OPENAI BUILD WEEKTEAM VIDYUT

VIDYUT means electricity. Like electricity flowing through every circuit, VIDYUT moves through every layer of a machine—geometry, electronics, control and environment—to uncover failures before physical trials.

Find integration failures before the physical trial.

A virtual proving workflow for autonomous machines—combining geometry, electronics, repeatable failure scenarios and engineering evidence.

BUILT BYSiddharthReshmaAbhinay

Physical 19-DOF humanoid robot used by the VIDYUT team

PHYSICAL MACHINE / 19 DOFGEOMETRY × ELECTRONICS × CONTROL

01 / THE PROBLEM

CAD describes the body.
It does not prove the machine.

Critical failures emerge where geometry, electronics and controller behaviour meet. Finding them during physical integration costs time, hardware and sometimes safety.

01Integration faults appear on the physical bench.

02Failure testing is costly, risky and difficult to repeat.

03Evidence is fragmented across disconnected tools.

02 / INTENDED USERS

For teams validating autonomous machines.

BUILDERS

Robotics & embedded teams

Developing drones, rovers, humanoids and physical-AI systems.

VALIDATORS

Test engineers

Responsible for repeatable failures, thresholds and reviewable evidence.

EXPLORERS

Smaller autonomy & space teams

Learning before costly trials without an enterprise-scale proving ground.

03 / THE WORKFLOW

One continuous engineering loop.

01DescribeGeometry + electronics
02ChallengeEnvironment + failures
03ExecuteSIL + controller bridge
04ProveTelemetry + evidence

STEP 01 / MACHINE SETUP

Give the model a body, electronics, and interfaces.

Geometry and wiring are separate sources. VIDYUT combines them into one reviewable machine manifest before any test is allowed to run.

REFERENCE MACHINES

Start from a working adapter

Drone is the primary hackathon demonstration.

HUMANOID ADAPTER / LIVE HARDWARE

19‑DOF humanoid · right gripper channel

Run the NUCLEO-G474RE/PCA9685 hardware proof from inside this working adapter. The virtual gripper changes only after the controller acknowledges the physical PWM command.

VIDYUT.GRIPPER.V1

0.0° ACKNOWLEDGED

Controller not connectedNUCLEO-G474RE / PCA9685 / CH0

Robot is secured, servo rail is 6.0 V, grounds are common, and I can cut physical power immediately.

Waiting for VIDYUT gripper controller…

MACHINE PACKAGE

Import engineering sources

ZIP / JSON / YAML / URDF / SDF / STEP / glTF / KiCad

+Drop a CAD, robot description, netlist, or VIDYUT packageVIDYUT ZIP packages are CRC-checked and expanded in the browser. Reference PDFs are recorded; structured engineering values still require confirmation.

MECHANICAL MODEL

Geometry and physical properties

Engineer confirmed

Machine name Family Geometry format Units Total mass (kg) Friction coefficient Drag coefficient Centre of mass XYZ (m)

ELECTRONICS MANIFEST

Components

ID	Component	Role	Input V min	Input V max	Load A	Bus / address	Confirmed

WIRING EDITOR

Power and signal connections

Kind	From component.pin	To component.pin	Signal	Frequency Hz

For large assemblies, upload the JSON manifest or KiCad netlist, then use this editor only for corrections.

STEP 02 / SCENARIO BUILDER

Choose a world, then compose the failures.

Select any compatible tests. The deterministic executor owns timing and results; GPT-5.6 can draft a constrained plan.

ENVIRONMENT

Default and space environments

DEFAULT TEST LIBRARY / 25

Select one or many tests

Test nameExecutable fault hookSeverity %Duration seconds

Measurable assertion

TRUTH POSITIONX 0 / Y 0 / Z 0

CONTROLLERNOMINAL

ALTITUDE AGL0.0 mVERTICAL SPEED0.0 m/sMOTION CONSTRAINT3D FLIGHT

TruthTargetDrag to orbit / wheel to zoom

Safety margin100%

Inside envelope

Estimate error0 m

Truth vs controller

Attitude0 deg

Configured limit

Power reserve100%

Live energy model

TELEMETRY

Deterministic evidence capture

SafetyBatteryError

Find integration failures before the physical trial.

CAD describes the body.
It does not prove the machine.

For teams validating autonomous machines.

Robotics & embedded teams

Test engineers

Smaller autonomy & space teams

One continuous engineering loop.

Give the model a body, electronics, and interfaces.

Start from a working adapter

19‑DOF humanoid · right gripper channel

Import engineering sources

Geometry and physical properties

Components

Power and signal connections

Choose a world, then compose the failures.

Default and space environments

Select one or many tests

Deterministic evidence capture

Engineering evidence report

Batch scenario coverage

Find integration failures before the physical trial.

CAD describes the body.It does not prove the machine.

For teams validating autonomous machines.

Robotics & embedded teams

Test engineers

Smaller autonomy & space teams

One continuous engineering loop.

Give the model a body, electronics, and interfaces.

Start from a working adapter

19‑DOF humanoid · right gripper channel

Import engineering sources

Geometry and physical properties

Components

Power and signal connections

Choose a world, then compose the failures.

Default and space environments

Select one or many tests

Deterministic evidence capture

CAD describes the body.
It does not prove the machine.