Module 5 — AI + IoT | AI & Robotics Training

Topic 01 · Connected things

When everyday objects talk to the internet

IoT means Internet of Things — devices that send small updates (temperature, door open, location) so apps can show charts or take action.

💡 Like Module 4: Robots and factory arms send sensor readings over the network. IoT connects everyday objects the same way.

What makes something “IoT”? It is not just “has Wi‑Fi.” It means: (1) sensor or actuator, (2) send data, (3) store, show, or act.

Three parts: (1) Device measures. (2) Gateway or cloud collects. (3) App or AI alerts or commands.

Messages are tiny — often dozens of characters, not video.

MQTT uses topics like school/greenhouse/sensor3/temp. Devices publish; apps subscribe.

Actuators turn fans, locks, or relays on — IoT can command back.

Figure — Measure → send → store → show or act.

Figure — Many devices publish; many apps can listen.

Where IoT shows up — industries and goals

Place	Measured	Why	Example
Home	Temp, door, leak	Comfort, security	Thermostat, leak alarm
School	CO₂, occupancy	Fresh air, save power	Classroom monitor
Farm	Soil, weather	Water when needed	Field sensor
Factory	Vibration, heat	Prevent breakdown	Motor tag
Trucks	Temp, GPS	Food safety	Cold chain
City	Traffic, air	Planning	Smart parking

Message styles — types and uses

Style	Good for	Plain English
MQTT	Many small updates	Publish/subscribe channels
HTTP	Simple apps	Like loading data from a URL
WebSocket	Live dashboards	Server pushes updates fast
LoRaWAN	Long range, battery	Slow small packets from fields

Rules first “If temp high 20 min, alert” needs no machine learning.

Clear topics Paths like school/lab/sensor1 help when you have many devices.

Local hub A Pi in the building can work when internet drops.

Common IoT sensors — what they measure and what they are used for

Sensor	Measures	Used for	Example device
Temperature	Heat	Fridges, greenhouses, trucks	Smart thermostat, cold chain tag
Humidity	Moisture in air	Mould risk, plant care	Greenhouse monitor
Door / window	Open or closed	Security, heating waste	Smart door sensor
Motion (PIR)	Movement	Lights on, intruder alert	Hallway sensor light
Light level	Brightness	Auto blinds, street lights	Smart office lighting
Vibration	Shaking	Machine about to break	Factory motor monitor
CO₂ / air quality	Stale air	Fans in classrooms	Classroom air monitor
Water leak	Moisture on floor	Stop flood damage	Under-sink alarm

Example — cold truck: A sensor sends temperature every few minutes. If it stays too high, the driver gets a text. Simple rules work; AI can add “predict breakdown from vibration.”

Example — school greenhouse: Soil moisture every 10 minutes; web chart; email if dry 2 hours on weekend.

Example — leak sensor: Wet floor sends one message; phone buzzes in seconds. No camera required.

Gateways and hubs — types and what they do

Device	Role	Used for
Wi‑Fi board (ESP32)	Sensor + internet in one	Student projects, prototypes
Home hub / bridge	Translates radio to Wi‑Fi	Some smart bulbs, sensors
Industrial gateway	Collects many factory sensors	Plant floor to cloud
Phone as gateway	Bluetooth device → phone → cloud	Wearables, some health gadgets

Sample JSON message (what many systems send):

{"device_id": "greenhouse-3", "temp_c": 24.1, "humidity": 58, "battery_pct": 72, "time": "2026-05-16T14:30:00Z"}

Your project should show one real message like this in the README — judges understand the system faster.

Name three smart devices at home

What does each measure? Where does the data go? Could someone else see that data?

Try it · Draw your IoT chain on paper

Device → internet → storage → app → action. Label one sensor reading and one possible alert.

Topic 02 · Smart devices

“Smart” can be simple or fancy

A smart thermostat might just follow a schedule, or it might learn your routine. More “intelligence” needs more data and testing.

Rule-based: If room empty 30 minutes → lights off.

Predictive: Heat the room before you arrive because it learned your pattern.

Digital twin: A computer model of a building or machine that updates when real sensors update — used to ask “what if?”

Types of smart devices — what they do

Device	What it does	Used for
Smart plug	Turn power on/off remotely	Lamps, fans, coffee machine
Smart bulb	Dim or colour change	Mood lighting, schedules
Smart thermostat	Control heating/cooling	Save energy, comfort
Video doorbell	Camera + speaker at door	See visitor, talk from phone
Smart lock	Lock/unlock with phone or code	Home entry without keys
Smart speaker	Voice commands	Music, timers, smart home control
Smart meter	Electricity use	Bills, saving energy
Wearable	Steps, heart rate	Fitness trends
Pet feeder	Schedule + low food	Feed when away

Example — campus classroom: CO₂ sensor + fan control keeps air fresh during exams without running fans all night.

Situation	Use	Why
Clear threshold	Rules	Fast, auditable
Weekly patterns	Learning	Hard to write all rules
Safety alone	Human + rules	Do not auto-lock exits on AI guess

Topic 03 · Edge vs cloud

Where should the “thinking” happen?

Edge = on the device or nearby box (fast, private). Cloud = big computers far away (powerful, needs internet).

Latency is delay. Edge can answer in milliseconds; slow internet makes cloud feel sluggish on doorbells.

Bandwidth is how much data you send. Sending “person: yes” is cheaper than uploading every video frame.

Privacy: On-device processing can keep video at home; cloud needs encryption and clear delete rules.

Figure — Filter on device; upload clips only when needed.

How fast is fast enough?

Job	Target	Where
Doorbell person	Under 1–2 s	Edge
Energy report	Hours OK	Cloud
Factory e-stop	Milliseconds	Edge / wired, not slow cloud

Example — fitness watch: Counts steps on wrist; syncs summary when phone has Wi‑Fi.

Figure — Many products use both: quick guess on device, hard cases sent to cloud.

When to use edge vs cloud

Choose	Good when	Example
Edge (on device)	Need fast reply, privacy, or spotty internet	“Someone at door” in 1 second
Cloud	Need long history or very large AI	Monthly energy report for whole building
Both	Device filters, cloud confirms	Doorbell: local detect, cloud stores video

On device Good for quick reactions (doorbell person detection). Saves privacy. Limited battery.

In cloud Good for huge models and storing history. Needs good Wi‑Fi.

Mix both Device spots “maybe a person”; cloud double-checks hard cases.

Example — coffee shop counter: Count cups locally, send only hourly totals to cloud — saves phone-style data bills.

Topic 04 · Staying safe

More devices = more doors to lock

Default passwords, old software, and fake messages are real problems. Good habits matter as much as AI.

Safety habits — what they protect

Habit	What it stops	Example
Change default password	Strangers logging into camera	password “admin” bots
Update firmware	Known security holes	OTA update from maker
Separate guest Wi‑Fi	Visitors reaching your devices	Guest network for friends
Encrypt video	Leaked private footage	Doorbell cloud with lock icon
Confirm big actions	Fake sensor data causing harm	Human OK before factory stop

Change default passwords; use unique ones.
Update firmware (OTA = over-the-air updates).
Do not put private video on open internet without encryption.
Segment networks: guest Wi‑Fi separate from factory machines.

Attack surface: Every camera and plug is a door. IoT is often forgotten after setup — still using factory passwords.

Fake data: Spoofed “temperature OK” on a cold truck can spoil food. Big actions need human OK or a second sensor.

Common risks — types and fixes

Risk	How	Fix
Default password	Bots scan “admin”	Change on day one
Old firmware	Known bugs	Auto-updates if offered
Open camera	No login on internet	Firewall, strong account
Fake MQTT	Spoofed message	Passwords; sign messages

Privacy — sensitive data

Data	Sensitivity	Care
Room temp	Lower	Do not publish openly
Indoor camera	Very high	Consent, encrypt, limit retention
Speaker mic	Very high	Clear when listening

Inventory List every lab device — know what is still online after term ends.

No secrets in Git Never commit real Wi‑Fi passwords to public repos.

Guest network Visitors should not reach your cameras.

Example — visitor Wi‑Fi: Hospital guests cannot reach medical devices. Same idea at home for cameras.

Example — camera botnet: Thousands of cheap cameras with password “admin” were hijacked. Lesson: boring security stops big disasters.

Discuss · What could go wrong if someone fakes sensor data?

Think heating, doors, or factory stops. Who should confirm big actions?

Checklist · Before you leave devices running after class

Changed default passwords?
Documented what each device does?
Turned off test cameras pointing at people?
Removed API keys from public project repos?

Check Your Understanding

Quick Knowledge Check

10 easy questions on smart devices, the internet of things, and safety. Instant feedback on every answer.

Score: 0 / 0

Module Wrap-Up

Key Takeaways

Module 5 in short: devices talk to the cloud (or locally) — good data and passwords matter.

01 IoT = everyday things online, often sending sensor readings.
02 Smart doorbells, thermostats, and lights are familiar examples.
03 Edge = work on the device; cloud = bigger computers far away.
04 MQTT is a common way devices send short messages.
05 Digital twins are computer copies that update with real sensors.
06 Change default passwords and install updates (OTA) to stay safer.
07 Bad or fake data can trigger wrong actions — security matters.
08 AI can spot patterns in sensor streams (heat, vibration, motion).

↑ Back to Top

📚 Further reading:
• MQTT specification overview — mqtt.org
• AWS IoT reference architectures — aws.amazon.com/iot
• TinyML community resources — tinyml.org

AI Meetsthe Internet of Things

When everyday objects talk to the internet

“Smart” can be simple or fancy

Where should the “thinking” happen?

More devices = more doors to lock

Quick Knowledge Check

Key Takeaways

AI Meets
the Internet of Things