BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Memento EPFL//
BEGIN:VEVENT
SUMMARY:Trust\, Sensing\, and Learning for Provable Multi-Robot Performanc
 e
DTSTART:20261105T110000
DTEND:20261105T120000
DTSTAMP:20260924T081055Z
UID:874f677fc78f4cb50c2452a8f6dfb9586ae40e14224a719b5754f9f0
CATEGORIES:Conferences - Seminars
DESCRIPTION:Stephanie Gil\nMulti-robot systems are physically embodied net
 works — they sense\, move\, and communicate through the physical world. 
 The bar for safe decision-making rises as these systems enter safety-criti
 cal\, real-world settings where they must perform well under uncertainty. 
 Our work shows that physicality is a resource against the two kinds of unc
 ertainty they face: intentional (or adversarial)\, where data is manipulat
 ed by malicious agents\, and natural\, where aspects of the environment ar
 e simply unknown. Most of this talk concerns intentional uncertainty. Here
 \, one way to exploit physicality is by using communication as a sensor. B
 ecause the signals robots exchange are difficult to forge\, they carry evi
 dence that can be cross-validated to yield a quantifiable likelihood that 
 an agent's data is trustworthy. This is the foundation of cy-trust\, in wh
 ich stochastic observations of trust model an agent's trustworthiness prob
 abilistically from physical rather than cryptographic evidence. Each neigh
 bor's contribution is then weighted by its trust value. Under this framewo
 rk\, we show that consensus\, distributed optimization\, and other core co
 ordination tasks admit almost-sure convergence with bounded deviation from
  their nominal performance\, even when malicious agents exceed half of a n
 ode's connectivity\, past the classical Byzantine bound. We support this f
 inding with both theory and hardware experiments under adversarial attack.
  Against natural uncertainty\, we show that real-time sensing can be folde
 d into rollout-based reinforcement learning\, where the same machinery rew
 eights futures rather than neighbors. We apply this idea to routing a flee
 t of robots to stochastically appearing demand and\, with Project CETI\, t
 o the first autonomous robotic rendezvous with sperm whales at sea. Finall
 y\, we preview some of our future work combining trust with long-horizon s
 equential decision-making\, targeting planning that stays provably resilie
 nt when the data informing the plan may itself be corrupted.\n\nStephanie 
 Gil is the John L. Loeb Associate Professor of Engineering and Applied Sci
 ences at Harvard University and an Associate Faculty member of the Kempner
  Institute. Her research focuses on trust and coordination in multi-robot 
 systems\, at the intersection of robotics\, communication\, and learning. 
 Her contributions have been recognized through the DARPA Young Faculty Awa
 rd (2024)\, the Office of Naval Research Young Investigator Award (2021)\,
  and the National Science Foundation CAREER Award (2019). She was named a 
 2020 Sloan Research Fellow for her work at the intersection of robotics an
 d communication. She earned her Ph.D. at CSAIL at MIT\, specializing in mu
 lti-robot coordination and control\, and her B.S. at Cornell University.
LOCATION:MED 2 1124 https://plan.epfl.ch/?room==MED%202%201124 https://epf
 l.zoom.us/j/64474989189
STATUS:CONFIRMED
END:VEVENT
END:VCALENDAR