Wencan Mao "FLEET-EI: Distributed Edge Intelligence for Autonomous Robotic Fleets"

This talk can be viewed via zoom. (Note: this talk will be recorded)

Abstract: Fleets of drones, ground robots, and connected vehicles are leaving the lab for fields, factories, and disaster sites. Their intelligence is split across robots, edge, and cloud, tied together by a wireless link that is shared, intermittent, and sometimes absent. Edge AI, multi-agent reinforcement learning (MARL) and networking research each optimize one layer and assume the others away, so policies trained under ideal links fail exactly when connectivity does.
This talk argues that sensing, learning, communication, and control should be optimized jointly against one resource budget. It first presents results that support this view: task-aware MARL trajectory planning for factory AGVs (Best Paper Runner-up, IEEE INFOCOM 2026 workshop) and imitation-based triple deep Q-learning for smart-farming UAVs (IEEE ICRA 2026), interference-aware multi-UAV path planning (IEEE TVT), two-timescale network slicing for vehicular edge computing (IEEE TMC), and a language model for semantic-aware network optimization (IEEE Network). It then outlines FLEET-EI, a three-year plan with three contributions: 1) coordination in which communication is an action with a cost, and feedback may not arrive; 2) distributed learning that runs within the mission’s own energy and bandwidth budget; 3) an edge-hosted language-model agent that turns operator intent into feasible missions, checks them with a verifier, and explains what the fleet did. The methods will be validated on real fleets in agriculture and urban or indoor logistics.
 

Bio: Wencan Mao is a Joint Postdoctoral Researcher at the National Institute of Informatics (NII) and the University of Tokyo, Japan, where she leads a UAV research group of four doctoral students and one master’s student. She holds a D.Sc. (Tech.) in Electrical Engineering (2023) and an M.Sc. in Mechanical Engineering (2019) from Aalto University, where she was also a Postdoctoral Researcher in Computer Science (2023–2024). Her research spans multi-agent reinforcement learning, large language models, edge computing, wireless networks, and robotics. She has 26 peer-reviewed publications, including 18 journal articles (e.g., IEEE TMC, TCOM, TVT) and 8 conference papers (e.g., IEEE ICRA), and received a Best Paper Runner-up Award at an IEEE INFOCOM 2026 workshop. She facilitates collaborations between Asian and Nordic universities, works with Aalto, the University of Helsinki, the University of Cambridge, and CUHK, is co-PI on three submitted grant applications (two international, with partners in Germany, Poland, and Spain), and serves on the IEEE GLOBECOM and ICC technical program committees.