Internet of Things and Smart Device Networks

1The subject and structure of the lecture6 min

The Internet of Things (IoT) describes the integration of physical objects, sensors, actuators and embedded systems into communication infrastructures and digital services. Unlike a traditional computer network, where data is mainly entered by users, an IoT system automatically and continuously collects information about real objects and processes.

Learning outcomes

• Describe the layered architecture of an IoT system (device, edge, cloud)
• Choose a network topology (star, tree, mesh) based on the application requirements
• Compare IoT communication technologies based on range/rate/consumption
• Distinguish the MQTT model (publish/subscribe) from the CoAP model (resource-based)
• Calculate the energy budget of a radio transmission, including retransmissions
• Describe the lifecycle of an IoT device and the associated security requirements

2From embedded system to IoT node8 min

A traditional embedded system performs a well-defined function and can operate completely in isolation - a temperature controller, a motor drive module. By adding connectivity, the system can transmit measurements, receive configurations, report errors, cooperate with other devices and be updated remotely.

Sensors turn physical quantities into digital data; actuators do the reverse. For a physical quantity x(t), acquisition generates samples x[k] = x(kT_s). The data can be filtered locally, z[k] = F(x[k], x[k-1], ...), and the result can generate a command u[k] = G(z[k], r[k]).

3Characteristics and requirements of IoT systems10 min

IoT systems combine resource-constrained devices, heterogeneous networks and distributed services. An architecture must work not just for a prototype with a few nodes, but under real operating conditions: packet loss, power interruptions, unpatched devices, gateway failures, a growing number of nodes.

Scalability

Energy, latency and reliability

Local intelligence and autonomy

Local processing reduces the volume of transmitted data and lowers latency - instead of transmitting every sample, the node can send only average, minimum, maximum values, events or the result of a classification.

Functional autonomy also implies the ability to keep essential services running while disconnected from the cloud - the node or gateway can temporarily store measurements, apply local rules and retransmit the data once the connection is restored.

4The architecture of an IoT system10 min

A practical model for IoT architecture includes four layers: devices and physical processes, communication, edge and gateway, services and applications - a model for organizing responsibilities, not a single universal standard.

The upstream flow carries the data: sensor → node → gateway → service → application. The downstream flow carries configurations, thresholds, commands, updates and acknowledgments - together they form the control loop measurement → decision → command → process.

The role of the gateway

The gateway connects the local device network to the IP or cloud infrastructure: protocol conversion, message aggregation, temporary storage, node authentication, terminating secure connections and handling updates. It is not mandatory in every system - a Wi-Fi or cellular device can talk directly to the cloud - but it becomes useful when the nodes use local protocols, have limited resources, or must operate without Internet access.

5Smart sensor nodes10 min

The sensor node is the direct interface between the IoT system and the physical environment. The main components are similar regardless of the application: sensors and conditioning circuits, an analog-to-digital converter, a microcontroller, memory, a transceiver, a power management circuit.

Validation can detect out-of-range values, sensor communication errors, incomplete data or physically impossible variations. A useful message does not contain only the measured value - it usually includes the node identifier, the timestamp, the unit of measure, the battery status and the sequence number.

Cyclic operation

6Topologies and the organization of IoT networks10 min

Topology describes how devices are logically connected and the paths information travels through - it affects consumption, latency, fault tolerance and the ability to add new nodes. It must be distinguished from the communication technology: a network based on IEEE 802.15.4 can be organized as a star, a tree or a mesh, depending on the protocol used.

In a star topology, the peripheral nodes do not relay other devices' messages - they can stay in low-power modes most of the time, which is why the star topology is common in Wi-Fi, BLE with a central device, LoRaWAN and cellular IoT networks.

The energy of a node with a routing function: E_node = E_own + E_reception + E_relay - nodes placed close to the gateway may relay a significant part of the network traffic and can drain faster than peripheral nodes, an important asymmetry to anticipate at design time.

7Communication technologies for IoT10 min

The communication technology must be chosen based on the application, not just on the advertised maximum range - the real parameters depend on frequency, transmit power, antenna, obstacles and interference.

IEEE 802.15.4 defines only the physical layer and medium access - Zigbee, Thread and 6LoWPAN add network and application functions on top of it. Thread is oriented toward IPv6 (it joins an IP network through a border router); Matter is an application-layer protocol for interoperability, which can run over Thread, Wi-Fi or Ethernet - it is not a new radio technology.

Technology selection should go through: the distance and propagation environment, the size and frequency of the messages, the acceptable latency, the power source, the available infrastructure, the link and energy budget, and the spectrum and security restrictions - in this order, not starting from a technology preferred in advance.

8IoT protocols and software stacks10 min

The radio technology establishes how bits are transmitted; a complete IoT system needs additional protocols for addressing, routing, transport, security and exchanging information between applications - a layered stack: physical/medium access, network/adaptation, transport, security, application.

IPv6 and 6LoWPAN

IPv6 uses 128-bit addresses (N_IPv6 = 2¹²⁸), but in networks with small frames, transmitting full IPv6 headers produces significant overhead. 6LoWPAN introduces an adaptation layer: header compression, fragmentation and reassembly for reduced frames.

MQTT - publish/subscribe

MQTT is based on an intermediary broker: a publisher publishes messages on a topic, and any subscriber subscribed to that topic receives them, without the producer knowing the recipients directly.

// a node publishes to:
agriculture/zone1/node12/temperature

// an application subscribes to:
agriculture/zone1/+/temperature

CoAP - resource-based model

CoAP (Constrained Application Protocol) is conceptually close to REST, with GET, POST, PUT, DELETE operations on resources addressed by a URI (coap://node12.local/sensors/temperature).

Regardless of the protocol, a useful message must specify the quantity identifier, the value, the unit of measure, the timestamp and the format version - real interoperability requires compatibility at several levels at once: connectivity, protocol, message structure, and data meaning.

9Edge, gateway and cloud infrastructures8 min

The processing of an IoT system can be distributed among nodes, gateways, edge servers and cloud infrastructures - not all data must be sent to the cloud. Choosing where to process depends on the maximum acceptable latency, the data volume, the available connectivity, privacy and the cost of communication.

Aggregation and offline operation

Instead of transmitting every sample, the gateway can aggregate: the average, minimum, maximum, standard deviation, event count, or only the values exceeding a threshold - x̄ = (1/N)Σxᵢ.

After communication is restored, the messages must be retransmitted without losing order and without uncontrolled duplication - the timestamp, the sequence number and the retransmission flag are essential for the application to correctly reconstruct the history.

10Managing IoT devices8 min

Administering an IoT fleet covers every operation from manufacturing the device to retiring it from service.

Enrollment assigns the device a verifiable identity - symmetric keys, asymmetric key pairs, digital certificates or an identity stored in a secure element. Every device must have a unique identity: using the same key for an entire product family drastically multiplies the impact of compromising a single unit.

During operation, the firmware version, battery level, communication quality, resets and sensor errors must be monitored - a typical diagnostic message contains exactly these fields, similar to measurement messages, but oriented toward the device's internal state rather than the application's data.

Firmware updates (OTA)

Over-the-air updates allow fixing vulnerabilities without physical access to the device: downloading the image, verifying authenticity and integrity, installing, rebooting, confirming correct operation.

When a device is retired, its identities must be revoked and its keys and sensitive data erased - a step often skipped, but just as important as the initial enrollment.

11Security of IoT systems8 min

IoT security must be designed for the whole system and its entire lifecycle - protecting only the communication is not enough if the firmware can be modified or if every device uses the same key.

An attacker can act on the device, the radio link, the gateway, the cloud services, the user application or the update process - the threats include interception, message tampering, replaying old messages, impersonating a node and installing malicious firmware.

The least-privilege principle applies here too: a node that publishes temperature readings must not be able to modify other devices' configuration or administer the broker - every component receives only the rights strictly necessary for its function.

12Representative IoT applications4 min

Representative domains for IoT include industrial automation and predictive maintenance, smart agriculture and environmental monitoring, urban infrastructure (Smart City), smart buildings, digital health and energy management.

Each domain combines the concepts of this lecture differently: agriculture typically combines LoRaWAN (wide coverage, infrequent messages) with local edge processing and offline operation tolerant of connectivity interruptions; industrial automation often combines local mesh networks with gateways that aggregate and transmit only relevant events; smart buildings frequently use Zigbee or Thread for the high node density over a small area.

13Frequent mistakes5 min

• "A mesh network is more reliable than a star network, because it has more paths." Not automatically - the reliability of a path with H hops is P_delivery = p^H, which decreases with the number of hops if there are no retransmissions or actively managed alternative paths. Evaluate the real reliability of the routing protocol, do not assume that more paths automatically means more reliability.
• "MQTT and CoAP are interchangeable, the choice depends only on message size." False - the choice depends on the interaction architecture: distribution through a broker (MQTT) vs. direct access to a resource (CoAP), a structural criterion, not just an efficiency one. Analyze the application's real interaction flow before choosing the protocol.
• "Encrypting the communication is enough for a secure IoT system." No - security must cover the entire lifecycle: Secure Boot, signed updates, unique per-device identities and the least-privilege principle, not just the communication channel. Build the threat model for the entire chain (device → network → gateway → cloud → application), not just for the data transport.

14Summary and glossary5 min

An IoT system extends an embedded node with an identity, connectivity and services - but this extension brings explicit trade-offs between latency, energy, reliability and scalability, which must be evaluated together, not in isolation. The layered architecture (device, edge, cloud) allows intelligent distribution of the processing, and choosing the right location depends on how critical and how time-sensitive a decision is. The network topology (star, tree, mesh) and the communication technology (BLE, Zigbee, LoRaWAN, cellular) must be chosen starting from the application's real requirements - range, rate, energy, available infrastructure - not from the popularity of a technology. MQTT and CoAP cover different interaction models (publish/subscribe versus resource access), and 6LoWPAN adapts IPv6 to resource-constrained networks. Managing an IoT fleet covers the whole lifecycle - enrollment, configuration, OTA updates, retirement - and security must be designed for this complete cycle, not just for the communication channel.

15Self-check questions6 min

1. What turns an isolated embedded system into an IoT node?
2. Why should a critical function with strict latency requirements not be moved exclusively to the cloud?
3. What is the difference between the star and mesh topologies, from the reliability perspective?
4. Why does the number of hops in a mesh network not automatically guarantee higher reliability?
5. What is the structural difference between the MQTT model and the CoAP model?
6. What role does 6LoWPAN play in an IEEE 802.15.4 network?
7. List the stages of an IoT device's lifecycle.
8. Why is an unprotected hash not enough for validating a firmware update?

16Closing the series2 min

This was the last lecture of the Embedded Systems series. The eleven lectures have covered the full path - from a processor's architecture and its energy consumption, through synchronization and real-time operating systems, hardware acceleration and harvesting energy from the environment, to integrating a node into a complete IoT network, with its management and security.

The concepts from this lecture - topologies, MQTT, gateways and device management - become real code and real measurements in Laboratory 04, where you build a complete IoT node, from sensor to dashboard.