" Towards super-flexible software that self-heals, self-optimizes, and self-protects "
Software systems are increasingly intertwined with physical environments and social infrastructure. Autonomous software that can flexibly address various changes by self-healing, self-optimizing, and self-protecting is a key enabler for the next generation of systems. In our laboratory, we develop techniques for building such “super-flexible software” across domains including cloud-native systems, robotics, federated learning, and human-machine interaction.
Research topics
We conduct research on autonomous software systems that can flexibly adapt and evolve in response to runtime changes. Our research spans four main areas: self-adaptive systems, formal controller synthesis, generative AI for software engineering, and human-autonomous system interaction.
Self-Adaptive Systems
Modern software systems closely interact with external entities (e.g., users, external services, and physical entities), whose changes occur at runtime and are impossible to be completely predicted at development time. Self-adaptive systems address this by (1) monitoring environmental changes, (2) analyzing and planning how to modify the system's structure or behavior, and (3) executing those modifications at runtime. Based on the MAPE-K loop and models@run.time techniques, we also apply self-adaptation to federated learning security and user preference adaptation. -> Details
Discrete Controller Synthesis
Discrete Controller Synthesis (DCS) automatically generates a mathematically guaranteed controller from a Labeled Transition System (LTS) environment model and given safety/liveness properties. We pursue algorithm efficiency improvements (on-the-fly synthesis, differential synthesis, LTS minimization) and AI-guided synthesis combining reinforcement learning and LLMs, with applications to robot control, cloud-native systems, and Systems-of-Systems. -> Details
LLM & Generative AI for Software Engineering
We integrate large language models (LLMs) and generative AI into software engineering across multiple research areas. Key topics include generative AI for self-adaptive systems (ACM TAAS 2024), LLM-guided discrete controller synthesis, LLM-enhanced evolutionary computation and AutoML, and LLM-based security applications. -> Details
Human-Autonomous System Interaction
We study the perception, trust, and interaction design between humans and autonomous systems such as self-driving vehicles, drones, and robots. Research topics include eHMI (External Human-Machine Interface) design and evaluation, take-over request (TOR) modalities in semi-autonomous driving, and collaborative communication between humans and robots. -> Details
International collaboration
We have active collaborations with world-leading research groups in overseas universities and institutes. -> Details