Munich Embedded 2025

Abstract

Reproducible builds are becoming essential for firmware security as supply chain attacks target embedded systems. This session demonstrates practical reproducible build implementations in the Yocto Project and Zephyr RTOS, showing how to achieve bit-for-bit identical firmware builds in production environments. The Yocto Project has achieved 100% reproducibility for OpenEmbedded-Core using SOURCE_DATE_EPOCH and deterministic build processes. Zephyr RTOS provides reproducible builds for MCU firmware, addressing unique challenges in resource-constrained environments. Both projects prove that deterministic firmware builds are achievable at scale.

We will look at CI/CD integration patterns, and how they might complement or contradict reproducibility. The presentation demonstrates some common errors that break reproducibility, including timestamp variations, non-deterministic file ordering, build path dependencies, and environment-specific configurations, providing practical solutions for each challenge.

Abstract

The fragmented bootloader ecosystem leads to fragmented hardware and OS support. Instead of improving bootloaders, features are hacked into the OS (e.g. UEFI compression) and user space (e.g. systemd-boot). Community-based open source bootloaders exist but lack widespread adoption. x86 is dominated by broken standards, while aarch64 requires specialized knowledge due to numerous chip manufacturers. No broad community project offers complete binary boot chains from ROM to OS.

This talk explores what such a community might look like, based on existing projects including mainstream Linux distribution tools and systemd, rather than specialized embedded tools. Moore's Law means embedded and consumer devices are converging. Embedded chip manufacturers face more competition and use older technology, making them less secretive. Combined with poor manufacturer software experience, this creates opportunity for new community development.

Abstract

Rust is starting to gain more and more traction in safety-critical domains. Being a system language that offers memory safety and a well-designed type system makes Rust a feasible and highly valuable language for software projects that must fulfill the most stringent safety requirements. However, a great language alone is not enough for projects in such domains. The whole development process and tooling must adhere to strict standards, which are especially challenging to achieve for open-source software. High costs for standards and existing tooling leads to a high entry barrier that is hard to overcome as individual or small company.

This talk presents a template project for safety-critical Rust projects, to demonstrate the current state of (open-source) Rust tooling needed for safety-critical certifications. The template itself builds on permissively licensed open-source Rust tooling that may be used for activities such as compiling, requirements traceability, testing, and code coverage. Additionally, the template should help in the prioritization of tool development, by listing missing tool certifications and missing tooling for certain activities.

Abstract

Testing automated construction robots presents unique challenges due to the unstructured, dynamic nature of construction sites. This talk highlights our multi-level testing approach, combining simulation-based testing (SIL, MIL), hardware-in-the-loop (HIL) integration, and real-world testing on construction sites.

Our process begins with extensive simulation-based testing to explore system behavior under various conditions before deploying hardware. HIL testing then incorporates real components within the simulation loop, enabling early validation of hardware-software integration.
Our approach incorporates continuous feedback loops: insights from physical and HIL testing refine simulation models and testing strategies. This iterative process enables efficient, risk-aware development, gradually increasing system confidence while maintaining realistic conditions.

Our presentation shares practical insights into building testing frameworks for novel robotic systems from scratch and demonstrates how established techniques can be adapted to fundamentally different application domains.