Skip to hardware brief
PICI
Hardware brief Tokyo / 2026
pici.co ↗

PICI / Physical intelligence / Technical brief 01

§H01 System boundary

Humanoid hardware is a systems problem.

The useful question is not whether a robot resembles a person. It is whether its body can sense, move, recover, and work safely inside the constraints people already built.

01

A compact platform is a chain of coupled choices: load path, energy, control, and operator trust.

Humanoid system boundary diagram A monochrome front elevation of a humanoid robot with callouts for sensing, actuation, energy, and contact. FIG. 01 / PLATFORM BOUNDARY REFERENCE ELEVATION / NTS SENSOR FIELD CONTACT / HAND STATE ESTIMATION ENERGY + COMPUTE GROUND CONTACT Z + 02.1 Z 00.0
Plate 01 Working abstraction / system boundary

§H02 Taxonomy

DOC / PICI-HW-001 REV / 01

System map

Start with the constraints.

Humanoid hardware is legible when the body is read as three linked layers, not as a single machine.

01

Structure

Load paths, joints, range, service access, and the geometry that defines where a body can work.

BODY / LOAD
02

Actuation

Motors, transmissions, brakes, and compliance convert energy into controlled contact.

JOINT / FORCE
03

Intelligence

Sensors and control software turn uncertain surroundings into bounded, recoverable action.

STATE / ACTION

§H03 Mobility

LENS / LOCOMOTION REF / 02

Mechanism plate A

The leg is a force-routing problem.

A leg has to carry the body, absorb error, and place the foot without asking the environment to become perfectly known.

READOUT Fixed geometry → controlled contact

The mechanism earns useful mobility by managing force at the foot, not by adding degrees of freedom alone.

Lower-body force routing diagram A monochrome leg diagram showing the hip, knee, ankle, actuator path, and ground reaction vector. A / FORCE ROUTING HIP ACTUATOR KNEE JOINT ANKLE / CONTACT GROUND REACTION REFERENCE PLANE / Z = 00 LOAD
Fig. A Joint chain / sagittal abstraction
Primary load
Body mass + external contact
Control surface
Foot placement and torque
Failure mode
Slip, saturation, or recovery loss

§H04 Manipulation

LENS / CONTACT REF / 03

Mechanism plate B

The hand is a sensing surface.

Useful manipulation depends on knowing when an object has been found, loaded, and released. Force is part of the signal.

READOUT Position → contact state

A reliable grasp is a closed loop between geometry, friction, force, and the next task state.

Manipulation contact-state diagram A monochrome gripper abstraction showing object contact, force vectors, and a closed control loop. B / CONTACT STATE FORCE SENSOR OBJECT LOAD CONTACT EVENT FEEDBACK SENSE CLASSIFY ADJUST HAND / OBJECT / TASK
Fig. BEnd-effector / contact abstraction
Primary signal
Contact onset + force
Control surface
Grip, slip, and release
Failure mode
Misalignment or object loss

§H05 Decision matrix

LENS / TRADE-OFFS REF / 04

Spec lens

Every capability spends a resource.

Hardware choices become credible when their costs stay visible alongside their intended benefit.

Representative humanoid hardware trade-offs
DecisionAddsCostsPICI lens
More degrees of freedomReach + task coverageMass + control burdenUse only where contact requires it
Higher peak torqueAcceleration + recovery marginEnergy + thermal loadSize for the task envelope
Softer complianceContact tolerancePosition precisionPair with measured force
More onboard computeLocal response + resiliencePower + heatKeep the control loop close

§H06 Principles

FIELD / INDUSTRIAL BASE / TOKYO

Build for the work that remains when the demo ends.

01 Make the environment legible.

02 Keep failure modes recoverable.

03 Treat operator trust as a system requirement.

§H07 Next discussion

A hardware brief is a starting point.

For technical questions, pilot discussions, or senior engineering conversations, contact PICI.

founders@pici.co