COMPAS FAB 2.0 release#458
Open
gonzalocasas wants to merge 561 commits into
Open
Conversation
…into wip_process
…into wip_process
…into wip_process
…into wip_process
…into wip_process
Strip the redundant type from numpydoc Parameters/Attributes entries across compas_fab.robots, relying on the signature/property annotations that mkdocstrings already renders (it shows the type from the signature, so the docstring copy only duplicated it and drifted — and the `[`X`][path]` cross-ref syntax rendered as raw markdown in those sections anyway). Where the type lived only in a `# type:` comment, convert it to a real inline annotation so it still renders: Duration (time_.py) and Inertia (inertia.py). Returns/Yields blocks and the TargetMode enum members are left as-is (no signature slot to move those types to). Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Strip the redundant type from numpydoc Parameters/Attributes entries across compas_fab.backends where the signature/property already carries the annotation that mkdocstrings renders. Removes a lot of leftover RST `:class:` / `:obj:` type markup that rendered as raw text in those sections. Entries whose signature is not yet annotated are left untouched (the type would otherwise be lost) and will be handled by adding the annotation. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Same numpydoc cleanup as the robots/backends commits, applied to the remaining modules (scene base objects, ghpython scene object, utilities). Strip the docstring type where the signature already carries it. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
…e too Finish the numpydoc type move for entries the automated strip had to leave because the signature/attribute carried no annotation to fall back on. Add the missing annotations and drop the docstring copy: * PlannerInterface.__init__ `client` -> Optional[ClientInterface] * RosClient.__init__ host/port/is_secure/transport * register_models_into_cell (ghpython.cell_builder) robot_cell/tools/rigid_bodies Also fix docstring bugs surfaced along the way: * AnalyticalSetRobotCell.set_robot_cell documented its params under an `Attributes` heading (should be `Parameters`). * AnalyticalKinematics documented `base_frame_offset`/`flange_frame_offset` attributes that don't exist (the attributes are `base_frame`/`flange_frame`). * inverse_kinematics_spherical_wrist documented `frame` (the param is `target_frame`). Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The lazy-loader guard only bound the `pybullet` name when the module was not yet imported; when something had already imported the real module (e.g. the user `import pybullet` in the same Grasshopper script), being in `sys.modules` never created a local binding, so `pybullet.connect(...)` raised `NameError`. Bind the real (cached) module in that branch. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
…pute `ensure_value_list` / `ensure_boolean_toggle` did AddObject + AddSource inline during the running solution. Those mutations only commit when the solution finishes, so the follow-up solves still read SourceCount == 0 and kept adding more widgets (observed 2-4 stacked value lists per target_mode input, and duplicate load_geometry toggles). Move the whole creation into the ScheduleSolution callback (between solves, where AddSource commits) and guard with a sticky pending flag so only one is ever scheduled - the same deferred+tracked pattern ensure_dynamic_value_list already uses. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The component reads the RobotCellObject's child scene objects directly, but those are only built when a state is applied via update(). With no cell_state wired, update() was skipped and the draws hit a None object. Default to robot_cell.default_cell_state() when none is wired, restoring the previous behaviour of drawing the cell in its default configuration. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The state held by the planner's client is whatever was last set internally (typically a stale/default one), so surfacing it as an output invites wiring a wrong state into planning. Output only robot_cell and robot_model; build the state explicitly via Default Cell State / the Attach* components instead. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
RigidBodyLibrary.floor() built a flat plane at Z=0, which collides with the robot base under distance-0 collision checking, making every configuration fail. A base's collision geometry can dip several mm below the origin (the UR5 base_link_inertia reaches about -7 mm), so add a `clearance` parameter that sits the floor below Z=0, defaulting to 1 cm. touch_links remains the robot-agnostic alternative (documented). Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Demonstrates collision-checked IK against a static rigid body: the UR5 IK now plans with a `RigidBodyLibrary.floor()` present in the cell. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
…onChain Promote ActionChain's private `_cell_signature` helper to a public `RobotCell.structural_signature()` method - a stable fingerprint (robot model name + sorted tool/body ids -> sha256) of a cell's structural identity. ActionChain now calls the method; the duplicated function and its `hashlib` import are removed. The MoveIt Planner GH component reuses it to decide when the planning scene needs re-uploading. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
MoveIt's planning scene is defined by the URDF/SRDF loaded into move_group, so the cell can only come from ROS - splitting "load the cell" from "make the planner" was artificial and needed a ros_client pass-through hack to enforce load-before-plan ordering. The MoveIt Planner now loads the cell from ROS itself and the separate Load Robot Cell From ROS component is removed. MoveIt Planner inputs: ros_client, tools, rigid_bodies, load_geometry (auto-toggle), options, reload. Outputs: planner (first), robot_cell, detected_distro. The base cell is fetched once and cached (reload to refetch); the scene is re-uploaded only when the cell's structural signature changes (or on reload / planner rebuild). Advanced load parameters (urdf/srdf param names, HTTP file server URL) move to a new MoveIt Planner Options component, backed by a tiny `MoveItPlannerOptions` class in compas_fab.ghpython - a plain (non iterable) class so it crosses a Grasshopper wire as one object rather than being exploded like a dict/namedtuple. The default flow is just RosClient -> MoveIt Planner. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The ROS Client component now auto-creates a Boolean Toggle on its `connect` input when nothing is wired, defaulting to False so a freshly-dropped client does not try to connect until the user flips it on. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
When planning fails with no (partial) trajectory, `trajectory` is None but `_viz()` still called `trajectory.to_frames_and_polyline(...)`, raising a NullReferenceException instead of just flagging the component. `_viz` now returns empty geometry for a None trajectory, on both the cached-replay and just-planned paths. Applied to Plan Motion and Plan Cartesian Motion. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
MoveItPlanner now implements the CheckCollision interface - the same planner.check_collision(state) as PyBulletPlanner - backed by MoveIt's /check_state_validity service (moveit_msgs/GetStateValidity). It does a stateless collision + constraint check of the state against the planning scene currently loaded into move_group (no planning, no IK) and raises CollisionCheckError listing the colliding body pairs. Adds the GetStateValidityRequest / GetStateValidityResponse message wrappers and mixes MoveItCheckCollision into MoveItPlanner. Live-tested against a running MoveIt: a folded UR config reports the colliding link pairs; a zero config returns collision-free. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
…state When inverse kinematics fails, the component now (if the planner supports collision checking) runs check_collision on the start state and, when it is in collision, appends the colliding pairs to the error - turning an opaque "no IK solution" into an actionable cause. Planners without the feature are handled gracefully (BackendFeatureNotSupportedError / AttributeError), and the diagnostic never masks the original error. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
…ners Consolidate the IK collision hint into a single `collision_diagnostic` helper in `compas_fab.ghpython` and reuse it from Inverse Kinematics, Plan Motion and Plan Cartesian Motion. On a failure it runs a concise collision check on the start state - listing the colliding pairs (capped, e.g. `a<->b, +2 more`) or noting the state is collision-free / likely unreachable - and never masks the real error. Planners without collision checking degrade gracefully. Each of the three components also gains a `debug_info` output that mirrors the warning/error messages as one string, so they can be read in a wired panel instead of only the component balloon. The IK error message is now just `str(e)` (the backend message is already descriptive). Also guards Plan Cartesian Motion's `fraction` read against a None trajectory. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
…ctory guards Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
JointTrajectory.to_frames_and_polyline gained an optional `target_mode` (defaults to TargetMode.ROBOT, coercing None -> ROBOT), threading it into the per-point forward kinematics instead of hardcoding ROBOT. The Plan Motion and Plan Cartesian Motion components pass the target_mode of their target / waypoints, so the previewed planes/polyline match what was planned (e.g. the tool-tip path for a TOOL-mode target). The mode is cached with the trajectory so replays preview the same frames. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Co-Authored-By: Claude Opus 4.8 <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
4 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.
This PR superceedes #456 and contains the entire Project Theseus + ROS 2 / MoveIt 2 support.
What type of change is this?
Checklist
Put an
xin the boxes that apply. You can also fill these out after creating the PR. If you're unsure about any of them, don't hesitate to ask. We're here to help! This is simply a reminder of what we are going to look for before merging your code.CHANGELOG.mdfile in theUnreleasedsection under the most fitting heading (e.g.Added,Changed,Removed).invoke test).invoke lint).compas_fab.robots.CollisionMesh.