What is a Cyber-Physical System?

Cyber-physical systems bridge the physical and digital worlds, where there’s a deep connection and interaction between the physical systems and the computational elements.

Cyber-physical systems now span different industries and technology applications — from autonomous vehicles (AVs) to smart grid technology — and use real-time data analytics and complex algorithms to control physical assets. Overall, cyber-physical systems integrate a level of intelligence into physical systems that enable them to make autonomous decisions, adapt to changing environments, and optimize their performance in real time.

There are several examples of cyber-physical systems being used to control the critical operations of advanced technology systems today.

Autonomous Vehicles

In autonomous vehicles, many physical and digital elements work tightly together. We are still a long way from seeing fully autonomous (level 5) vehicles, so autonomous vehicles require some level of human input to ensure passenger safety.

Autonomous vehicles have many sensors, from imaging cameras and radar to light  detection and ranging (lidar) sensors, which help the vehicle perceive its environment and make critical operational decisions. This can include “seeing” the presence of obstacles — for example, pedestrians, other cars, or sudden objects in the road — and changes in weather that alter how the vehicle behaves on the road.

However, autonomous vehicles wouldn’t function the way they do with physical elements alone; the software elements are just as important. Because there are so many different types of sensors, the software has to perform sensor fusion functions that homogenize the data into a single format that can be read, analyzed, and acted upon. Advanced algorithms — a mixture of classical algorithms, artificial intelligence (AI), and machine learning (ML) — need to be used in case there is a disparity among the sensor data. If different sensors perceive different things, then these algorithms are responsible for finding which sensors have provided truthful data and which have provided inaccurate data.

Then, another hardware layer needs to be present consisting of graphics processing units (GPUs) that can process the sensor fusion algorithms and any other control algorithms, such as if the car should brake or move the steering wheel when there’s an obstacle, that the vehicle uses to control the car’s physical elements. There have already been specific automotive GPUs developed by NVIDIA — the Jetson Xavier — that are efficient at detecting pedestrians, detecting other cars, and understanding the environment from the sensor data.

Smart Grids

Smart grids are a key example of cyber-physical systems and automation in the energy industry. They use a range of physical assets around the grid, especially sensors, to monitor many different grid aspects. These assets include smart inverters, solar panels, wind turbines, smart meters, legacy infrastructure, distribution and transmission lines, and battery energy storage systems (BESSs). While some aspects of smart grids are autonomous, operators are alerted to any issues before they occur (through predictive maintenance) or when they arise so action can be taken swiftly.

Aside from all these physical assets, smart grids have hierarchical layers of software architecture that connect utility companies, customers, and distribution system operators (DSOs). This software architecture uses a mixture of classical algorithms and machine learning to analyze, monitor, and optimize different assets in the grid. Many grids today contain legacy infrastructure that was not built for the digital world, but these can also be retrofitted with Internet of Things (IoT) sensors to make them compatible with newer digital-ready architecture. As well as monitoring existing assets, smart grids can monitor the integration of newly distributed energy resources (DERs), such as different renewable energy generators, to ensure they are working properly once integrated into the grid.

Sensors collect data on energy use in real time, which allows the operator to better optimize the distribution of energy around the grid. This can be in response to a range of environmental factors, including:

• Electric vehicles (EVs) using a lot of local power at the same time
• Renewable energy systems producing too much local power that can’t be stored (leading to transmission bottlenecks)
• Outages from weather and other damage to critical infrastructure

Without a tight network of entwined physical and digital architecture, smart grids would not be possible, and the energy industry would have to continue using grids that require a lot more manual upkeep.

While cyber-physical systems are closely related to IoT and can be used interchangeably at times, the general consensus is that there are small differences between the two types of technologies. However, the lines between what constitutes IoT versus a cyber-physical system have become blurred in recent years as the two technologies have further converged.

Some ambiguity remains, but the consensus is that the level of integration between the physical processes and digital assets is the main differentiator. This table shows some of the specific instances where IoT and cyber-physical systems differ:

The architecture of cyber-physical systems includes a mixture of physical hardware and software components, which provide intelligent, responsive machine behavior. The main components of cyber-physical systems are:

• Actuators
• Communication networks
• Computational nodes
• Control algorithms
• Sensors

The hybrid nature of these systems means they can perform many intelligent and real-time operations. Some of the key operational features and advantages of cyber-physical systems include their ability to:

• Operate in real time and quickly respond to changes in the physical environment for critical tasks that require immediate action.
• Exchange data and coordinate actions through connected devices.
• Learn from their own data to optimize their behavior.
• Possess advanced control and monitoring capabilities.
• Exert a high degree of control over the system’s physical components.
• Make informed, data-driven decisions using sensor data.
• Use a high heterogeneity with many different components to perform a variety of functions.

The Role of Digital Twins in Cyber-Physical Systems

One of the best examples of how the physical world influences the digital world and vice versa is digital twins. This virtual environment is influenced in real time by real-world data and assets. On the other hand, digital twins can simulate, optimize, and better understand how physical assets can be improved through simulating how they behave in different environments and scenarios. This virtual prototyping capability means that digital twins can be used to simulate many cyber-physical system applications.

Data is initially fed into the digital twin to build the environment. The resulting digital twin environment can be a small system, a large system like an autonomous vehicle, or a complete manufacturing facility. Once built, simulations can be run on various aspects of the environment to see how the assets will function in the real world.

Digital twins can also co-exist with physical assets that they are modeling. By feeding the data from the physical asset into the model in real time, it enables engineers to run different scenarios about their physical asset. This could include looking at how a vehicle behaves in different driving environments, distributing power in and out of an EV charging station, simulating smart city environments, running predictive maintenance operations in smart buildings, or changing manufacturing protocols in smart manufacturing environments.

Digital twins have many architectural layers, including both software and physical layers. This includes:

• A main physical layer that includes the sensors that are attached to the physical asset being simulated and produces all the relevant data in real time.
• A layer that is a combination of hardware and software, which gathers, cleans, and stores the data.
• The application layer, which is a software layer that enables the data to be presented to the owner of the physical asset and focuses on small operational details, such as temperature increases in manufacturing environments.
• The interaction layer, which is a software layer that contains AI/ML algorithms and gives deeper insights into the operation of the physical system.

The software layers are also supported by various levels of physical hardware to run the simulations and present the data to users.

Additionally, digital twins contain data network and communication architecture to connect the sensors to the data management platform.

With physical systems being brought more into the digital world, safeguarding data against malicious cyberattacks is more important than ever. One of the challenges of building cyber-physical systems is that there is a lack of standardization of components, which leads to weak cybersecurity in many of these systems. This is a crucial issue because these systems harvest and store a lot of data (including personal data in some applications), so it could lead to serious data breaches if any hackers gain unauthorized access to this data. However, the challenges that exist for cyber-physical systems are different depending on whether they are fixed systems or mobile systems. Organizations will be required to implement DevSecOps practices to ensure that security challenges are handled at all levels.

Fixed Systems

In systems that have fixed infrastructure, such as smart grids, a lot of new and potentially unsecured entry points are being created across the digital network due to the sheer number of sensors and IoT devices being installed. A lot of this is trying to merge old and new technology. Much of it is not standardized, and the cybersecurity protocols are not always robust, leading to vulnerabilities.

In the case of smart grids, the decentralized and interconnected nature of many parties in the network — utility providers, consumers, DSOs — means that a network hack at one small, unsecured node could potentially gain access to the entire network. This could include sensitive customer data due to networks containing billing and smart meter information and smart grids using cloud computing and open-protocol standards.

The smart grid network also contains a lot of information on the operational side of the grid from transmission and distribution, energy generation control, EV, and electricity market data, which means that any attack could potentially bring down the grid. One of the dangers of smart grids not having robust IT protocols is that shutting down operations is much more catastrophic than bringing down a website, as the potential effects are a lot more deadly for the local area — for instance, shutting off power to local hospitals and other emergency services.

Mobile Systems

The situation is a bit different for mobile cyber-physical systems, such as autonomous vehicles. Autonomous vehicles are newer technologies that are being built with more robust IT protocols directly integrated into the designs. Additionally, autonomous vehicles are designed to behave like closed systems, making each entry point of a cyberattack isolated, monitored, and resilient.

However, autonomous vehicles could face challenges in the future when vehicle-to-vehicle communication becomes a reality. At that point, it will be a lot easier to infiltrate the vehicle network and cause issues, so robust protocols will need to be in place as this develops. One of the challenges, again, will be standardization. Protocols are being developed, but they’re either too insecure or not being adopted by OEMs, so there will soon need to be standard protocols in place so software can be continuously patched and updated in line with potential cyberthreats.

Simulation can help develop different cyber-physical systems, the components used in them, and the environments they are used in. Ansys, part of Synopsys, has a number of tools that help simulate the vast spectrum of cyber-physical systems and bring legacy equipment into the modern day through digital engineering. These tools include but are not limited to:

Ansys Fluent fluid simulation software and Ansys Mechanical structural finite element analysis software: Uses the asset layer of cyber-physical systems and plays a big role in designing the asset itself.

Ansys Speos CAD integrated optical and lighting simulation software and Ansys Zemax OpticStudio optical system design and analysis software: Designs the optical components — lenses, etc. — used in the cameras and charge couple device (CCD) sensors of autonomous vehicles. OpticStudio software is used for developing the lenses while Speos software looks at how the overall sensor behaves. Ansys Lumerical FDTD advanced 3D electromagnetic FDTD simulation software can also be used to investigate small-scale optical effects.

Ansys AVXcelerate software: Simulates the sensor environment in autonomous vehicles and can be used for robotics in general. Primarily used for vehicle testing, it can cover camera, lidar, radar, and inertial measurement units (IMUs). AVXcelerate software can also use data from Speos software to look at how different sensors interplay and work together in autonomous vehicles.

The Ansys Twin Builder simulation-based digital twin platform and Ansys TwinAI AI-powered digital twin software: Creates digital twins and enables physical assets to be managed and monitored in real time inside the virtual environment. They can be used to design new physical assets as well.

Ansys Medini Cybersecurity SE system-oriented cybersecurity analysis software: A model-based, integrated security and cybersecurity solution used for the development, certification, and maintenance of cyber-resilient products.

If you’d like to discover how you can develop and monitor cyber-physical systems through advanced simulation, contact our technical team today.

Related Resources