HIRO: A Humanoid Research Platform for Real-World Robotics
How Puneet Melwani and his team built a humanoid robotics research platform to explore autonomous manipulation, human-robot collaboration, and adaptive task execution across manufacturing, warehouse, and service environments.
What Is HIRO?
HIRO — Humanoid for Robotic Operations — is a humanoid robotics research platform engineered by Puneet Melwani and his team to explore how humanoid form factors can operate in environments that have historically resisted automation. Where fixed-purpose robots excel at repetitive, structured tasks, HIRO was built as a research vehicle for tackling variability: irregular objects, unpredictable layouts, and close proximity to human coworkers in real-world assembly lines, warehouses, and service settings.
The platform's defining characteristic is its adaptability — designed to operate tools, fixtures, and workstations built for people without redesigning the workspace around the robot. That design philosophy is what makes HIRO a versatile research platform, capable of exploring applications across multiple environments without bespoke retooling for each one.
Unstructured Environments Don't Wait for Perfect Conditions
Most automation succeeds by constraining the problem — fixturing parts, standardizing inputs, isolating the robot from people. HIRO needed to succeed without those constraints.
Adaptive Manipulation
Tasks requiring variable grip strategies, adaptive motion planning, and real-time decision-making that traditional fixed-axis robots cannot replicate.
Irregular Object Handling
Parts and items that vary in shape, weight, and orientation — the opposite of the standardized inputs most industrial robots expect.
Operating Alongside Humans
Safe, intuitive coexistence with human coworkers in shared workspaces, without fixed guarding or rigid choreography.
A Research Platform Built for Adaptability
Every subsystem in HIRO exists to support research into adaptive, autonomous operation in unpredictable environments.
Multi-Axis Upper Body
A high degree-of-freedom upper body gives HIRO a motion range sufficient to interact with tools and workstations designed for human operators.
Depth-Sensing Vision
Stereo and depth cameras build a real-time 3D understanding of the workspace, enabling object recognition and grasp planning on the fly.
Force/Torque Sensing
Joint and end-effector force feedback let HIRO modulate grip strength in real time, handling delicate and heavy items with the same hands.
Modular End-Effectors
Interchangeable hand and gripper modules let HIRO switch between precision pinch grips and full-hand grasps for different task types.
Gesture-Based Interface
Human operators can direct, pause, or redirect HIRO using natural gestures, reducing the need for specialized programming interfaces on the floor.
ROS Architecture
A modular ROS-based control stack coordinates perception, planning, and actuation, and lets new skills be added without re-architecting the system.
From Concept to Cross-Disciplinary Build
HIRO was never a single-discipline project. Bringing a humanoid platform from concept to working prototype required mechanical, electrical, software, and AI engineering to move in lockstep rather than in sequence — a structural lesson and decision made early in the program.
Mechanical Engineering
The skeletal structure, joint design, and actuator selection had to balance strength, weight, and range of motion — without exceeding power budgets that would make the platform impractical to run continuously.
Electrical Engineering
Power distribution, motor controllers, and sensor wiring needed to fit within tight volumetric constraints while keeping thermal loads manageable across 24/7 operation cycles.
Software & AI
Perception, grasp planning, and motion control software were developed iteratively against real test rigs — not simulation alone — so that lessons from early hardware failures fed directly back into the control architecture before the next build cycle.
Each iteration of HIRO moved through this same loop: build, test on real tasks, instrument failures, redesign. That iterative discipline — rather than any single breakthrough component — is what got the platform from early concept to a robot capable of independent deployment across three distinct environments.
Three Environments, One Platform
HIRO's adaptability as a research platform translated into exploration across sectors that don't typically share robotic platforms.
Deployment environments: manufacturing, warehouse, hospitality
Productivity gain over manual baseline in deployed tasks
Continuous operating capability without shift changeovers
Open architecture enabling rapid experimentation and iteration
An Early Mover in the Humanoid Robotics Wave
HIRO's development began well ahead of the current wave of commercial humanoid platforms now drawing attention from major technology and automotive companies. Building a humanoid research platform before that wave meant working without the benefit of mature commercial supply chains, off-the-shelf actuator ecosystems, or established design patterns — every subsystem decision had to be validated independently rather than borrowed from a known-good reference design.
That early positioning has paid off in practical terms: the research lessons on actuator reliability, thermal management, and autonomous task planning that informed HIRO's design are now directly applicable as the broader humanoid robotics market matures and demand accelerates across logistics, manufacturing, and service industries. The platform continues to serve as a testbed for new approaches to human-robot collaboration and autonomous manipulation.
Related Case Studies
See how other robotics platforms we've built are solving real operational problems.
Exploring Humanoid Robotics for Your Operation?
Tell us about the tasks that don't fit traditional automation, and we'll help you scope what's realistic today versus what to plan for next.