In-orbit computing promises faster delivery of satellite data but is not a universal solution. Examine the conditions under which onboard processing is advantageous and where ground processing remains preferable.
New · District CollectorCan you hold the district for 72 hours?Floods, a heatwave, an outbreak, a grievance queue. Take the decisions, see the consequences.Play →In this answer
In-orbit computing means the satellite analyses its own data and sends down only the most relevant information. This cuts the delay caused by downloading data only during short ground-station pass windows [1]. Its value depends on the use: it helps most where speed matters more than having the complete data.
Where onboard processing is advantageous
- Time-critical disasters: flood maps of submerged roads and at-risk settlements help responders only if they arrive in time [1].
- Simple, well-defined tasks: in NASA's Dynamic Targeting test on CogniSAT-6, a satellite looked ahead along its path and skipped cloudy scenes within about 90 seconds, with no human input [2].
- Downlink bottlenecks: as data volumes rise, the gap between capturing data and using it widens. Screening on board saves scarce bandwidth, storage and power [1][2].
- Data-heavy radar: ISRO lists onboard SAR processing as a research area [5]. SAR can see through monsoon cloud, which makes it India's strongest use case.
Where ground processing remains preferable
- Irreversibility: a frame deleted in orbit is lost for good. Ground archives can be re-analysed later with better methods for climate records and recalibration.
- Changing analysis needs: a model trained before launch may misread a new kind of flood. Ground systems can be upgraded every day.
- Wrong filter, wrong moment: floods come with heavy cloud, so an optical cloud filter may throw away exactly the disaster frames.
- Complex tasks not yet proven: only cloud avoidance has been tested in flight. Detecting wildfires, storms and volcanoes is still only planned [2].
- Ground-side fixes exist: ISRO's AGEOS station in Antarctica gives polar-orbiting IRS satellites more frequent chances to send data down [4].
- Hardware limits: onboard power, heat, radiation tolerance and mass limit how much computing a satellite can carry.
Way forward
- First set delivery-time targets for each user group, as NASA's latency study did [3]. Then choose between onboard processing, polar stations and relay satellites by cost.
- Use a hybrid model: send urgent summaries at once and keep the raw frames on the onboard recorder for the next pass [4].
- IN-SPACe could require private operators to keep a log of discarded frames for audit.
In-orbit computing is therefore a tool for specific tasks, not a replacement for ground processing. A hybrid "screen in orbit, verify on ground" design, backed by clear delivery-time targets, can make Earth observation faster without losing data integrity. This would strengthen India's disaster resilience and its space economy.
Sources
- 1The Hindu, "When orbit processing makes sense — and when it doesn't" (explainer, 17 Nov 2026)pass-window delay; flood-mapping timeliness; sending only relevant data; rising data volumes
- 2NASA JPL, "How NASA Is Testing AI to Make Earth-Observing Satellites Smarter"Dynamic Targeting on CogniSAT-6; decision in under ~90 s; only cloud avoidance tested so far, event detection planned
- 3NASA GSFC Applied Sciences, *Study on Data Latency Needs and Requirements*delivery-time needs set for each user group
- 4ISRO, *Space India*, July–December 2016AGEOS Antarctic station; IRS onboard solid-state recorder and pass-based downlink
- 5ISRO, *Research Areas in Space 2023* (RESPOND)onboard SAR and signal processing listed as research areas
Practice
7 questions on this article
Check the answer for each question, or reveal all at once.