BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//NextToU//EN
CALSCALE:GREGORIAN
BEGIN:VEVENT
UID:1c204626-dc31-4b9d-b712-c45d7aaddadc@nexttoyou
DTSTAMP:20261106T174501Z
DTSTART:20261111T213000Z
DTEND:20261111T223000Z
SUMMARY:MECH Seminar Series Presents: Dr. Bolei Deng
LOCATION:Maxfield Hall\, Maxfield Hall
DESCRIPTION:Toward Electronics-Free Systems through Mechanical Intelligenc
 e\n\nAbstract:\n\nIntelligence in engineered systems is often associated 
 with electronic sensors\, processors\, and controllers. In this talk\, I 
 will show how the physical structure of a system can itself play an activ
 e role in sensing\, communication\, and control. I will discuss how mecha
 nical design can shift part of this functionality from electronics into t
 he body\, creating new opportunities for simpler and more adaptable syste
 ms.\n\nFirst\, I will show how linear mechanical waves can control comple
 x behavior in structures and robots without distributed electronic contro
 llers. By tuning the frequency of vibrational signals\, we selectively ad
 dress mechanical elements and coordinate their responses. I will then dis
 cuss how this approach extends to modular robots\, allowing physically co
 nnected modules to communicate through vibration and enabling reconfigura
 ble architectures.\n\nSecond\, I will discuss how structural design can e
 nhance a system’s sensing capabilities. Our approach is performance-drive
 n brain–body co-design: gradients of the neural decoder’s prediction erro
 r are propagated through a differentiable mechanical simulator to optimiz
 e the body’s geometry alongside the decoder. This co-training reduces mea
 n force-prediction error by factors of 4.2 in simulation and 1.7 in exper
 iments\, with an optimized single-sensor body outperforming the original 
 body equipped with two sensors.\n\nTogether\, these examples highlight a 
 \n\nhttp://next-to-u.app/events/1c204626-dc31-4b9d-b712-c45d7aaddadc
URL:http://next-to-u.app/events/1c204626-dc31-4b9d-b712-c45d7aaddadc
END:VEVENT
END:VCALENDAR