Wireless Low Power Edge Devices and Transient Computing Systems with Energy Harvesting

Since the prediction of Gordon Moore in 1965, known as Moore’s Law [1], the in- tegration density and performance of integrated circuits have increased continuously. As a result, integrated circuits are enabling ever-new possibilities for decades and are being used in more and more areas of applications. This led to a fast growing demand of micro- electronic systems ranging from, e.g., edge devices for the Internet of Things (IoT) over industrial control systems up to computationally intensive machine learning applications. However, within the last years, the increased environmental awareness and the efforts to create a more sustainable world have shed light on the darker side of this development: The increasing energy demand for computing. In fact, the global demand for computa- tional energy and battery capacity is increasing exponentially. In order to counteract this development, experts call for fundamental changes in microelectronic systems design that enable significantly better energy efficiencies of electronic devices [2]. Suggestions range from an increased use of advanced computing platforms that allow for application-specific hardware-software co-designs up to an improvement of edge computing capabilities to gradually replace conventional computing. More radical proposals request for a consistent utilization of renewable energy sources [3] or propose specific paradigm changes in com- monly battery-powered application fields by calling for a battery-less internet of things [4]. This thesis proposes new design approaches for energy and environmentally friendly embedded systems in different areas of application that are fit for the future. For the in- dustrial area, this thesis addresses model predictive control (MPC) systems. A new op- timization strategy for the resource-intensive MPC algorithms is demonstrated to enable their implementation on low power and low-cost FPGAs. Regarding edge computing, new design approaches are presented to improve the computing capabilities for mobile battery- powered devices. These approaches enable ultra-low power communication security fea- tures for IoT systems and present new precise time synchronization principles for wireless sensor networks. A clear emphasis of this thesis is on the improvement of embedded systems that are solely powered by energy harvesting. Such battery-less systems are also known as inter- mittent or transient computing systems, as they do not have a reliable energy basis. One of the major issues that hamper the widespread use of this sustainable system concept is the reactive nature of state-of-the-art transient computing approaches, i.e., the system is com- pletely and utterly dependent on the transiency of its energy harvesting source. As a solu- tion, this thesis proposes new hardware and software design methods towards a proactive transient computing. These methods are dedicated to kinetic and thermal energy harvesting technologies and enable a predictive and adaptive power loss management to pave the road for yet infeasible applications fields.