Research Projects

2021 - now Research collaboration with the Division of Computer Engineering (University of Aizu)

• Evolutionary algorithms for neuromorphic systems

2021 - now Research collaboration with the Chair of Integrated Systems (TU Munich)

• Hardware Learning Classifier Tables (reinforcement learning) for self-adaptive mixed-critical applications (in collaboration with UC Irvine and TU Braunschweig)

• Methodological developments of evolutionary algorithms

2021 - now Research mentor at TU Munich

• • Exploring hardware-acceleration of operating system functionalities, such as inter-process communication (MSc. Twardzik, Chair of Integrated Systems)

  • Game theoretic framework for safe autonomous systems (MSc. Grobelna, Chair for Theoretical Computer Science)

2016 - now Scientific collaborator of the CoDE-SMG department (ULB-EPB)

• Development of a decision aiding methodology for finding consensus among multiple decision-makers and application with industrial partners (case studies from cities of Oslo and Brussels, in collaboration with the Vrije Universiteit Brussel)

2021 - 2023 Safety of AI in autonomous systems (Fraunhofer IKS)

• Dynamic risk management of autonomous systems using game theory

• Adversarial attacks to uncover unsafe behavior of AI

• Uncertainty estimation for safe autonomous systems (ICON LOPAAS project in collaboration with Fraunhofer IESE and the University of York)

• Safety case for autonomous lane merging/changing manoeuvre using neural networks and reinforcement learning

• Safety case for automated guided vehicles operating in shop floors

2018 - 2021 Post-doctoral Research Projects at TU Munich

AI-based methodologies for embedded (mixed-critical) systems

• • • • • • Fine-grained power modelling and prediction for multicore processors (DFG TCRC 89 “Invasive Computing”)

          • Genetic algorithm-based rule evolution of Hardware Learning Classifier Tables (reinforcement learning) for self-adaptive mixed-critical applications (in collaboration with UC Irvine and TU Braunschweig)

          • Design space exploration and decision-making for task mapping of a fault-tolerant hybrid TDM/packet-switched NoC with protection switching

Optimization for approximate computing applications

• Multi-objective optimization and decision-making for approximate computing parameters of FPGA-based image processing in cameras (in collaboration with Arri and Smartray)

• Design space exploration for parameters of compressed CNNs on embedded automotive multicore platforms (in collaboration with BMW)

Decision-making processes for computer resource management

• Cache memory replacement policies using voting-based mechanisms (DFG TCRC 89 “Invasive Computing”)

Student and project supervision

• Master thesis supervision (MSc. Çetinkaya) : “Machine Learning for Improved Commonality Optimization in Passive Crash Safety” (in collaboration with BMW)

• Master thesis supervision (MSc. Shah): “Enhancing the Representability of Driving Profile Synthesis using Stochastic Optimization” (in collaboration with BMW)