PYTHON · NUMPY

Capstone: Design Ada, a Companion Humanoid

Pull all six modules into one coherent humanoid design for a named work-and-play role, and prove the design is internally consistent, strong and stable enough to work, articulate and efficient enough to play.

01Challenge

Design Ada, a companion humanoid for a senior-living community: on shift she carries trays and tends a light repair bench (work); off shift she leads a movement class and plays games (play). Set every parameter of her design so that all seven requirements pass, each one tracing back to a module you have already mastered.

02Model

Everything in this course was one decision inside a larger design. Now you make them together. Meet the brief: Ada, a companion humanoid for a senior-living community. On shift she does real work, carrying trays, tending a light repair bench, moving safely among residents. Off shift she plays, leading a seated movement class, running gentle forms, playing games that keep people sharp and delighted. One body has to do both, and the two roles pull on each other: the strength and stability that make her useful, the articulation and expressiveness that make her a joy, the safety and economy that let her share the floor with the staff she amplifies rather than replaces. That is the whole point of designing embodied AI: not any single subsystem, but a coherent whole that earns its place at work and its welcome at play. Your capstone is a design review. Set every parameter of Ada's design and satisfy every requirement, each one tracing back to a module you have already done. When they all pass together, you have designed a humanoid.

03Guided practice

Each failing requirement points back to a module. Fix them in the design dict: M1 Body. Raise knee_stiff so stiffness * crouch >= (body+payload)*g (actuator). Widen foot_half to at least com_offset (support base). Reach is met when torso_h + arm_reach >= 1.10 (kinematics). M2 Move. Raise step_reach to at least bump_v / sqrt(g/torso_h), the capture point (stepping). M3 Behavior. Keep demo_budget >= task_feats so the serving skill is learnable on the backbone (the bridge). M4/M5 Play. Give her dof >= 20 for expressive movement and eff_target >= 0.5 (dance / agency). M6 Work. Set alloc so the human-robot split beats either alone AND leaves the care tasks (comfort-care, games) with people (amplify, not replace).

04Feedback

PASS WHEN PASS when every requirement is met: Ada's design is internally consistent, strong and stable enough to work, articulate and efficient enough to play, and staffed so she amplifies the people around her.

If your run fails, check:

Body requirements fail: check the actuator (stiffness x crouch vs weight), the foot width vs the CoM offset, and the arm reach vs the 1.10 m shelf.
The move check fails: her step must reach the capture point, step_reach >= bump_v / sqrt(g/torso_h).
The work split fails: give the robot the non-care tasks, keep comfort-care and games human, and confirm the makespan beats all-human and all-robot.

05Retrieve & space

Which single equation sizes the actuator that holds Ada at her working crouch? (stiffness x deflection = mass*g).
Where must Ada step to arrest a nudge, and how does it scale with her height? (the capture point, bump_v / sqrt(g/height)).
Why keep the care tasks with people even when the robot is faster? (amplify not replace; tacit, human-valued work).

06Mastery & project

Integrate all six modules into one coherent, internally-consistent humanoid design for a named work-and-play role, and prove it meets every requirement at once.

Ada is the course artifact: a humanoid specified end to end, body to behavior to deployment. Take the design into the live tools and make it move.

GO FURTHER · THE LIVE PLAYGROUND

Teach Ada's walk in Forge ↗ — evolve a gait on-device for the body you just specified
Pilot the biped ↗ — drive a full-physics humanoid and feel the design come alive

← Amplify, Not Replace Back to course →

PYTHON · NUMPY

Capstone: Design Ada, a Companion Humanoid

The capstone design review. Tune Ada's design dict until every requirement, one per module, passes.

You read

the arrays and values already in scope

You change

the code you write in each cell

Fixed

the dataset and the checks that grade you