Embedded software for low-power connected devices

Boost efficiency, reliability, and performance with scalable RTOS solutions.

Wearable device

Developing embedded software for ultra-low-power embedded devices is a complex task as it involves optimizing energy consumption, efficiently utilizing limited resources, balancing real-time performance, and managing software complexity.

Challenges are plentiful, including low memory footprint, sensor integration, battery management, and secure low-power operations.

Ultra-low-power real-time operating systems (RTOS), such as CSEM μ169 or Zephyr, are key enablers of low-power, secure, and future-proof embedded software.

Real-time operating systems and low-level software

Various approaches exist for developing low-power software for embedded devices. Bare Metal Software runs directly on the hardware, bypassing intermediary layers. Although it lacks an operating system or scheduler, it suits simple, resource-constrained devices. However, its response times to external events can be unpredictable. To enhance multitasking capabilities, access peripherals, and maintain precise timing, a software scheduler allocates CPU time efficiently. In contrast, an RTOS sits atop the hardware, providing scheduling, resource management, and inter-task communication. RTOS enables complex, real-time applications with predictable execution and meets critical task deadlines.

History of RTOS at CSEM

The journey began in 1992 with µKOS, originally created for the Khepera mobile robot. Remarkably, µKOS was already operating these robots in advanced space applications as early as 1997 at JPL. Since then, it has been employed in various miniature and low-power systems at CSEM. Specifically, µ169 represents the latest version of the fully customizable µKOS microkernel OS, designed to provide a multitasking environment.

Over the years, µ169 has been ported to numerous microcontrollers, and it now runs on the latest RISC-V and ARM platforms. The primary goal of μ169 is to offer tools and platforms for rapidly developing real-time applications that can adapt seamlessly to hardware evolution. Moreover, the OS is available for a wide range of microcontrollers, facilitating straightforward application porting. Researchers and developers can obtain a customized distribution of the OS through a facilitated license upon request.

The μ169 package comprises a collection of expandable software modules and tools. At its core lies the real-time micro kernel, which is highly portable, scalable, preemptive, and configurable. It is designed for microprocessors, microcontrollers, and DSPs.

What sets the μ169 package apart is its unprecedented ease of use. It comes complete with 100% C99 source code and comprehensive documentation. Moreover, it runs on a wide range of processor architectures, and you can download port examples upon request.

One notable feature of the μ169 package is its ability to adjust the system’s memory footprint based on your design requirements, thus conserving valuable memory space. Additionally, the Nano version is specifically tailored to minimize memory usage, making it suitable for embedding in ASICs.

T-Touch Connect Solar and T-Touch Solar Sport run on a CSEM RTOS: connected, secured, autonomous, and long-lasting.

The T-Touch Connect Solar by Tissot. © Tissot

The T-Touch Connect Solar by Tissot.

The T-Touch Solar Sport by Tissot. © Tissot

The T-Touch Solar Sport by Tissot.

Benefits

  • Versatile console-based interface
  • Console tools for on-the-fly debugging, analysis and software download
  • Easy and rapid porting to new CPU / platform
  • Allows building of secured and reusable applications
  • Rapid path between off-the-shelf microcontroller boards and ASICs
  • Optimal usage of the time and memory resources
  • Available sources with a long-term support license
  • Supported cores: Cortex-M0, -M0+, M3, -M4, -M7, -M23, -M33, RISC-V, RV32, and RV64
  • Ready ports: STM32xxx, Ambiq Micro Apollo2, ON Semiconductor RSL10, Canaan Kendryte K210, Nordic Semiconductors nRF5xx

Key specifications

  • Portable: The µ169 μkernel runs on a large num­ber of processor architectures, such as ARM Cortex-Mx, RISC-V RV32 & V64, CSEM icyflex. It can take advantage of the user/machine spaces and of specific secure units such as MPUs
  • Scalable: The µ169 μkernel allows almost unlimited tasks and kernel objects. The kernels’ memory footprint can be scaled down (Nano ver­sion) to contain only the features required for your application, typically ~6–10 KBytes of code space and ~1 KByte of data space
  • Reliable: The µ169 package includes debug­ging features that reduce development time. The μkernel can provide extensive range checking (memory leaks, stack limits, etc.)
  • Efficient: µ169 can include valuable runtime statistics, making the internals of your application observable. Identify performance bottlenecks, and optimize power usage, early in your development cycle
  • Toolchain: The µ169 package is developed using full open-source tools. The supported compiler goes up to the latest gcc, but other proprietary tools could be easily adapted. All the building scripts are available for macOS, Window 10, Ubuntu.

For specific needs, CSEM also offers expertise in developing applications using a customized scheduler or other embedded operating systems such as FreeRTOS and Zyphyr, drawing on their experience in resource optimization and system security.

Comparison of key features and specifications of CSEM µ169 RTOS, Zephyr RTOS, and FreeRTOS.

Comparison of key features and specifications of CSEM µ169 RTOS, Zephyr RTOS, and FreeRTOS

RF protocols and stacks

Over the last 25 years, CSEM has actively engaged in the communication protocol field, creating robust solutions using either proprietary concepts or established standards. A notable example of its expertise is the WiseMac Low Latency and Low Power protocol. WiseMAC is a low-power protocol for WSN.

However, in the domain of ultra-low-power software development, CSEM’s expertise in communication protocols extends beyond customized solutions. Over the years, CSEM has successfully implemented solutions based on various well-known protocols such as LoRaWAN, UWB (Ultra-Wideband), and Bluetooth. This wealth of experience has been informed by CSEM's work in developing its own silicon technology, which in turn has contributed to the advancement of low-power features in other widely used protocols like 802.15.4 and Wirepas. By combining proprietary concepts with established standards, CSEM continues to play a pivotal role in shaping efficient and reliable communication solutions for diverse applications, including aeronautics, water management, home automation, and safety.

Plane flying above the clouds

Mesh wireless networks are used in airplanes to connect distributed sensors.

Self-configuring wireless mesh network for Hinni water management.

Self-configuring wireless mesh network for Hinni water management.

Key figures and benefits

  • Low latency, ultra-low power (several years of operation without changing the battery)
  • Long range with multiple short hops
  • Multi-frequency wireless solutions
  • Peer-to-peer and converge-cast routing
  • All nodes are sensors and routers
  • Multiple destinations (sinks) with multiple routes (depending on node density)
  • Gateways to standard communication systems (Internet/IP)
  • Guaranteed transmission (error management, repetitions, etc.)
  • Secure & robust (routing, frequency hopping, firmware update, etc.)
  • Integrated into CSEM RTOS
  • Dual mode (TDMA/CSMA), standard/optimized operations 

A proven technology

With our autonomous wireless systems for continuous monitoring and control, you can benefit from the following features:

  • Compliant with regulations
  • Real-time, secure
  • Robust, reliable
  • Small footprint, low power
  • Energy management and energy harvesting
  • up to 10 years of operation and more

The RF propagation properties can also be used for precise indoor localization, for which CSEM has developed a range of technologies.

Tiny Machine Learning

Tiny Machine Learning (tinyML) is revolutionizing the world of machine learning by bringing the power of artificial intelligence to the smallest embedded devices. This innovative approach enables complex data processing and decision-making directly on microcontrollers with minimal power consumption. CSEM’s expertise in embedded software combined with tiny machine learning, unlocks new possibilities for real-time analytics and enhanced IoT applications.

Our expertise in tinyML is successfully applied and realized across a diverse set of applications:

Smart weather station Aurora

Aurora is a low-power, intelligent, low-cost, and easy-to-install weather station. By listening to the surrounding environment, Aurora estimates the rain and wind intensity without the need for any mechanical moving part. Aurora collects audio signals with an integrated microphone and processes this data through tinyML inference directly on the device to accomplish this.

Robust gesture recognition

We design and implement sophisticated tinyML algorithms that accurately interpret human gestures, even in challenging environments. These are optimized to ensure gestures can be recognized seamlessly, even on resource-constrained devices. Coupled with cutting-edge hardware solutions based on time of flight (ToF) and mmWave radar, our gesture recognition systems offer unparalleled precision and reliability. This combination ensures seamless and intuitive user interactions across a wide range of applications.

CSEM’s smart acoustic weather station Aurora

CSEM’s smart acoustic weather station Aurora, deployed next to the lake of Neuchâtel in Switzerland

Maximize performance with CSEM’s embedded software

Harness the power of hand-crafted embedded software built for power efficiency, scalability, and security. Explore seamless integration with cutting-edge hardware and achieve superior embedded system performance. 

Let’s build the future together.

Publications and Research

Journal articles

CSEM scientific and technical reports