The Role of Simulation Technology in Modern Embedded System Design

In the design process of embedded systems, simulation integrates into various engineering methodologies. In the classical waterfall approach and its extension—the V-model—simulation is primarily applied during the verification and integration phases. It is used to ensure that the system design meets both functional and non-functional requirements. This involves unit testing, interface validation, and the simulation of edge cases.

In Model-Based Design (MBD) environments, simulation becomes a core tool. The development process begins with the creation of a functional model (e.g., in Simulink), which then serves as the basis for automatic code generation. This code can be immediately tested using Software-in-the-Loop (SiL), Processor-in-the-Loop (PiL), or Hardware-in-the-Loop (HiL) simulations. They enable continuous validation without the need to build physical prototypes. According to MathWorks case studies, using SiL and PiL simulations can reduce overall development time by up to 35% in safety-critical applications such as automotive or medical devices. Similarly, in Agile methodologies—where development is iterative and organized into short sprints—simulation supports rapid prototyping and testing of new features without interrupting ongoing hardware development.

The key benefits of simulation in embedded systems engineering

Despite the wide range of available techniques and tools, the true value of simulation often remains underestimated—particularly in the early stages of project planning. Why is simulation still overlooked at the point where it can deliver the greatest impact? Gaining a proper understanding of simulation can serve as a starting point for more informed architectural decisions, better resource planning, and more effective collaboration across interdisciplinary teams—especially when guided by best practices in simulation-driven design:

In embedded systems engineering, various types of simulation are applied depending on the project stage, the purpose of the analysis, and the required level of detail needed to assess system correctness. The fundamental classification consists of three main categories:

• Hardware simulation focuses on replicating the physical components of a system, primarily digital elements such as microprocessors, peripheral devices, and programmable logic. During the design phase of integrated circuits or FPGA-based systems, Register Transfer Level (RTL) simulation is used to analyze the behavior of digital logic with clock-cycle precision. This approach allows designers to detect logical errors, timing hazards, or bus conflicts before synthesis. Another widely used method is FPGA prototyping, where the modeled circuit is implemented directly on a physical FPGA device to verify functional behavior using real clock signals and peripherals, thereby closely approximating real-world operating conditions.
• Software simulation involves modeling the execution of application code in a simulated environment. One common technique is CPU instruction emulation, where binary code compiled for a specific architecture (e.g., ARM Cortex-M, RISC-V) is executed in an emulator such as QEMU or Renode. As João Bittencourt of STMicroelectronics aptly put it: “Every minute spent simulating in QEMU saves an hour in the lab.” This enables debugging, resource usage analysis, and functional validation without the need for physical hardware. In parallel, simulation of device drivers and real-time operating systems (RTOS) is performed to evaluate task behavior, interrupt handling, synchronization mechanisms, and memory management.
• Hardware-software co-simulation encompasses hybrid approaches such as SiL, where application code runs within a simulated environment; PiL, which involves execution on a physical CPU while interacting with simulated components; and HiL, which allows for testing of the actual embedded hardware interfacing with a simulated external environment. As Felix Möller, lead test engineer from Continental AG noted, “We use HiL to crash a virtual car into a wall 1000 times a day, so we don’t have to do it even once in the real world”. In more complex scenarios, Co-Simulation is used—integrating multiple simulators (e.g., for hardware and physical environment models) into a single, synchronized simulation framework. Simulation can help in validating integration points early and safely, before real-world systems are available or stable enough for live interaction.

If you’d like to explore the intricacies of HIL, we encourage you to check out our dedicated article:

What is Hardware-in-the-Loop (HIL) Testing And Simulation? A Complete Guide for Engineers

Optimizing simulation strategies: using functional simulation and behavioral simulation

It should be started by stating that physical models describe system phenomena in a way that closely mirrors reality—they account for parameters such as voltage-current characteristics of electronic components, thermal effects, mechanical properties, or electromagnetic interference. One might then ask: where are they most commonly used? Due to their complexity and computational intensity, they are typically used in low-level simulations, such as analog circuit analysis (e.g., SPICE) or modeling of hardware IP blocks.

Functional models focus on representing the logical behavior of the system without tracking its internal physical states. For instance, a processor model might implement an instruction set and track variable values, without modeling actual execution time or signal delays. These models are widely used for algorithm verification, protocol simulation, and testing of high-level software.

Behavioral models occupy a middle ground between functional and physical models—their aim is to reflect the system’s behavior while considering key timing, logical, or state-based parameters. An example would be a driver model that reacts to input events based on predefined logic, with response times either statically defined or dynamically computed.

An essential aspect of model selection is the level of abstraction. Abstract models omit timing and physical details in favor of simplified logic, while cycle-accurate or time-accurate models simulate instruction execution time, signal propagation delays, and bus latency. Choosing between these approaches involves a trade-off between simulation realism and computational performance. Time-accurate models offer higher fidelity but are often impractical in large systems due to long simulation runtimes. On the other hand, abstract models allow faster iterations and early-stage design validation, at the cost of lower prediction accuracy.

In real-world projects, a multi-level modeling approach is often employed, where different subsystems are modeled at different levels of detail—for example, implementing control logic as a behavioral model while modeling hardware interfaces with cycle-accurate fidelity. This strategy enables effective complexity management and allows engineers to tailor simulation scope and resolution to the specific goals of the design phase.

How to choose a simulation tool? Commercial and open-source solutions in practice