: f32) -> bool { | ||
| match c.normal { | ||
| Some(n) => { | ||
| frame.saturating_sub(c.recency) <= recency_window && ray_unit.dot(&n).abs() < graze_cos | ||
| } | ||
| Some(n) => ray_unit.dot(&n).abs() < graze_cos, | ||
| None => false, |
Contributor
There was a problem hiding this comment.
Removing the recency check lets any voxel with an old fitted normal be spared from grazing-ray clearing forever. A stale planar obstacle that is later seen only at grazing angles can keep its occupied health much longer than before, leaving old geometry in the map.
…/navigation-dev-2
❌ 2 Tests Failed:
View the full list of 2 ❄️ flaky test(s)
To view more test analytics, go to the Test Analytics Dashboard |
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
3 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.