TrustFarm AI — Silent Force
TrustFarm AI

Silent Force · Trust before advice

Waking the 3D farm…

Showcase System 3D flow Evidence Team 3D circuit
TrustFarm AI
ICBT · INNOVA 2026 · IEEE Computer Society

TrustFarm AI

Trusted AI irrigation and farm energy

Low-cost field sensors fail quietly — and the model still produces an answer. TrustFarm verifies every reading before the AI is allowed to advise.

3 hforecast horizon
36 × 5min steps
0–100trust score
60 stelemetry publish
Demo data unless hardware is connected.
Scroll to explore
Project showcase · INNOVA 2026

See the film. Read the deck. Open the build.

The full project walkthrough, the original submission and the live 3D hardware twin — together, without the long scroll.

Project film · OneDrive

Trust farm project - Present video

OneDrive

A guided walkthrough of the problem, the trust engine, the 3D twin and the live demo.

Download video

OneDrive blocks embedded playback. Open the full film in a new tab, or download it for offline viewing.

Original submission

Project Presentation

25 slides
TrustFarm presentation slide 1
01 · Title & vision1 / 25

Rendered preview of the original PPTX — all 25 slides, unchanged.

Rendered slide preview. Wording is unchanged from the submitted PPTX.

Interactive hardware · Three.js

Interactive 3D digital twin.

A real browser-based hardware model — not a rendered mock-up.

ESP32Labelled ESP32, probes, INA219, relay, battery and generator.
3D circuit
TrustFarm 3D circuit with ESP32, soil probes, relay, battery and generator Live 3D twin Open the model →
Solution

One loop: sense, verify, forecast, advise, power.

Trust-gated AI plus energy-aware irrigation advice — demonstrated in a live 3D twin.

Sense

ESP32 + INA219

Verify

Trust Engine

Predict

3-hour forecast

Advise

Explained action

Power

Energy gate

0/100Trust score
Green 0–100
score = 100 − penalties
Data-quality gate

The model is not always allowed to answer.

A 0–100 trust score decides whether the forecast becomes advice, a request for a manual check, or a pause and an alert.

GreenAdvice allowed — irrigate or prepare, with a reason.
AmberPredict, but ask for a sensor check. No automatic irrigation.
RedPause and alert. Sensor data is not trusted.
Green≥ 80

Advice allowed

Amber50–79

Predict, but ask for a manual check

Red< 50

Advice paused; alert raised

3-hour forecast

Soil moisture, 36 steps ahead. A GradientBoosting model forecasts soil moisture; a balanced classifier turns it into Low / Medium / High need.

Energy-aware advice

Water only when power allows. Battery state and generation can turn “irrigate now” into “delay or use manual supply”.

Live system flow

Watch a reading become trusted advice.

One 3D pipeline: field sensors, MQTT gateway, trust gate, forecast and farmer app — with a fault cycle that pauses advice.

Live trust pipeline GREEN · advice allowed
Drag to orbit · scroll to zoom
Open the full 3D circuit

FAULTFault simulation runs every 18 s

Evidence

A working prototype, labelled honestly.

Apps, datasets, models and results — with measured data separated from synthetic tests.

One app, three views.

Live overview, 24-hour analytics and read-only alerts — on desktop and mobile.

OverviewAnalyticsAlerts
Live dashboards · 3D twin · trust historyNot yet: push notifications · pump control · production TLS

Four datasets. One honest evidence label.

Public data supports development. It is not a substitute for a field pilot.

AquaIntelMeasured AgriField-ManipurSynthetic Biogas Swine FarmSynthetic Semantic-Aware IoTSynthetic

What the numbers actually show.

Forecast error improves against persistence. Fault detection is still below the target.

52.3%lower MAE vs persistence
59.5%lower RMSE vs persistence
47.5%sensor-fault detection · below 90% target
Synthetic held-out results, not field measurements. Do not present them as field performance.

Implemented in code: field ingest, trust gate, forecast, advice, monitoring and apps. Proposed / supervised build: digester, gas cleaning, generator, charge controller and BMS.

Team · ICBT

Four disciplines. One integrated system.

AI & backend · web & mobile · IoT & communication · deployment & integration.

KK

G. W. W. M. R. Kithara Kirinda

AI & Backend · Team leader

RM

W. P. A. Rasith Maleesha

Web & Mobile Applications

SB

B. V. D. T. Senmitha Binaragama

IoT & Communication

KA

B. W. A. M. Kaushalya

Deployment & Integration

“Trust before advice.”
INNOVA 2026 · ICBT

Trust before
advice.

Watch the film, open the original deck, or explore the live 3D circuit. Demo data unless hardware is connected.