AI-Powered Robot Cognition Modeling for Trustworthy Humans-Robots Collaboration
Abstract: Human-supervisory robots have been widely deployed in mission scenarios, such as disaster response, agriculture, and manufacturing. However, this symbiotic partnership is fragile and often unstable. These robots frequently fail in task executions and human collaborations due to real-world uncertainties and dynamics. This discrepancy arises because the real world inevitably differs from theoretical expectations, presenting challenges to planned missions, such as interrupting executions, ignoring priorities, and disordering workload allocations. These uncertainties broadly exist in human intents, obstacle distributions, knowledge levels, etc. Humans excel in collaboration by forming a trustworthy collaborative partnership, characterized by comprehensive goal understanding and high error tolerance, thereby encouraging a flexible and stable collective intelligence. Inspired by this human approach, to enhance the trustworthiness of human-robot teams, it is critical to rethink the current design of robot motion planning and delve deeper into the general “cognition processes” behind these motions. This research aims to mimic the human cognitive thinking process to make robots more resilient to failures; it innovates in both the partnership paradigm and trustworthiness of robot-human collaborations, moving towards generally applicable robot systems in the real world.
In my talk, I will discuss "AI-Powered Robot Cognition Modeling," specifically focusing on trust, intent, and agility models, for designing a generally trustworthy human-robot collaborative system. Based on psychology theory, human cognitive behavior studies, and cutting-edge AI algorithms and hardware systems, cognition is modeled to develop a preliminary "self-awareness" in robots. This enables them to reflect on their behaviors from a human perspective, analyze collaboration ineffectiveness, and then build a reliable and high-performance mixed human-robot team. First, I will introduce a trust-aware learning control model for multi-robot systems to self-repair their faulty behaviors by estimating human expectations in scenarios like "human-multi-robot cooperated disaster rescue." Second, I will present a meta-learning model that aids robotic systems in quickly understanding human preferences and adapting swiftly. Third, I will discuss an agility-aware power supply model to demonstrate how cognition models can accelerate physical robot systems. Lastly, I will briefly introduce my latest theoretical work on modeling cognitions using quantum artificial intelligence and large language models.
BIO. Dr. Rui Liu is an Assistant Professor in the College of Aeronautics and Engineering (CAE), Kent State University, Ohio, since 2019 Fall. He directed the Cognitive Robotics and AI lab (CRAI, ruiliurobotics.weebly.com), with a focus on cognitive robotics research -– designing the "Mind" for robotics and AI systems for seamless cooperation with a human. From 2018 Fall to 2019 Spring, Rui received his postdoc training at Carnegie Mellon University, Pittsburgh PA, the Robotics Institute at the School of Computer Science, advised by Prof. Katia Sycara. In 2018 Spring, he got his Ph.D. degree from the Colorado School of Mines, Golden CO, ME, advised by Prof. Xiaoli Zhang. Before that, He graduated from Shanghai Jiao Tong University in China and worked as a research engineer in the Chinese Academy of Science.
He has finished more than 50 peer-reviewed publications on human-Robot/Multi-Robot teaming. He currently serves as the Associate Editor in the prestigious journal IEEE Robotics and Automation Letters (R-AL) and conferences IEEE ICRA, IROS, and Ro-Man, a committee member for multiple leading robotics/AI conferences AAAI, AAMAS, MRS; he is also a reviewer for some high-impact journals/conferences, including leading AI/robotics conferences and Journals such as IEEE Transactions on Robotics. In the leading robotics conference IEEE ICRA 2015 and IEEE RO-MAN 2019, his works have been nominated as the best paper award. Dr. Liu’s research is funded by various federal and state agencies such as NSF, AF, ARL, NASA, etc.
Zoom Link: https://wpi.zoom.us/s/92946413217