THANU — The Humanoid Robot · Aananthal Technologies

What Is THANU

Not a toy.
Not a tool.
A collaborator.

THANU is Aananthal's flagship humanoid robot — designed from first principles to interact naturally with humans in real-world environments. Unlike industrial robots that repeat fixed motions, THANU perceives, reasons, adapts, and responds.

Built on a foundation of multi-sensor fusion, computer vision, and real-time AI inference, THANU processes its environment at every moment — adjusting its gait on uneven ground, understanding spoken language in context, and learning from each interaction to become more effective over time.

check_circle32 degrees of freedom — natural human-like movement
check_circleMulti-modal AI: vision, language, proprioception, touch
check_circleEdge AI inference on NVIDIA Jetson platform
check_circleContinuous learning — improves with every deployment

Humanoid robotics research — THANU is built from the ground up in Bangalore

Intelligence Stack

How THANU Thinks & Moves

THANU's intelligence pipeline runs across five layers — from raw sensor input to physical action — in under 200ms.

sensors

01

Perception Layer

RGB-D cameras, LiDAR, IMU, microphone arrays, and tactile sensors deliver a continuous multi-modal stream of environmental data.

chevron_right

memory

02

Sensor Fusion

All sensor streams are synchronised, calibrated, and fused into a unified world model — updated at 100Hz using an Extended Kalman Filter.

chevron_right

psychology

03

AI Reasoning

Scene understanding, natural language parsing, and intention recognition run simultaneously on the NVIDIA Jetson Orin platform.

chevron_right

directions_walk

04

Motion Planning

Model Predictive Control generates stable, adaptive trajectories in real time. Gait is adjusted dynamically for terrain, load, and speed.

chevron_right

smart_toy

05

Physical Action

Torque-controlled servo actuators execute the motion plan — with real-time force feedback closing the loop back to perception.

Capabilities

What THANU Can Do

smart_toy

Bipedal Locomotion

Dynamic walking on flat, uneven, and inclined surfaces. Adaptive gait generation using model predictive control and real-time stability estimation.

record_voice_over

Natural Language

Conversational AI for task comprehension and response. Understands context, follows multi-step instructions, and responds with synthesised speech.

visibility

Computer Vision

Real-time object detection, depth estimation, facial recognition, and scene understanding. Operates in variable lighting and complex environments.

pan_tool

Dexterous Manipulation

5-fingered hands with tactile feedback. Capable of grasping, assembling, and manipulating objects with human-like precision and gentleness.

self_improvement

Continuous Learning

Reinforcement learning from physical interaction. THANU improves its movement efficiency and task performance with every deployment cycle.

sensors

Multi-Sensor Fusion

RGB-D cameras, LiDAR, IMU, microphone arrays, and joint encoders fused at 100Hz into a unified world model for robust real-world navigation.

battery_charging_full

8h Battery Life

High-density lithium battery pack with intelligent power management. Hot-swappable design for continuous operation in commercial deployments.

shield

Safety Systems

Multi-layer collision avoidance, force limiting on all joints, emergency stop, and human proximity detection for safe human-robot collaboration.

Hardware

Sensor Suite

camera

RGB-D Cameras

2× Intel RealSense D435 · 30fps depth

Object detection, depth mapping, face recognition

radar

LiDAR

16-channel · 360° · 100m range

Spatial mapping, obstacle avoidance, navigation

rotate_90_degrees_ccw

IMU Array

6-axis · 1000Hz · redundant pair

Balance, posture estimation, fall detection

mic

Microphone Array

8-channel beamforming · 360° pickup

Speech recognition, sound localisation

touch_app

Tactile Sensors

192 pressure points per hand

Grip force control, texture sensing, safety

settings

Joint Encoders

32× absolute encoders · 18-bit resolution

Proprioception, torque feedback, gait control

R&D Progress

Gait Research Milestones

Bipedal locomotion is the hardest problem in humanoid robotics. We started from first principles — modelling the physics before writing a line of control code. Here's where we are.

Static Balance100%

Controlled Walking (flat)78%

Incline Navigation45%

Uneven Terrain30%

Stair Climbing15%

Gait research in progress at Aananthal Labs, Bangalore — 2026

Specifications

Technical Specifications

Height

180 cm (target)

Weight

65 kg (target)

Degrees of Freedom

32 (full body)

Compute Platform

NVIDIA Jetson Orin NX

Operating System

ROS 2 Humble on Ubuntu

Battery Life

8 hours continuous

Response Latency

< 200ms perception-to-action

Walking Speed

1.2 m/s (target)

Payload Capacity

5 kg (arm)

IP Rating

IP44 (target)

AI Stack

AI Modules Inside THANU

01

Visual Perception Model

YOLOv8-based real-time object detection + depth estimation. Identifies people, objects, obstacles, and semantic spaces at 30fps.

PyTorch · YOLOv8 · CUDA

02

Natural Language Understanding

Fine-tuned language model for task comprehension, question answering, and context-aware conversation. Runs entirely on-device.

Transformer · On-device LLM · Speech-to-Text

03

Gait Control Policy

Reinforcement learning policy trained in simulation (Isaac Gym) and transferred to hardware. Handles dynamic perturbations and terrain changes.

PPO · Isaac Gym · Sim-to-Real

04

Manipulation Planner

Model-based trajectory planning for arm and hand motion. Grasp synthesis using point cloud inputs. Force-compliant execution.

MoveIt2 · IKFast · Force Control

05

Continuous Learning Engine

Online adaptation layer that fine-tunes models from deployment feedback. Improves task success rates without full retraining.

Online RL · Federated Learning · Experience Replay

Lab Progress

From Whiteboard to Metal

Prototype Gen 1.1

Full body structure assembled · Q1 2026

Arm Assembly

5-DOF arm with tactile hand

Control System

Jetson Orin + servo drivers

Applications

Where THANU Will Work

home

Home Assistance

Daily task assistance for elderly and differently-abled individuals — fetching, navigating, and communicating with natural ease.

Elderly care assistance
Mobility support
Home management

factory

Industrial Operations

Safe collaboration with human workers in manufacturing — handling tasks too repetitive, hazardous, or precise for human operators.

Assembly line support
Quality inspection
Hazardous operations

school

Research & Education

Open platform for AI and robotics research. Universities and labs can build on THANU's architecture to advance humanoid science.

University research
AI curriculum labs
Open platform development

storefront

Retail & Hospitality

Customer-facing service robot — guiding, answering questions, and handling routine interactions in commercial spaces.

Information kiosk
Hotel concierge
Retail navigation

From the Lab

The Team on THANU

"The moment THANU first maintained balance through a lateral push — that was the moment I knew we were on the right path. Bipedal balance isn't one problem. It's a hundred nested problems, each solved in milliseconds. We're solving them one by one."

"Running inference on embedded hardware for real-time gait control with no cloud dependency — that was the design constraint that forced us to be better engineers. Every millisecond of latency we cut makes THANU more alive."

"India building a humanoid robot from scratch isn't a crazy idea. It's an inevitable one. The question was never if — only when, and who would do it first. We chose to be that team."

History

Research Timeline

Q3 2025

Project Inception

Team formation, first principles research, biomechanics literature review, simulation environment setup

Complete

Q4 2025

Simulation & Design

Full body CAD, simulation in Isaac Gym, initial gait policy training, sensor integration design

Complete

Q1 2026

Hardware Prototype

Gen 1.0 body structure assembled, actuator calibration, static balance demonstrated in lab conditions

Complete

Q2 2026

Gait R&D

Active bipedal locomotion research, sim-to-real transfer, early walking trials on flat terrain

In Progress

Q4 2026

Multi-terrain Testing

Stairs, inclines, uneven surfaces, human interaction safety protocols

Upcoming

2027

Commercial Prototype

Gen 2.0 refined body, full sensor suite, production-ready AI stack, partner deployment programme

Planned

Questions

Frequently Asked

Talk to us directly.

Not yet. THANU is in active R&D. We're targeting a partner deployment programme in 2027. Research institutions and strategic partners can express interest now through our contact page.

That's a future goal. We're currently focused on reliable bipedal locomotion and natural interaction in controlled environments. Full end-to-end autonomy in unstructured environments remains an active research area that will be tackled progressively.

THANU is designed from first principles by an Indian team, built for environments and use cases specific to India and South Asia. We're not adapting a foreign platform — we're building natively for our context, at a fraction of the cost of Western competitors.

We're open to university partnerships, joint research programmes, and industry collaborations. Reach out through our contact page with your institution and area of interest.

Positioning

THANU vs The Field

Dimension

THANU

Global Competitors

Research Robots

Origin & Context

India-first design

Western contexts

Lab-only

Cost Target

Accessible price point

$100,000+

N/A

Edge AI

Fully on-device

Mixed

Cloud-dependent

Open for Research

Yes (planned)

Closed

Limited

The Humanoid
Robot.

Not a toy.
Not a tool.
A collaborator.

How THANU Thinks & Moves