Projects

Thesis submitted October 2024

IoT-Driven Smart Supply Chain

Undergraduate research into post-harvest loss in Sri Lankan farming — field interviews, requirements, and the design of a sensor-to-dashboard system.

Research & system design · 2024 · Sole researcher & designer

What it is

My final-year dissertation for the BSc in Management Information Systems. It asks how IoT-driven supply-chain management could reduce post-harvest losses for Sri Lankan farmers, and answers in two halves: qualitative research into where the chain breaks, and the design of a system to give it visibility.

The problem

A large share of produce is lost between field and buyer — to transport, storage and handling — and farmers often sell at prices far below what consumers eventually pay. The chain is fragmented across farmers, transporters, wholesalers and retailers, and almost nothing is measured along the way, so decisions are made without data.

From fieldwork to a system design

  1. 01

    Field research

    Interviews and case studies with agricultural experts at Gannoruwa and with farmers and sellers at Peliyagoda Manning Market.

  2. 02

    Requirements

    Questionnaire mapped to research objectives; stakeholder onion model; use cases such as tracking produce from farm to warehouse.

  3. 03

    Sense & track

    Environmental sensors (temperature, humidity, soil moisture) and RFID/GPS tracking proposed for produce in transit.

  4. 04

    Store & analyse

    Time-series storage (InfluxDB) behind a Python/Django API, with a simple trend rule turning readings into recommendations.

  5. 05

    Dashboard

    React prototype with views for farmer data entry, price lists, stock and sales, charted with Recharts.

My role

Sole researcher and designer: research design, interviews and case studies, requirements and stakeholder analysis, system design (use cases, ERD, architecture), and a React dashboard prototype.

Decisions that mattered

  • Start with the people, not the sensors. The research began with interviews at the market and with agricultural experts, not with hardware. A recurring answer was that farmers would adopt IoT tools if they reduced manual labour and made yield prediction more accurate — and that supply-chain managers needed real-time logistics data on a simple dashboard. Those answers became the requirements.
  • A time-series store for readings. Sensor readings are append-only, timestamped and queried by time range. The design stores them in a time-series database (InfluxDB) indexed on timestamp, with the API querying windows such as the last two weeks for a given sensor.
  • Different scales, different entry paths. Small farmers need an interface that is intuitive and asks for very little input; larger operations want as much as possible captured automatically through sensors. The requirements treat these as two paths into the same data.

What is not finished

This was academic research. The dashboard prototype ran against simulated and placeholder data; nothing was deployed in the field, and I make no claim about measured reductions in loss.

Built with

  • Qualitative research
  • Systems analysis
  • ERD & use-case modelling
  • React
  • React Router
  • React Bootstrap
  • Recharts
  • Axios
  • Django (designed)
  • InfluxDB (designed)