A solar sensor pod for river borders that cannot be fenced, and the tools to plan and run a fleet of them.
Floating or on stilts, each pod carries thermal, day camera, radar and acoustic sensors on 300 Wp of solar modules. Before one goes in the water, Prahari tells you whether it stays powered through the monsoon, how many you need for a stretch, and what they cost.
Live figures load from the latest export.
Much of the eastern border is water. 34.6 km of Assam's riverine border has no fence, 30.37 km of it in Dhubri (Assam Tribune, 6 Jul 2026). In the Sundarbans, BSF began a technical study in July 2026 to fence about 90 km, 71 km of it along the wildlife sanctuary (The Hindu, via SuperKalam).
Rivers do not hold still for a fence or a camera tower. Chars covered about a third of the Brahmaputra's channel area in Assam in 2017–18, and they form and vanish every few seasons (NASA and NESAC, 2018). Sensors here need their own power, their own link and a way to be moved.
The Home Ministry's Smart Border Project plans drones, radars and cameras along about 6,000 km of frontier (ETV Bharat, 22 May 2026), with pilots at 7 to 8 sites (Tribune, 6 Jun 2026). A riverine node is one of the pieces such a pilot needs.
- Reference design v0. Sensors, edge compute, radios, power, mast, mooring and enclosure, with the draw of every part in three operating profiles and a bill of materials in rupees: ₹18.0 lakh a floating pod, indicative.
- Power budget. Five years of daily sunlight at Gosaba, the Raimangal estuary, Hasnabad, Dhubri and Mankachar, run through a battery model one day at a time. It reports outage days, power-save days, the worst month and the smallest array that works.
- Coverage planner. Real river centrelines for the Ichamati, the Sundarbans and the Brahmaputra at Dhubri. Set the spacing and sensor ranges; it measures coverage by sensor, lists the gaps, and finds the fewest pods for full coverage and their cost per kilometre.
- Fleet console. 24 simulated pods with 14 days of hourly telemetry, 13 health rules and a maintenance queue that plans daylight boat trips, most urgent first.
- Listen and scan. The hydrophone and a duty-cycled radar run all the time at low power.
- Look on cue. Thermal and the zoom camera slew to what the cheap sensors found.
- Decide on board. An edge AI module detects and tracks. Tracks and events leave the pod; video stays on it.
- Report. VHF data or LoRa to a shore relay, then to Sangam for fusion and to the console for operators.
- Stay up. A supervisor drops to power save below 40 % charge. Health rules turn faults into boat runs.
Every number in the planner and the calculator comes from formulas on the methods page, implemented in the pipeline and in the browser, with a test that checks they agree.
- The monsoon sets the battery, not winter. Loading the five-year record…
| Source | What it gives | Licence | Status |
|---|---|---|---|
| ERA5 reanalysis via Open-Meteo | Daily sunlight, temperature and rain, 2021–2025, five sites | CC BY 4.0 | Real |
| NASA POWER | Monthly sunlight, an independent cross-check | Open data | Real |
| OpenStreetMap via Overpass | River centrelines for three stretches | ODbL | Real |
| Open-Meteo forecast API | Hourly sunlight that drives the simulated pods' solar input | CC BY 4.0 | Real |
| Prahari fleet simulator | Pods, telemetry, faults and tasks, seeded | Trinetra | Simulated |
| Reference design v0 | Power draws and indicative prices | Trinetra | Indicative |
- Deployment engineers choose the array, the battery and the hull for a site, and see the margin over five real monsoons.
- Outpost commanders see which water a plan covers, by night and by sensor, and where the gaps are.
- Maintenance teams get a ranked task list and a day-by-day boat plan per stretch that reaches the urgent pods first.
- Planners and procurement get a BOM, a cost per kilometre and the approvals to line up before a pilot.
- Pods count and track contacts: boats, debris, wildlife. No face recognition and no identity matching.
- Video stays on the pod in a 72-hour buffer unless an operator flags an event.
- Software recommends; trained people decide. A person in the water is planned for as a rescue target.
- Non-biocidal hull coatings, because the Sundarbans pods sit in a tiger reserve and World Heritage site.
- The international boundary is not drawn. On the Brahmaputra the planning line stops 1 km short of it; where the Ichamati is the boundary, pods are planned 40 m in from the Indian bank. See the doctrine.
- The design is a reference for planning, not a product specification, a quote or a vendor choice.
- Sensor ranges are planning values; each must be measured on the actual stretch.
- ERA5 is a 25 km reanalysis and differs from NASA POWER by up to 32 % in winter; size with the conservative series.
- River lines come from OpenStreetMap and move every monsoon. Re-plan on a fresh survey.
- The fleet console runs on simulated pods. It shows how the rules and queue work, not real failure rates.
Pilot offer: two pods and a shore relay on one stretch for one monsoon, logging on-site sunlight and measured detection ranges against these plans. contact@trinetra.in