Lightbot 0

About

About Lightbot 0

Lightbot 0 is Light Origins’ compact humanoid research platform for learning movement that uses the whole body and contact with the surrounding environment. Its LightParkour system combines walking with climbing and vaulting skills in one policy. A chest-mounted depth camera and the robot’s own motion feedback guide the controller, which runs onboard at 50 Hz and selects when to switch between walking and more complex movements.

The research transfers skills developed in simulation onto the physical robot, with demonstrations on stairs, narrow supports and obstacle courses. Training begins with short human-motion examples, adjusts them to physical constraints and expands them across different terrain. Light Origins also shows an early, separately trained policy carrying objects while walking. Lightbot 0 remains a research platform. Numerical obstacle-success rates in the report come from simulation.

Capabilities

Source-backed capabilities

  • Operating environments: Research obstacle course with narrow wooden supports and stepping stones, Physical research staircase
  • Control modes: Autonomous
  • Multi-terrain locomotion

    Research demonstration - 2026-08-03 - Walk across narrow supports and stepping stones while selecting footholds from onboard depth observations. - environment: Research obstacle course with narrow wooden supports and stepping stones

    Source
  • Stair climbing

    Research demonstration - 2026-08-03 - Go up and down stairs using the same onboard policy as whole-body locomotion. - environment: Physical research staircase

    Source

Research

  • Light-Loco-Parkour: Versatile Perceptive Whole-Body Locomotion via Multi-Skill Distillation

    2026-08-01 - Research Publication - Light Origins’ first public research release demonstrates one onboard policy for locomotion and contact-rich parkour on the Lightbot 0 humanoid.

    Source

Resources

Timeline

  • Light Origins publishes Light-Loco-Parkour for whole-body humanoid locomotion

    2026-08-01 - Research Publication - The policy uses only onboard depth sensing and a velocity command to select walking, balancing, climbing, stepping down or vaulting without a runtime skill label, state machine or motion graph. A Real2Sim2Real pipeline expands sparse human-motion seeds across terrain variations, unifies multiple experts and distils privileged geometry into a recurrent depth policy. The same policy transferred without real-world fine-tuning to indoor and outdoor tests on Lightbot 0, including unseen obstacle variations.

    Source

Context

Comparable Systems

  • TORO

    DLR Institute of Robotics and Mechatronics - Anthropomorphic Biped

  • TALOS

    PAL Robotics - humanoid robotics

  • TLIBOT

    Sichuan Tianlian Robotics Co., Ltd. - Anthropomorphic Biped

  • 4NE1

    Neura Robotics - humanoid robotics

  • ALICE 4

    AeiROBOT - humanoid robotics

  • AGIBOT A2

    AgiBot - humanoid robotics

Related Systems

  • Atlas

    Boston Dynamics - humanoid robotics

  • Calvin-40

    Wandercraft - humanoid robotics

  • T800

    ENGINEAI - humanoid robotics

  • MenteeBot

    Mentee Robotics - humanoid robotics

  • NAVIAI i3

    Zhejiang Humanoid - humanoid robotics