[BioCam]

The Problem

The Mantis arm (two arms driving one head, plus Z) has to be brought up after every power-on: energised, told where it is, homed. Each unit's arm is calibrated at the factory, so values from one unit put another's head in the wrong place. And a stop has to actually halt the device: abort what is queued, stop every motor and park, before the link closes.

PR Content/Solution

• mantis.arm (v4), built when the absolute encoder board answers. Its calibration (each arm motor's offset, encoder reference and encoder correction) comes from the unit's configuration folder or a saved file, never from a recording; without it, bringing the arm up refuses before anything is sent.
• setup() brings the arm and Z up as in the recorded v4 startup: energise, read where the arms are from the absolute encoders, set the following-error limits, move to idle, home Z. It only starts near idle; anywhere else it refuses and unpowers the arms. setup(skip_arm=True) reads only, nothing moves.
• stop() halts the device: aborts the sequencer, stops every motor, parks an arm that was brought up (Z down to 20 mm), then releases the clamp. Each step runs even if one before it fails.
• Kinematics (arm_kinematics.py): device units, inverse and forward kinematics and the encoder correction, as pure maths.
• Reads: request_head_position, request_joint_position, request_absolute_joint_position, request_link_angles.
• Resource model: the Z column and the arm's links join the 3D model, placed from a pose read.
• LED bar: the bring-up shows in setup's colour.

Usage

mantis = Mantis(name="mantis", declared_configuration_json="my_mantis.json")
await mantis.setup()  # brings the arm and Z up
await mantis.arm.request_head_position()  # the idle position
await mantis.stop()  # halt, park, release

await mantis.setup(skip_arm=True)  # reads only

On the device

A Mantis v4 on Windows (D2XX), head clear, no chip:

	result
read-only pose	head at (14.983, 31.180) mm from the absolute encoders, idle within 0.02 mm
no arm calibration	refused before anything was sent, nothing moved
cold bring-up at idle	no fault; arms within 10⁻⁷ device units of idle; Z homed twice; the same commands in the same order as the recorded v4 startup
reads after the bring-up	head (15.000, 31.177, −0.003) mm
stop()	abort, three motor stops, Z read by its encoder before the arms moved, Z down to 20 mm (19.996 read back)
stop in the middle of five queued Z moves	the item already playing finished, the rest never played, then the park
second setup in one process, after a USB reconnect, from the saved file alone	the same bring-up each time; the saved file alone sent the same 155 commands in the same order
arm moved away from idle by hand	refused with the head's position, no move sent, the arms unpowered

🤖 Generated with Claude Code