History & Evolution of Humanoid Robots
From the first bipedal walking experiments in the 1970s to today's AI-powered dynamic athletes—explore the remarkable journey of humanoid robotics.
Based on comprehensive academic research including IEEE publications
Three Stages of Humanoid Robot Evolution
The Foundation Era
Basic Bipedal Walking
The nascent advancement of humanoid robots, focusing on fundamental walking capabilities and establishing basic control principles.
Key Developments
- Waseda University pioneered with WAP (Waseda Automatic Pedipulator) series
- WL (Waseda Legged) and WABIAN robots achieved fundamental walking
- WABOT (Waseda Robot) became first full-scale humanoid
- Research focused on bipedal walking functionality and foundational control
Notable Robots
The Integration Era
Sensing & Intelligent Control
Highly integrated systems emerged with rudimentary sensory capabilities, enabling robots to perceive their surroundings and make simple adjustments.
Key Developments
- Honda's ASIMO2000 became first globally influential humanoid
- Integration of sensing and intelligent control technologies
- Robots gained ability to anticipate movements and adjust center of gravity
- Sony's QRIO became first humanoid capable of running
Notable Robots
The AI & Dynamics Era
Highly Dynamic Motion & Intelligence
Breakthrough progress with highly dynamic motion, advanced cognitive capabilities, and AI-driven control enabling complex autonomous behaviors.
Key Developments
- Boston Dynamics Atlas demonstrated human-like perception and decision-making
- Dynamic abilities like navigating obstacles, backflips, and parkour
- Tesla Optimus aims to revolutionize mass production
- Integration of VLMs and LBMs for natural language understanding
Notable Robots
Global Research Landscape
Japan
Morphological Simulation & Service Robots
Pioneered humanoid development with focus on intricate bionic robots for domestic and service applications.
Key Research Institutions
- Waseda University — WABOT, WABIAN series
- AIST — HRP series (HRP-2, HRP-4, HRP-5P)
- University of Tokyo — Kenshiro, Kengoro
Major Companies
United States
Brain Mechanisms & Dynamic Control
Focused on understanding human brain mechanisms and functional simulation, achieving significant advancements in dynamic motion and complex environments.
Key Research Institutions
- MIT — Planar biped, 3D biped, Kogo
- NASA — Robonaut 2, Valkyrie
- IHMC — M2V2, Nadia
Major Companies
China
Rapid Development & Commercialization
Made significant progress with various universities leading research, now expanding into commercial humanoid development.
Key Research Institutions
- Beijing Institute of Technology — BHR series (BHR-5)
- NUDT — Forerunner (2000), "Blackman"
- Zhejiang University — Wukong, table tennis robots
Major Companies
Europe
Cognitive Research & Collaboration
Strong focus on human cognition, artificial intelligence research, and safe human-robot interaction.
Key Research Institutions
- Italian Institute of Technology — iCub, COMAN, WALK-MAN
- German Aerospace Center (DLR) — TORO, Justin
- INRIA (France) — POPPY (3D printed)
Major Companies
South Korea
Advanced Mobility & Competition
Strong presence in robotics competitions with innovative approaches to stability and mobility.
Key Research Institutions
- KAIST — KHR series, HUBO, DRC-HUBO
Major Companies
Control Methods Classification
Traditional Methods
ZMP-Based Control
Zero Moment Point method generates stable gait patterns by ensuring the total moment of forces equals zero at a point on the ground.
Dynamic Model-Based
Uses mathematical models of robot dynamics to compute control commands for stable movement.
Optimization-Based Methods
Model Predictive Control (MPC)
Frames control as real-time optimization, predicting future states to determine optimal actions.
Whole-Body Control (WBC)
Blends upper and lower body tasks, prioritizing execution while respecting physical constraints.
Trajectory Optimization (TO)
Optimizes entire motion trajectories for efficiency and stability.
Bionic Methods
Central Pattern Generator (CPG)
Inspired by biological neural circuits that generate rhythmic movements without sensory feedback.
CMAC
Cerebellum-Model-Articulation Controller mimics cerebellar function for motor learning.
Learning-Based Methods
Reinforcement Learning (RL)
Robots improve through trial and error, learning optimal policies from reward feedback.
Imitation Learning
Robots learn by observing and mimicking expert demonstrations.
Behavior Cloning (BC)
Directly learns policies from demonstration datasets using supervised learning.
Inverse Reinforcement Learning (IRL)
Learns reward functions from expert demonstrations, then optimizes policies.
Notable Humanoid Robot Specifications
Technical specifications of representative humanoid robots from research institutions and industry worldwide.
Research Platforms
| Robot | Year | Institute | DOF | Mass (kg) | Height (m) | Mobility |
|---|---|---|---|---|---|---|
| WABIAN-2 | 2005 | Waseda University, Japan | 41 | 64.5 | 1.53 | Biped |
| HRP-2 | 2004 | AIST, Japan | 30 | 58 | 1.54 | Biped |
| HRP-4C | 2009 | AIST, Japan | 42 | 43 | 1.58 | Biped |
| iCub | 2004 | IIT, Italy | 53 | 25 | 1.04 | Biped |
| COMAN | 2015 | IIT, Italy | 25 | 31 | 0.95 | Biped |
| Kengoro | 2016 | JSK, Japan | 68 | 56 | 1.7 | Biped |
Commercial/Industrial
| Robot | Year | Institute | DOF | Mass (kg) | Height (m) | Mobility |
|---|---|---|---|---|---|---|
| ASIMO 2011 | 2011 | Honda, Japan | 57 | 48 | 1.3 | Biped |
| ATLAS | 2016 | Boston Dynamics, USA | 28 | 82 | 1.65 | Biped |
| Optimus Prime | 2023 | Tesla, USA | 40 | 56 | 1.73 | Biped |
| Valkyrie | 2013 | NASA, USA | 44 | 44 | 1.9 | Biped |
| REEM-C | 2014 | PAL Robotics, Spain | 68 | 70 | 1.6 | Biped |
| DRC-HUBO | 2015 | KAIST, South Korea | 33 | 80 | 1.75 | Biped/Wheeled |
Child-Sized/Education
| Robot | Year | Institute | DOF | Mass (kg) | Height (m) | Mobility |
|---|---|---|---|---|---|---|
| NAO | 2009 | Aldebaran, France | 25 | 4.5 | 0.57 | Biped |
| DARwIn-OP | 2011 | Robotis/UPenn | 20 | 2.8 | 0.45 | Biped |
| Poppy | 2014 | INRIA, France | 25 | 3.5 | 0.83 | Biped |
| NimbRo-OP2 | 2018 | AIS, Germany | 20 | 19 | 1.35 | Biped |
| Surena-Mini | 2017 | CAST, Iran | 23 | 3.3 | 0.53 | Biped |
Chinese Robots
| Robot | Year | Institute | DOF | Mass (kg) | Height (m) | Mobility |
|---|---|---|---|---|---|---|
| BHR-5 | 2012 | BIT, China | 30 | 63 | 1.62 | Biped |
| Wu Kong | 2011 | Zhejiang Univ., China | 30 | 55 | 1.62 | Biped |
| GOROBOT-III | 2007 | HIT, China | 70 | 90 | 1.58 | Biped |
| THBIP-I | 2002 | Tsinghua, China | 32 | 130 | 1.8 | Biped |
| Blackman | 2005 | NUDT, China | 36 | 63.5 | 1.54 | Biped |
Actuator Technologies
Different actuator types enable humanoid robots to move with varying degrees of force, precision, and compliance.
Electric (DC Motors)
Brushed and brushless DC motors providing precise control and robust torque
Hydraulic
Fluid-powered actuators excelling in demanding tasks and dynamic maneuvers
Pneumatic (Artificial Muscles)
Muscular mechanisms providing lightweight and adaptable motion
Series Elastic Actuators (SEAs)
Compliant actuators with elastic elements for safety and energy storage
Shape Memory Alloys (SMAs)
Materials that contract when heated, providing noiseless actuation
Sensor Systems & Perception
Multi-sensor fusion enables humanoid robots to perceive and interact with their environment effectively.
Vision Systems
Monocular and stereo cameras for environmental perception
- Object detection
- Obstacle avoidance
- SLAM
- Human recognition
Inertial Measurement Units (IMUs)
Accelerometers, gyroscopes, and magnetometers for body orientation
- Attitude measurement
- Angular velocity sensing
- Balance control input
Force/Torque Sensors
Sensors measuring contact forces and joint torques
- Contact detection
- Manipulation force control
- Ground reaction forces
Tactile Sensors
Touch-sensitive arrays for manipulation and interaction
- Object grasping
- Texture recognition
- Slip detection
Range Sensors
LiDAR, ultrasound, and infrared for distance measurement
- 3D mapping
- Obstacle detection
- Navigation
Future Research Directions
Biological Mechanisms
Deeper understanding of human anatomy, physiology, and neural control mechanisms to inform robot design.
Biological Sensing
Highly accurate, affordable sensors for advanced perception combining vision, touch, and proprioception.
Biological Structure
Evolution from rigid to integrated rigid-flexible designs for more natural, adaptable movements.
Biological Materials
Biomimetic materials with self-adaptive and self-healing properties for enhanced durability.
Biological Control
Brain-inspired control combining optimization, continuous learning, and autonomous decision-making.
Biological Energy
Efficient conversion and utilization of energy sources for extended operational autonomy.
Key Technological Challenges
Based on IEEE/CAA research surveys, these represent the six major technological areas where humanoid robots face significant challenges requiring breakthrough innovations.
Biological Mechanisms
Robots can only partially replicate human musculoskeletal coordination
- •Human body complexity: skeletal, muscular, and nervous systems work in intricate coordination
- •Replicating natural gait patterns with 200+ bones and 600+ muscles
- •Understanding proprioception and vestibular feedback loops
Bio-inspired designs, tendon-driven systems, artificial muscles
Sensing & Perception
Multi-modal sensor fusion improving but not human-level
- •Integrating visual, tactile, auditory, and proprioceptive data in real-time
- •Environmental uncertainty and sensor noise handling
- •Achieving human-like situational awareness and reaction speed
Event cameras, distributed tactile sensors, neural network fusion
Structural Design
Trade-offs between weight, strength, and DOF count
- •Achieving human-like range of motion with 30-50+ DOF
- •Balancing structural rigidity with flexibility for safe interaction
- •Miniaturizing components while maintaining performance
3D printing, topology optimization, modular architectures
Materials
Carbon fiber and aluminum alloys dominant; soft materials emerging
- •Finding materials that are lightweight, strong, and impact-resistant
- •Developing artificial skin with touch sensitivity and durability
- •Heat dissipation in actuator-dense limbs
Carbon nanotubes, smart materials, self-healing polymers
Control Systems
Hybrid classical + learning-based approaches showing promise
- •Real-time computation for high-DOF whole-body control
- •Sim-to-real transfer gap for learned policies
- •Robust recovery from unexpected disturbances
Foundation models, end-to-end learning, MPC + RL hybrid
Energy & Power
Battery life typically 1-4 hours under active use
- •High energy consumption from multiple actuators (50-150W per joint)
- •Battery weight vs. capacity trade-off affecting mobility
- •Thermal management in enclosed humanoid bodies
Solid-state batteries, energy harvesting, efficient actuators
Academic Source
Content on this page is informed by: Y. Tong, H. Liu, and Z. Zhang, "Advancements in humanoid robots: A comprehensive review and future prospects," IEEE/CAA J. Autom. Sinica, vol. 11, no. 2, pp. 301-328, Feb. 2024.
Explore Further
See these concepts applied to real robots in our database