We plan to integrate and combine deep learning methods into/with conventional physical models in order to improve nonlinear modeling for electronic devices like e.g. communications and radio frequency (RF) transceivers. In particular, we will enhance or replace conventional signal processing models like e.g. Volterra series, Wiener-models, Hammerstein models or memory polynomials with new techniques of machine learning. Involved research fields are deep learning, deep unfolding, model –based neural networks, few shot learning, meta learning (transfer models from one device to another), sequence analysis using LSTMs and transformers/BERT. Reinforcement learning methods should be considered for improving models that have delayed signals, therefore also delayed error or reward signals. The goal is to enhance existing physical models by new deep learning techniques to make model simulation faster without losing precision and modeling capacity or to improve existing models using new machine learning approaches. Also online and tracking models should be investigated, that is, models that adapt to the current environment like temperature, humidity, power consumption, voltage deviations, fast changing system parameters etc. These online methods might be combined or substituted by few shot and meta-learning methods.
Furthermore, we also plan to enhance and/or substitute model-based signal processing blocks like e.g. pre-coding, digital pre-distortion, nonlinear distortion mitigation, or adaptive channel estimation/equalization, by machine learning approaches.
Technical skills: Master’s degree (for PhD position) in computer science, mechatronics, electrical engineering, mathematics, physics, or similar subject; knowledge in one or more of the following application domains: signal processing, machine learning
Social skills: willingness and ability to work in a team, solution-orientation, persistent researcher, well-structured and goal-oriented working style, reliability and excellent interpersonal skills, hands-on mentality
Personal skills: English level C1
Due to the ever increasing complexity of circuits and systems, the design of System on Chips starts at high levels of abstraction. Model-based design is a state-of-the-art design methodology which addresses this development. Here, initial drafts of the design are defined using modeling languages such as UML, SysML, etc. However, already at this stage severe design choices are made which will have a substantial impact, e.g., at the eventual cost of the design. But properly estimating those effects is hard to impossible using conventional methods, which is why those problems are thus far tackled, e.g., by “trial and error”, “gut feeling”, and “experience of the design engineer”. This frequently yields situations where, after months of implementations, design choices turn out to not satisfying the desired objectives—a severe threat for time to market which is crucial in the EDA industry.
Machine Learning research in the fields of deep learning, few shot learning, meta learning, reinforcement learning, sequence analysis methods like LSTM or Transformer/BERT should be performed and new methods to provide improved cost estimations should be developed. Onto these estimates, design choices can be based to obtain optimized SoCs, and eventually to automate further steps of the design process. Possible industry partners (providing correspondingly needed data and use cases) are available. Results generated by this project will directly be applied in industrial practice.
Research will be focused on building blocks for mm-wave applications like communications and radar. The candidate will innovate in either receive, transmit, or frequency synthesis in coordination with other researcher in the SAL mm-wave laboratory.
Focus a: MIMO Transceiver for Integrated Sub-THz Sensor Systems: Building Blocks and Integrated Circuits for MIMO Transceivers above 100 GHz.
Focus b: Next Generation/Alternative mmW-Sensing and -Imaging Concepts: Alternative and situation aware modulation schemes as well as imaging concepts, realizations, and algorithms.
Research will be focused on building blocks for sub-THz-sensing applications at 120, 230, 330 and 480 GHz, respectively, focused on RX-TX separation at highest frequencies as well as array optimization under the given constraints. The candidate will innovate in either core functional blocks, the full sensor system, array concepts together with researchers in the SAL mm-wave laboratory.
Research in the area of automata learning and verification in order to further automate the testing process of computer-based systems. Novel combinations of model learning, model-based test-case generation and runtime verification in order to test selected dependability properties of a given system-under-test. The research shall go beyond the logical correctness of a system’s functionality and also consider other quality characteristics, like latency and energy consumption.
Development of novel and theoretically sound machine learning techniques based on variational techniques, with a focus on robustness and implementability on embedded/low power devices.
With the advancement in CMOS technologies, multiple electronic systems are getting integrated on a single chip. The communication between these systems demands a dramatic increase in bandwidth over on-chip and chip-to-chip transmission lines. Energy and area efficient high-speed data transmission is becoming a challenge in multi-processor systems-on-chip (MPSoCs) and multi-input-multi-output systems (MIMOs).
Simultaneous bidirectional (SBD) signaling, where data streams are transmitted and received on both sides of a given channel, is an attractive approach offering higher pin efficiency at a given data rate. The available Ph.D. position is related to the development of SBD transceivers for MIMO applications. This research work focuses on an energy and area efficient SBD transceiver with adaptive echo/crosstalk cancellation for the support of a wide range of channels (on-chip and chip-to-chip transmission lines). The aim is to achieve a behavioral model of all-important effects and novel circuit architectures with adaptive solutions for attenuation of echo and crosstalk.
Research in the area of signal processing and physical layer aspects for industrial wireless sensor networks with the goal to characterize and rate the reliability and dependability of wireless links within a network. Additionally, centralized and cooperative localization methods, based on information extracted from the wireless communication between nodes should be applied for enhancing security, safety and reliability of the network
Factory communication is facing the leap of the new generation wired and wireless technologies e.g. TSN, WiFi6, 5G, and more, expanding from automation to the IIoT. In this context, this project focuses on theoretic work, simulation and empirical research to use for factory automation 5G and WIFI6 technologies. This will lead to the investigation on requirements and challenges for communication of sensors and actuators used in industrial environments and design of protocols and methodologies for critical control in factory of the future.