NoteTube

Problem Statement Session 2 | Bharatiya Antariksh Hackathon 2026
1:18:41

Problem Statement Session 2 | Bharatiya Antariksh Hackathon 2026

Hack2skill

8 chapters8 takeaways27 key terms7 questions

Overview

This video is the second problem statement explanation session for the Bharatiya Antariksh Yatra 2026 hackathon, organized by ISRO. It covers problem statements 9 through 15, with mentors explaining the technical challenges and expected outcomes. The hackathon is open to students, offering opportunities for learning, mentorship, potential internships at ISRO, and national recognition. Key dates and milestones, including registration deadlines and the grand finale, are also outlined. The session details specific problems in areas like adaptive optics, infrared image enhancement, satellite image retrieval, temporal resolution enhancement, network security, energetic particle forecasting, and solar flare prediction.

How was this?

Save this permanently with flashcards, quizzes, and AI chat

Chapters

  • Welcome to the third edition of the Bharatiya Antariksh Yatra 2026, an ISRO-powered national innovation initiative.
  • This session explains problem statements 9-15, following up on yesterday's session for statements 1-8.
  • The hackathon aims to bring student innovators together to build solutions for space technology.
  • Eligibility is for undergraduate, postgraduate, and PhD students; teams must have 3-4 members.
  • Key benefits include expert mentorship, potential ISRO internships, networking, and national recognition.
Understanding the hackathon's structure, eligibility, and benefits is crucial for potential participants to effectively engage and maximize their learning and opportunities.
Registration is done via the website, where participants form teams and choose one of the 15 listed challenges.
  • Develop and optimize algorithms for wavefront reconstruction and turbulence characterization using Shack-Hartmann sensor time-series data.
  • The problem is relevant to adaptive optics in ground-based astronomy, optical communication, and biomedicine.
  • A Shack-Hartmann sensor uses an array of lenslets to sample a wavefront and detect distortions.
  • Participants will receive time-series data of sensor frames to identify spot deviations and reconstruct the wavefront.
  • The reconstructed wavefront will be used to derive turbulence characteristics and an actuator map for a deformable mirror.
This problem addresses the challenge of atmospheric distortion in astronomical observations, aiming to improve image clarity through advanced optical techniques.
A plane wavefront produces a regular spot field on a detector, while a distorted wavefront causes displaced spots, which are used to calculate local slopes and reconstruct the wavefront.
  • Develop an end-to-end framework to super-resolve and colorize low-resolution thermal infrared (TIR) satellite images.
  • The goal is to improve object interpretation by predicting realistic RGB colors for TIR images.
  • Current low-resolution TIR images from satellites are difficult for humans and AI to interpret, hindering earth observation analysis.
  • The framework will involve a super-resolution model followed by a colorization model.
  • The dataset will be from Landsat 8 or 9, using co-registered TIR and RGB images.
Enhancing and colorizing infrared satellite imagery makes crucial earth observation data more interpretable, aiding in better analysis of urban areas, infrastructure, and environmental changes.
Input is a single-channel TIR image at 200m resolution; the output is a high-resolution (100m) TIR image and a corresponding colorized RGB image at 100m resolution.
  • Address the temporal resolution gap in satellite imagery by generating synthetic frames using AI/ML-based optical flow.
  • Current geostationary satellites provide data at intervals (e.g., 10-30 minutes), missing rapid events like cyclones, wildfires, and thunderstorms.
  • The approach involves using optical flow for motion vector estimation to interpolate intermediate frames between existing ones.
  • Participants should use thermal infrared bands (e.g., GOES-19 ABI Channel 13) for interpolation.
  • The final model should be applied to INSAT 3DR data to generate 15-minute interval frames, with deliverables including a trained model, interpolated frames, and a web dashboard.
Improving the temporal resolution of satellite data allows for near real-time monitoring of rapidly evolving natural phenomena, enhancing disaster response and prediction capabilities.
Given frames at 12:00 and 12:30 hours, the model should interpolate a synthetic frame for 12:15 hours.
  • Develop an autonomous, fully offline LLM-based co-pilot for predictive fault analysis in MPLS networks.
  • The model should forecast network failures proactively, explain issues in plain English, and not depend on internet connectivity or external APIs.
  • Participants need to simulate a multi-site MPLS network topology and perform fault analysis based on parameters like route flapping, queuing, and tunnel status.
  • An open-weight LLM model should be used with local playbooks and runbooks for RAG (Retrieval-Augmented Generation).
  • The system should correlate events, identify potential failure reasons, and provide explanations.
This problem aims to enhance network security and reliability by creating an offline predictive system that can anticipate and explain potential failures before they impact operations.
The co-pilot should be able to query information like 'What is this IP address?' and, by consulting local configuration files, identify the corresponding router and location.
  • Develop an AI/ML-based forecast model for energetic particle radiation, specifically 'killer electrons' (MeV energies), for ISRO's geostationary satellites.
  • The model should provide short-term (nowcasting) and long-term forecasts of harsh radiation experienced in geostationary orbit.
  • Key factors influencing the radiation environment include solar wind speed, interplanetary magnetic field (IMF), and density.
  • Data sources include GOES satellite data (electron fluxes > 2 MeV), solar wind data (e.g., from ACE/Wind), and GRACE satellite data for validation over Indian longitudes.
  • The expected outcome is a real-time GUI forecasting electron fluxes for ISRO's geostationary satellites.
Accurate forecasting of energetic particle radiation is critical for protecting sensitive satellite electronics from damage and ensuring the reliable operation of space missions.
Using historical solar wind parameters and past electron flux data, predict the electron flux for the next 30-45 minutes.
  • Develop models for real-time tracking (nowcasting) and advanced warning (forecasting) of solar flares.
  • The project utilizes combined soft and hard X-ray data from ISRO's Aditya-L1 satellite (SoLEXS and HELIOS instruments).
  • Solar flares are energetic explosions releasing charged particles that can cause geomagnetic storms, disrupting power grids, GPS, and communications.
  • Objectives include building detection algorithms for SoLEXS and HELIOS independently, creating a unified flare catalog, and developing a forecasting system with quantifiable lead time.
  • Evaluation metrics focus on flare detection accuracy, true positive rate, low false positive rates, and a lead time of 15-30 minutes for forecasts.
Predicting solar flares enables proactive measures to mitigate their disruptive effects on Earth's infrastructure and space-based assets, ensuring operational continuity.
Analyzing the impulsive flux enhancements in hard X-ray data (HELIOS) and gradual enhancements in soft X-ray data (SoLEXS) to detect and predict flare events.
  • Develop a unified model for cross-modal satellite image retrieval, enabling semantic similarity searches across optical, SAR, and multispectral images.
  • The objective is to retrieve similar regions across different sensor modalities, overcoming limitations of metadata-based search.
  • A deep learning model will learn a common embedding space for various modalities.
  • The model should accept a query image from one modality and retrieve top-K similar images from another modality (e.g., optical to SAR).
  • Evaluation metrics include F1 score and accuracy for same-model and cross-model retrieval, as well as average retrieval time per query.
Enabling cross-modal retrieval allows users to find relevant information across diverse satellite datasets, enhancing the utility and accessibility of remote sensing data for various applications.
Given a query image of an airport from an optical satellite, the model should retrieve the top five most similar airport images from a gallery of SAR satellite images.

Key takeaways

  1. 1The Bharatiya Antariksh Yatra 2026 hackathon provides a platform for students to innovate in space technology, offering significant learning and career development opportunities.
  2. 2Problem statements cover diverse and critical areas of space research, from improving astronomical observations to forecasting space weather events.
  3. 3AI and ML are central to solving many of these challenges, requiring participants to leverage techniques like deep learning, time-series analysis, and computer vision.
  4. 4Understanding and processing diverse satellite data formats (e.g., CDF, NetCDF) is a common requirement across multiple problem statements.
  5. 5The hackathon emphasizes both the technical development of solutions and the ability to clearly explain and present them.
  6. 6Successful participation involves not just technical skill but also understanding the 'why it matters' aspect of each problem statement.
  7. 7Participants are encouraged to explore existing literature and models but also to innovate and develop novel approaches.
  8. 8The ultimate goal for many problems is to create systems that can provide real-time insights or forecasts, aiding in critical decision-making.

Key terms

Bharatiya Antariksh Yatra 2026ISROHackathonShack-Hartmann Wavefront SensorAdaptive OpticsWavefront ReconstructionTurbulence CharacterizationThermal Infrared (TIR) ImageSuper-ResolutionImage ColorizationOptical FlowTemporal ResolutionAir GapMPLSLLM (Large Language Model)Predictive Fault AnalysisEnergetic ParticlesGeostationary OrbitSolar FlaresSpace WeatherAditya-L1SoLEXSHELIOSCross-Modal RetrievalSAR ImageMultispectral ImageEmbedding Space

Test your understanding

  1. 1How does a Shack-Hartmann wavefront sensor work, and what kind of data does it provide for turbulence characterization?
  2. 2What are the two main objectives of the infrared image colorization and enhancement problem statement, and why is this important for earth observation?
  3. 3Explain the concept of 'temporal resolution gap' in satellite imagery and how optical flow can be used to address it.
  4. 4What is the primary challenge in developing an 'air gap predictive co-pilot' for network security, and what role does an LLM play in this context?
  5. 5Why is forecasting energetic particle radiation important for geostationary satellites, and what types of data are typically used for such predictions?
  6. 6How do soft and hard X-ray data from instruments like SoLEXS and HELIOS differ, and how can this difference be leveraged for solar flare forecasting?
  7. 7What is cross-modal satellite image retrieval, and what are the key challenges in building a unified model for it?

Turn any lecture into study material

Paste a YouTube URL, PDF, or article. Get flashcards, quizzes, summaries, and AI chat — in seconds.

Summarize Your Own VideoSign up free

No credit card required

Problem Statement Session 2 | Bharatiya Antariksh Hackathon 2026 | NoteTube | NoteTube