Skip to content

motor_control_walkthrough.md

Latest commit

 

History

History
90 lines (70 loc) · 3.39 KB

motor_control_walkthrough.md

File metadata and controls

90 lines (70 loc) · 3.39 KB
Raw

Motor-control walkthrough

This walkthrough opens examples/motor_control_dual_jacobian/motor_control_dual_jacobian.piper in the editor and takes you through the moving parts: nodes, links, stages, modes, and the per-pin Bus pattern that lets one node span multiple stages.

The graph is regenerable from C++:

./build/tests/fixtures/piper_build_motor_dual_jacobian examples/motor_control_dual_jacobian/motor_control_dual_jacobian.piper

For an end-to-end engine run with a CSV/plot tail, see examples/filter_demo/, examples/am_radio/, or examples/pid_demo/ (each has its own runner + plot script).

Dual-channel with feedback: motor_control_dual_jacobian.piper

constant<float>(target_x) ----+
                              +-> jacobian_2x2(jacobian) -> motor(motor_a) -> probe<float>(pose_a)
constant<float>(target_y) ----+                          -> motor(motor_b) -> probe<float>(pose_b)

Open it:

./build/app/piper-editor examples/motor_control_dual_jacobian/motor_control_dual_jacobian.piper

The pipeline encodes a small inverse-kinematics-style flow: two cartesian-space targets feed a 2x2 jacobian whose two outputs drive two motors. Each motor's measured pose comes back as feedback.

What to look at:

  • Two stages, two roles. Setpoint flow lives in control; measured pose lives in feedback. Switch between them via the toolbar combo or right-click on empty canvas.
  • The Bus pattern. The motor node sits in the control stage by default, but its measured output is tagged with a per-pin override of [feedback]. Switch the display stage to feedback and you'll see most of the motor body dim while the measured pin stays bright -- and vice versa for control. This is how V2 represents a node whose pins live in different stages without duplicating the node.
  • Per-stage pipeline DAG. Inside each stage the dataflow is acyclic: control is purely setpoint -> jacobian -> command; feedback is purely motor.measured -> probe. The cycle (motor's command and measured belong to the same physical motor) only closes when you read the graph as a whole.

Try the play button

Hit play in the toolbar (between < and >). The display stage cycles every 2 seconds, alternating between control and feedback. Useful for visualizing a multi-stage pipeline at a glance.

Tweaking the example

Open the file, edit, save:

  • Click target_x, change value from 1.0 to 0.5. Ctrl+S saves over the example.
  • Add a node: right-click empty canvas -> Add node -> any type. Drag from a pin to wire it in.
  • Rename motor_a. Press Enter in the inspector's name field. The header updates immediately.
  • Toggle a node's mode: right-click the node -> Set mode (default) -> disable. Body fades; the engine skips the node's compute() on tick (its outputs hold their last value). See docs/architecture.md's "Modes" section.
  • Fan-in is allowed. Drag a second source onto an already-connected input. The editor permits it; whichever producer ran most recently in tick order wins (typically the later-stage one).

Lints

Open the Problems tab in the right panel. With the bundled example loaded cleanly the tab shows No problems.. Add a node and leave it disconnected -- the tab turns orange and lists:

* node 'NewNode' is disconnected
* input 'NewNode.in' has no source

Click any lint row to jump to the affected node.