MANUSH Research
MANUSH Labs — advanced research for embodied intelligence.
Research that turns intelligence into machines that perceive, reason, move, learn and safely act in the physical world.
Research Areas
Ten research areas
- Embodied AI AI + robotics integration: intelligence that perceives, reasons, moves, learns and safely acts in the real world.
- Robot Learning Learning from demonstration, experience, simulation and fleet learning.
- Dexterous Manipulation Hands, arms and tools: two-handed, precision and force-controlled manipulation.
- Computer Vision Perception and sensor fusion for machines operating among people.
- Spatial Intelligence 3D perception, SLAM and world modeling.
- Locomotion Walking, running, balance and whole-body control.
- Multimodal Models Vision, audio, language, touch, depth, force, motion and spatial data in one model.
- Robot Simulation Digital twins, synthetic data and reinforcement learning in MANUSH Sim.
- Safety Robot safety and governance — MANUSH SafeCore.
- Human-Robot Interaction Natural multilingual interaction, intention estimation and human handover.
Technology
Research technology
- Spatial Intelligence
World Models
MANUSH World — a physical-world model of objects, people, rooms, factories, roads, terrain, physics, motion, cause and effect, spatial relationships, tools and…
- Multimodal Models
Multimodal Foundation Models — MANUSH Mind
The central multimodal intelligence model: Vision, Audio, Language, Touch, Depth, Force, Motion and Spatial Data → Reason → Plan → Act.
- Dexterous Manipulation
Dexterous Manipulation
Two-handed manipulation, precision manipulation, force feedback, tactile sensing, tool use, object transfer and human handover.
- Locomotion
Legged Locomotion
Bipedal walking, stair climbing, uneven terrain, running, crouching, kneeling and getting up after falling — plus quadruped locomotion for extreme terrain.
- Robot Learning
Learning from Demonstration
Show it once → teach the robot the task. Converting human demonstrations into reusable robot skills.
- Safety
SafeCore Safety Kernel
Hard safety boundaries and runtime monitoring — high-level AI should never be allowed to bypass the low-level safety system.
Publications
Papers & technical reports
Research Divisions
MANUSH Labs — ten divisions
- 01MANUSH AI — Foundation models.
- 02MANUSH Robotics — Mechanical and electrical engineering.
- 03MANUSH Embodied Intelligence — AI + robotics integration.
- 04MANUSH Dexterity — Hands and manipulation.
- 05MANUSH Mobility — Walking, running, balance and locomotion.
- 06MANUSH Spatial — 3D perception, SLAM and world modeling.
- 07MANUSH Autonomy — Planning and decision systems.
- 08MANUSH Safety — Robot safety and governance.
- 09MANUSH Energy — Battery, motors and power electronics.
- 10MANUSH Materials — Actuators, structural materials and future robot bodies.
Research collaboration
Research with MANUSH.
MANUSH H1 is built for university/research partnerships — locomotion, manipulation, vision, voice, teleoperation, reinforcement learning, imitation learning and human demonstrations.