A signal in the flood.A lifeline in motion.
- Engine
- T+000 · —
- Open SOS
- —
- SAR data age
- awaiting pass
- D1 barrage (1958)
- —
- Rescued
- —
- Invariants
- —
Flood plans go stale in minutes.
Roads close, a bridge gives way, a shelter fills, and two people report the same bridge differently. FloodSetu treats every boat, ground team, shelter, bridge and SOS cluster as an autonomous agent with only partial information — and keeps re-planning, with a reason attached to every decision.
- Time to service01
- Unmet requests02
- Resource idle time03
- Route violations04
- Fairness across neighbourhoods05
- Recovery after contradictory updates06
Incidents announce. Boats bid. The best offer wins — in the open.
A Contract-Net auction with leveled commitment: each agent bids its time-to-service from its own map; awards weigh urgency, vulnerability, waiting time, fairness — and dam risk. Every award has a bid table and a “Why?”. Money has no input. That is tested, not promised.
BRIDGE · BOAT · SHELTER — INJECTING…
Satellites see what the network can’t — but never in real time.
Monsoon cloud blinds optical cameras, so FloodSetu leans on ISRO radar. Every product carries its acquisition time and age. Fusion is last-writer-wins by observation time: a two-hour-old image never overrides a fresher field reading, and satellite downlink still works when local towers are dark.
| Satellite | Operator | Sensor | Sees | Revisit | Latency | Data age · live |
|---|---|---|---|---|---|---|
| EOS-04 (RISAT-1A) | ISRO | C-band SAR | Flood extent through cloud | 17-day repeat · tasked | ~1h 50m* | — |
| NISAR S-band | ISRO · NASA | S-band SAR | Flood extent, flooded vegetation | 12-day (~6 d avg) | ~1h 35m* | — |
| INSAT-3DS + 3DR | ISRO · IMD | Geostationary imager | Rain cells (IMSRA), nowcast | 15 min (staggered) | ~30m* | — |
| Resourcesat-2A | ISRO | LISS-IV optical | Detail — when no cloud | 5-day | ~1h 00m* | — |
* Modelled from public sources (NRSC flood bulletins, NISAR disaster-product target <5 h, MOSDAC half-hourly IMSRA). Pass times are compressed into the scenario; latencies are the realistic part.
Over 1,280 large dams in India are more than 50 years old.
The National Register of Large Dams (CWC · NDSA, 2023) counts 224 dams older than a century and 1,065 aged 50–100. FloodSetu reads a register like it — year of completion, type, last inspection, hazard class — scores each dam, and marks its downstream area a high-risk zone. Requests inside get a documented priority bump; a release turns the zone very high.
| Dam (fictional) | Built | Age | Inspection | Risk |
|---|---|---|---|---|
| Jalnagar Barrage zone GHT, RLC · | 1958 | 68 yrs | needs attention | very high |
| Pahadi Nala Dam zone MCH · | 1994 | 32 yrs | satisfactory | moderate |
| Shanti Check Dam zone SHN · | 2011 | 15 yrs | satisfactory | low |
—
What if the phone drowns?
“I’m in the water” sends a CRITICAL SOS with GPS in under 200 bytes — no typing with wet hands. It queues offline and retries itself.
The open app pings while help is on the way. 90 s of silence on an active request means the phone is presumed submerged or dead.
The incident keeps its last-known location and building cluster. It is upgraded to critical and escalated in the same tick (safety invariant 10).
The assigned team is told to search the last-known position; the alternate number is called and an IVR/SMS callback goes out (mocked).
At ≤10% battery the last location is saved and the control room is warned before the phone dies.
SMS “SOS 4 C near ghat 2” works from a feature phone, and 112 stays one tap away.
No neighbourhood is left waiting. No rupee buys priority.
urgency × vulnerability × people × aging × verification × escalation × fairness (+ dam risk)
sees verified need — can never feed back. Enforced by an architecture test and a lint rule.
Every rupee, every kit — on a ledger anyone can re-verify.
simulated money and in-kind, by verified campaign
per verified area, k-anonymous
timestamped, attributed, SHA-256 hash-chained
In your pocket.
A native iPhone app (built, in the repo) and an installable web app: one-tap SOS with an offline queue, live flood alerts and river levels, live GPS to the team, and a rescuer mode with turn-by-turn to the victim.
Stored only as a salted hash. Mocked — nothing is sent.