Physical AI using Swarms of Flying Light Specks
Shahram Ghandeharizadeh - University of Southern California
Aug. 6, 2026, 10 a.m. - Aug. 6, 2026, 11 a.m.
MC 437
Hosted by: Paul Kry
The next frontier of human-computer and human-human communication and interaction is moving beyond the constraints of two-dimensional screens and head-mounted displays. We present a physical manifestation of spatial computing driven by Flying Light Specks, FLSs. It is an instance of Physical AI that materializes graphics and animations visible to the naked eye and supports bare-hand interactions. An FLS is a voxel-like entity that possesses mass, emits light, and generates thrust, enabling dynamic three-dimensional structures with behavior. This presentation outlines the core architectural challenges of swarms, including decentralized localization, spatial coordination, and the emerging need for social physics in human-swarm interaction, laying the groundwork for immersive holodeck-scale systems. A safe holodeck technology promises to transform fundamental human experiences, reshaping how we work, socialize, learn, manage our health, engage in physical activities, and perceive both physical and virtual worlds around us.
Shahram Ghandeharizadeh has been on the faculty of the USC Computer Science Department since 1990. In the mid 1990s when VHS and BETA tapes were dominant, his research group designed and implemented Mitra, the first streaming software system with the ability to scale to thousands of nodes (licensed by Panasonic). He has contributed to several pioneering systems, earning him the prestigious ACM Software System Award in 2009. Flying Light Specks, FLSs, are the latest addition to his body of work, supported by grants from the US National Science Foundation. This work formed the basis of a Frontiers Workshop at SIGGRAPH 2026. In addition, a poster describing visual effects using a special class of FLSs, known as LightBenders, received second prize at the SIGGRAPH 2026 Student Research Competition (SRC). A comprehensive paper on LightBenders has also been accepted to the main track of ACM Multimedia 2026.
