Mindful AI for Adaptive, Resilient CyberPhysical Human Systems (CPHS)
CyberPhysical Human Systems (CPHS) – ranging from small form-factor IoT devices to complex system-of-systems with humans-in-the-loop – manage complexity through hierarchically layered design abstractions. While this clean separation of abstraction levels eases the tasks of modeling, design, validation and verification, CPHS demand a tight coupling of computation, communication and control across the abstraction stack to meet energy, performance, reliability and security needs. Furthermore, the fast-evolving landscape of emerging computing substrates, coupled with highly dynamic behaviors operating in varying environmental conditions pose significant challenges to meet the (often conflicting) goals of resiliency, energy, heat, cost, performance, security, etc. AI is increasingly being used in an ad-hoc manner across the abstraction stack, falling short in adaptive scenarios, and also unable to support system evolution over time. I posit that CPHS need to incorporate mindful AI strategies to enable continual runtime learning and evolution, allowing the system to adapt dynamically and maintain overall resilience. A key feature of the mindful AI paradigm is computational self-awareness through introspection (i.e., modeling and observing its own internal and external behaviors) combined with both reflexive and reflective adaptations via cross-layer physical and virtual sensing and actuations applied across multiple layers of the system abstraction stack. This requires a fundamental change from current AI-infused layered computing to a cross-layer mindful AI paradigm that embodies self-awareness principles. In the past decade we have applied these principles across multiple CPHS projects spanning nanoscale computing, healthcare IoT, data center memory, and end-to-end autonomous system computational pipelines. The rise of generative and agentic AI, coupled with distributed autonomy poses both new challenges as well as opportunities for the ESWEEK research community. I will close with some thoughts on how mindful AI principles and cross-layer self-awareness might be applied to address these challenges.
Biography
Nikil Dutt is a Distinguished Professor (CS, Cognitive Sciences, and EECS) and the Associate Dean for Research (Donald Bren School of ICS) at the University of California, Irvine (UCI). He received a PhD from the University of Illinois at Urbana-Champaign (1989). His research interests are in embedded and cyber-physical systems, EDA, computer architecture & compilers, distributed systems, healthcare IoT, and braininspired architectures and computing. His research group has done foundational research in Architecture Description Languages (ADLs) for customized processors, software controlled memories, and self-aware, cross-layer design of adaptive, resilient computing systems. He has received over 20 best paper awards and nominations at premier EDA and embedded systems conferences, as well as multiple departmental and campus awards for excellence in teaching and mentoring at UCI. Professor Dutt has served as EiC of ACM TODAES, and AE for ACM TECS and IEEE TVLSI. He has served on the steering, organizing, and program committees of several premier EDA and Embedded System Design conferences, and has also served on several ACM (Publications Board, SIGBED, SIGDA, TECS) and IEEE (ESL) advisory boards. He is an ACM Fellow, IEEE Fellow, and recipient of the IFIP Silver Core Award.
Affiliation
Departments of Computer Science, Cognitive Sciences, and EECS
University of California, Irvine
Irvine, CA 92697-3435, USA
