The short answer
An embedded system is a dedicated computer built into a device: a microcontroller and its firmware controlling a washing machine, a thermostat or a car’s brakes. A cyber-physical system (CPS) is a larger engineered system in which computation, networking and physical processes are tightly woven together, with sensors and actuators forming feedback loops that monitor and control the physical world.
The relationship is mostly one of scale and viewpoint:
- An embedded system is a component. You look at it as hardware plus firmware.
- A cyber-physical system is a system. You look at the computers, the network and the physical process together, including how they affect each other over time.
Most cyber-physical systems are built from many embedded systems. Not every embedded system is a meaningful CPS on its own, though. A digital alarm clock is embedded, but there’s little physical process for it to control.
What is an embedded system?
An embedded system combines a processor (usually a microcontroller), memory, peripherals and software designed for one specific job inside a larger product. Typical characteristics:
- A dedicated function rather than general-purpose computing
- Tight limits on memory, processing power, cost and energy
- Direct interaction with hardware through sensors, actuators and buses like I2C, SPI and CAN
- Often real-time behaviour, meaning responses must happen within a known deadline
- Long, unattended operation
The four types of embedded systems
Embedded systems are commonly grouped into four types: standalone (a microwave controller), real-time (an airbag controller), networked (a smart thermostat) and mobile (a fitness tracker). The categories overlap, and our pillar on embedded systems and the Internet of Things covers each with examples. Real-time and networked embedded systems are the ones most likely to show up as parts of a CPS.
What is a cyber-physical system?
A cyber-physical system integrates computing and communication with physical processes so closely that you can’t understand either side on its own. Sensors measure the physical world, software decides what to do, and actuators change the physical world, which the sensors then measure again. The physical process and the computation keep influencing each other in a closed loop.
Common features of CPS:
- Feedback control over a physical process (motion, temperature, pressure, power flow, chemical reactions)
- Networking between multiple computing nodes, sensors and actuators
- Timing that matters: a correct command delivered too late can be as bad as a wrong one
- Concurrency: many things happen at once, in the physical world and in software
- Safety consequences: failures can damage equipment, the environment or people
Where the term comes from
The term “cyber-physical systems” emerged in the mid-2000s, closely associated with the US National Science Foundation, which launched a research programme around it. The idea was to treat the computing, networking and physical parts of systems like aircraft, power grids and medical devices as one engineering problem, instead of designing the software separately and bolting it on. A well-known textbook, Introduction to Embedded Systems: A Cyber-Physical Systems Approach by Edward A. Lee and Sanjit A. Seshia, reflects how closely the two fields are linked.
Some explanations describe CPS as “digital models of physical systems.” That’s closer to a digital twin, which is a related but separate idea. A digital twin is a software model that mirrors a physical system. A CPS is the real system that senses and acts. Digital twins are often used to design, test or monitor CPS.
Cyber-physical systems vs embedded systems: side by side
| Aspect | Embedded system | Cyber-physical system |
|---|---|---|
| What it is | A dedicated computer inside a product | An integrated system of computation, networking and physical processes |
| Scale | Usually one device or board | Often many devices, networks and physical subsystems |
| Focus | Hardware and firmware doing a task reliably | The interaction between software and physical dynamics |
| Physical interaction | Optional (some only process data or show information) | Essential, usually through feedback control |
| Networking | Optional | Usually central, linking sensors, controllers and actuators |
| Timing | Matters for real-time types | Matters across the whole system, including network delays |
| Analysis | Code correctness, resource use, worst-case execution time | Plus control theory, physical modelling, system safety |
| Typical disciplines | Embedded software, electronics | Plus control engineering, mechanical or electrical engineering, systems engineering |
| Examples | Microwave controller, digital watch, motor driver board | Smart grid, autonomous vehicle, industrial process control, surgical robot |
How they relate: embedded systems are the building blocks of CPS
The easiest way to see the difference is to zoom out from one controller.
Worked example: from a brake controller to a connected vehicle
- One embedded system. An anti-lock braking (ABS) controller is a microcontroller running firmware. It reads wheel-speed sensors and modulates brake pressure through valves. On its own it’s a classic hard real-time embedded system.
- A cyber-physical subsystem. Look at the ABS controller together with the wheels, tyres, road surface and hydraulics, and you have a feedback loop between software and vehicle dynamics. Its correctness depends on physics as much as code: how fast wheels lock up, how pressure builds in the lines, how long the control loop takes.
- A larger CPS. A modern car has many electronic control units (ECUs) for the engine, brakes, steering, airbags and driver assistance, connected over networks like CAN and automotive Ethernet. Stability control coordinates braking and engine torque across several ECUs. Now network latency, message priorities and interactions between subsystems are part of the design.
- A system of systems. Add vehicle-to-infrastructure communication, cloud services and over-the-air updates, and the car becomes part of a much larger cyber-physical and IoT system.
The microcontroller in step 1 didn’t change. What changed is how much of the surrounding physical and networked world you treat as part of “the system.”
What makes CPS engineering different
Writing correct firmware is necessary but not enough for a CPS. Extra concerns include:
- Physical dynamics. Engineers model the plant (the physical process being controlled) and design control laws that keep it stable. A bug-free program can still drive an unstable system.
- End-to-end timing. Sensor sampling, computation, network transmission and actuation all add delay. Too much delay, or too much variation in it (jitter), can destabilise a control loop.
- Concurrency and distribution. Multiple controllers act at once, often without a shared clock. Clock synchronisation and coordinated behaviour become design problems.
- Uncertainty. Sensors are noisy, actuators saturate, and the environment changes. Robust designs handle this gracefully.
- Safety. Many CPS fall under functional safety standards such as IEC 61508 (general industrial), ISO 26262 (road vehicles) or IEC 62304 (medical device software). These standards shape the whole development process.
- Security. An attacker who compromises a CPS can cause physical harm, not just data loss. That’s the focus of What is cyber-physical security?
Where IoT fits
The three terms overlap but emphasise different things:
- Embedded systems focus on the dedicated computer inside a device.
- IoT focuses on connectivity: devices sending data to and receiving commands from networks and cloud services. Many IoT applications mostly monitor, such as a temperature logger or a parcel tracker.
- Cyber-physical systems focus on closed-loop interaction with physical processes, often with tight timing and safety requirements.
A smart thermostat that adjusts heating is a small CPS built on an embedded system and connected through IoT. A soil-moisture sensor that only reports readings to a dashboard is IoT, but it’s a weak example of CPS because nothing closes the loop automatically. Industrial control systems may be strongly cyber-physical with very little internet connectivity at all. For the IoT side, see IoT System Architecture.
Examples
Embedded systems (not usually called CPS on their own):
- Digital watches and calculators
- TV remote controls
- A USB keyboard controller
- A network router’s management processor
Cyber-physical systems:
- Power grids with automated protection and load balancing
- Autonomous and driver-assisted vehicles
- Industrial process control in chemical plants and water treatment
- Fly-by-wire aircraft flight control
- Medical devices such as insulin pumps with closed-loop control
- Collaborative robots on a factory floor
- Building management systems controlling HVAC, lighting and access
FAQ
Is every embedded system a cyber-physical system? No. Many embedded systems interact with the physical world in only a minor way, or not at all beyond buttons and a display. The CPS label fits best when the system’s main job is to monitor and control a physical process through feedback.
Is a robot a cyber-physical system? Yes, robots are a textbook example. They combine embedded controllers, sensors, actuators and often networking, and their behaviour depends on both software and physical dynamics.
Is Python or C++ better for embedded systems? For microcontroller firmware and anything time-critical, C and C++ remain the standard because they give predictable timing and direct hardware control with little overhead. Python is widely used for testing, tooling, data analysis and prototyping, and MicroPython runs on some microcontrollers. Python also appears on Linux-based embedded devices where timing is less strict. Many engineers use both: C or C++ on the device, Python on the bench.
What does CPS stand for? In engineering, CPS usually stands for cyber-physical system, but the acronym has several other meanings in computing. We list them in What does CPS stand for in the context of computers?