Indian Space Research Organisation(ISRO)
Browse and strategize across all 11 official problem statements submitted by Indian Space Research Organisation(ISRO) for Smart India Hackathon 2026.
Problem Statements (11)
Match My TeamMulti-modal, Sun angle and scale invariant image correspondence using Chandrayaan-2 optical images (OHRC, TMC and IIRS)
Automated co-registration of satellite images captured at different seasons, sun illumination angles, sensor modalities (Optical vs SAR Radar), and spatial resolutions (Cartosat vs Sentinel) suffers from high geometric distortion, parallax errors, and feature matching failures over rugged Indian terrain. Build an AI-Driven Multi-Modal, Sun-Angle, and Scale-Invariant Image Correspondence and Geo-Registration Engine for ISRO that uses deep local invariant feature descriptors (SuperPoint / LoFTR) to achieve sub-pixel spatial alignment across heterogeneous satellite acquisitions.
SatQuery AI - An Interactive Vision-Language Assistant for Multimodal Remote Sensing Image Analysis through Text Queries
Scientists, urban planners, and disaster responders struggle to query massive archives of Indian satellite imagery on ISRO's Bhuvan and Bhoonidhi portals because finding specific geospatial insights (e.g. 'Show flood inundation in Kaziranga between June and August 2024' or 'Count solar farms in Rajasthan') requires specialized GIS software and manual band-math processing. Build 'SatQuery AI'—an Interactive Vision-Language Assistant and Spatial Grounding Engine for Multimodal Earth Observation for ISRO that understands natural language queries, performs zero-shot satellite object detection, and generates analytical GIS summary reports.
AI-ML based Intelligent Dead Reckoning system for seamless navigation
Planetary exploration rovers (like Chandrayaan lunar rovers) and autonomous vehicles navigating deep underground tunnels or dense urban canyons face prolonged GPS/NavIC satellite signal blackouts where classical Inertial Dead Reckoning accumulates rapid integration drift and position errors. Build an AI-ML Based Intelligent Dead Reckoning and Sensor Fusion Navigation System for ISRO that combines IMU accelerometer/gyroscope readings, wheel-odometry slip correction, and Visual Odometry using deep learned kinematic error models to maintain sub-meter positioning accuracy during GNSS denial.
Development of an AI-Based Virtual Camera Tracking System for Coarse Alignment of Mobile Free Space Optical Communication (FSOC) Terminals
Launch vehicle rocket tracking during satellite lift-off at ISRO (Satish Dhawan Space Centre SHAR) relies on ground optical tracking telescopes, where high-speed rocket acceleration, cloud occlusions, rocket exhaust plumes, and camera vibrations cause traditional image processing to lose optical track. Build an AI-Based Virtual Camera Tracking and Coarse-to-Fine Trajectory Prediction System for ISRO that combines deep object detection, multi-scale Kalman filters, and physics-based rocket launch kinematics to maintain continuous lock-on tracking through cloud occlusions.
AI-Driven Anomaly Detection in Component Burn-In & Screening
Space-grade electronic components (FPGA, microprocessors, power MOSFETs, RF MMICs) used in ISRO satellite payloads undergo mandatory high-temperature accelerated electrical burn-in and screening (125°C for 168+ hours), where subtle early-life component degradation (gate oxide leakage, electromigration, thermal runaway) goes unnoticed in massive multi-channel voltage/current logs. Build an AI-Driven Anomaly Detection and Reliability Screening Platform for ISRO that analyzes high-frequency electrical telemetry during burn-in testing to flag micro-anomalies and predict space mission component infant mortality.
On-device Visual Perception for Light-weight Browser Agents
Autonomous AI browser agents executing automated web tasks (booking satellite launch permits, scraping public Earth observation catalogs, automating web portals) rely on heavy server-side Vision-Language Models (like GPT-4V) that take 5-10 seconds per webpage action, consume high cloud API costs, and leak user credentials to external servers. Build a Lightweight On-Device Visual Perception and Web Grounding Engine for Browser Agents for ISRO that runs quantized small multimodal models directly in the web browser/edge device to locate clickable UI buttons, parse complex forms, and execute multi-step web workflows in sub-500 milliseconds.
Low Latency and Efficient Voice Activator for Edge Devices
Spacecraft cockpits, defense helmets, and battery-powered satellite edge devices require hands-free voice command activation, but existing wake-word models consume high memory, have noticeable latency (>500ms), and trigger frequent false activations on background engine noise and cabin chatter. Build a Low-Latency, Ultra-Low Power Voice Activator and Wake-Word Engine for Edge Devices for ISRO in C++/Rust that operates on microcontrollers (<256 KB RAM) with sub-100 millisecond activation latency, streaming acoustic keyword spotting, and zero-shot custom wake-word enrollment.
iTantra -Indian Multilingual TTS & STT Aided Neural Transceiver Radio Access for low bitrate links
Disaster management responders, remote maritime fishermen, and defense field units across India communicate over low-bandwidth tactical VHF/HF radios and satellite transponders where voice communication is distorted by static noise and limited by language barriers across multi-state teams. Build 'iTantra'—an Indian Multilingual TTS & STT-Aided Neural Transceiver Radio Platform for ISRO that compresses voice into ultra-compact semantic text tokens (sub-100 bps), translates speech in real time across 12+ Indian languages, and resynthesizes voice in authentic speaker timbre.
AI Human Activity Recognition for On-board BAS Experiments
Astronauts aboard the Indian Space Station (BAS - Bharatiya Antariksh Station) and Gaganyaan crew module conduct complex scientific experiments under microgravity where floating postures, zero-g kinematic dynamics, and tight spacecraft camera angles make standard Earth-based action recognition models fail. Build an AI-Powered Human Activity Recognition (HAR) and Protocol Compliance Tracking Platform for ISRO that uses 3D skeleton pose estimation, multimodal IMU wearable tracking, and spatial action graphs to monitor astronaut scientific experiments in microgravity.
DepthWizard - Single-View Height Estimation and 3D Flythrough
Disaster responders, defense intelligence officers, and urban planners working with ISRO satellite data frequently have access to only single optical monocular satellite images (with no stereo pairs or LiDAR elevation data available), making 3D elevation analysis and building height measurements impossible. Build 'DepthWizard'—a Single-View Height Estimation and 3D Photorealistic Flythrough Generation Platform for ISRO that uses deep monocular depth foundation models, solar shadow geometry trigonometry, and 3D neural rendering to reconstruct 3D cityscapes from a single 2D satellite image.
ORCA Marine EcOsystem Reasoning with Collaborative Agents
Marine ecosystem conservation and fisheries management across the Indian Ocean require complex multi-disciplinary reasoning across sea surface temperatures, chlorophyll ocean blooms, coral bleaching, illegal fishing vessel tracks, and climatic El Niño oscillations that no single AI model can analyze holistically. Build 'ORCA' (Marine Ecosystem Reasoning with Collaborative Agents) for ISRO—a multi-agent vision-language reasoning platform that coordinates specialized AI agents (Oceanographer Agent, Fisheries Agent, Coral Health Agent, Vessel Surveillance Agent) to solve complex marine conservation inquiries.