Livestock Health and Movement Monitoring Collar with LoRa Alerts
A low-power collar for dairy farmers that classifies cattle behaviour such as grazing, ruminating, walking and lying from motion data, and alerts when an animal drifts from its own baseline. An advanced IoT and machine learning final year project where battery life over LoRa shapes every decision.
Advanced
Needs a team, real planning, and a supervisor who knows the area.2 semesters · 3 students
7 to ship
3 optional extrasSuggested stack
What you should ship
- A collar prototype logging 3-axis acceleration and running an on-device classifier for grazing, ruminating, walking, lying and standing
- A labelled dataset from at least 5 animals, with video or observer notes aligned to sensor timestamps
- A classifier validated on animals it was not trained on, reporting per-class precision and recall
- A measured power budget per subsystem (sensor, CPU, radio, sleep) with projected and measured battery life
- LoRa transmission of hourly behaviour summaries, not raw data, with a documented duty cycle and packet loss survey on the farm
- A dashboard showing daily time budgets per animal and alerts when rumination or activity deviates from that animal's recent baseline
- An enclosure and strap design that survives two weeks of outdoor wear, with failures recorded
If you have time left
- Solar energy harvesting to extend deployment beyond one battery
- GPS fixes a few times a day to show grazing distribution across a field
- Linking alerts to vet records to see how early behaviour changes preceded diagnosed illness
The problem farmers have spotting sick animals early
Illness in cattle usually shows up in behaviour before it shows up as obvious symptoms. A cow that ruminates less, eats less or lies down more than usual may be heading towards mastitis, a digestive problem or lameness. Catching that a day early can mean a cheaper treatment and less milk lost. A farmer with 40 animals cannot watch each one all day, and on small and mid-sized farms, including many in Pakistan, checks are done by eye at milking time.
Commercial monitoring exists and works, but it is priced and supported for large dairy operations in Europe and North America. Smaller farms, and research groups at agricultural universities, need something cheaper and more open.
What already exists and the gap
- SenseHub Dairy (from MSD Animal Health, formerly Allflex) uses collars and ear tags to track rumination, activity and heat, aimed at large commercial dairy herds.
- CowManager uses an ear sensor to monitor eating, rumination, activity and ear temperature.
- smaXtec uses a bolus that sits in the stomach and measures internal temperature and activity.
These are closed systems sold with a farm subscription and local dealer support, built around large barns with fixed receivers. The gap is an open, low-cost collar with transparent classification that works on pasture over long-range radio, runs for months, and whose alerts are tuned per animal rather than to a herd average.
What you build
- Collar: microcontroller, accelerometer, LoRa radio and battery in a weatherproof box on a neck strap. It samples motion, classifies behaviour in short windows on the device, and keeps running totals.
- Gateway: one LoRa gateway at the farm forwarding summaries to a server over mobile data, with store and forward if the internet drops.
- Server and dashboard: each animal's daily time budget (grazing, ruminating, lying, walking), trends over weeks, and alerts when today looks unlike that animal's last 7 to 14 days.
Users are the farmer or herd manager, who sees alerts on a phone, and a vet or researcher, who sees the history.
The technically interesting part
Power drives everything. The collar must run for months on a battery an animal can wear, so streaming raw acceleration is out. That forces feature extraction and classification on the device, with only summaries sent over LoRa. Measure each subsystem with a current meter or power profiler, build a budget table, then choose sample rate, window length and radio interval from it. That table is one of the most valuable pages in your report.
The second part is per-animal baselines. Animals differ; a fixed rumination threshold will constantly flag naturally quiet cows and miss changes in active ones. Compare each animal against its own recent history, for example with a rolling median and deviation, and show how many alerts each approach would raise.
How to evaluate it
- Classification: leave-one-animal-out accuracy, per-class precision and recall against observed ground truth. Rumination is the class that matters most.
- Power: measured current per mode, projected battery life, and actual runtime of at least one collar over the deployment.
- Radio: packet delivery rate across the farm at different distances and with animals lying down.
- Alerts: alerts per animal per week with fixed thresholds against per-animal baselines, reviewed with the farmer or a vet.
Suggested plan
- Month 1: secure access to a farm or a veterinary university herd; build a logger-only collar.
- Month 2: collect and label data with video; train offline classifiers in Python.
- Months 3 to 4: port the chosen model to the microcontroller; power measurements and optimisation.
- Month 5: LoRa link, gateway, server and dashboard.
- Months 6 to 7: farm deployment, alert tuning.
- Month 8: evaluation and write-up.
Path to a product
The first customer is a mid-sized dairy farm or a livestock research group at an agricultural university that wants open data. A low-cost collar sold with a per-animal monthly fee for the dashboard fits farms that cannot afford commercial systems. Livestock departments and development NGOs running dairy programmes are realistic pilot partners, and the same hardware extends to goats and sheep.
Scope warning
Behaviour classification and alerting only. Do not build a herd management system with breeding, feed and milk records, and do not claim automated veterinary diagnosis. Arrange farm access first: without ground truth labels, there is no project.
Learn what you need on Sefism
- Machine learning roadmap : classification, validation and features for time series.
- Machine learning cheat sheet : metrics and model choices at a glance.
- Google Colab : free notebooks for training before you port the model.
- OpenIntro Statistics : baselines and deviation for per-animal alerts.
- Programming roadmap : C and low-level fundamentals for firmware.
Get help with this idea
This is a demanding project with a real dependency on a farm and on hardware lead times. Tauseef can help you decide what to cut, pick hardware your team can source, and plan an evaluation that holds up with a small herd. If you are new here, create a free account first.
Ideas and guidance, not finished projects
These are project ideas and scoping guidance, published free for students to use as a starting point. I do not build, write, or sell final-year projects, and I do not complete coursework for anyone. Take an idea, make it yours, and build it.
Not sure this one fits you?
I help students pick an idea that matches their skills and their deadline, then scope it down until it is finishable. Guidance only, never done for you.
Stuck on something specific?
Writing only gets you so far. If you want an answer to your situation rather than the general case, book a session and we will work through it together. Sessions are free for approved Sefism members, and a few slots open each week.
Follow along
New writing, resources and project ideas land here first.