NSF CAREER Award for Networked 4D Spatial Intelligence Architecture
ECE Assistant Professor Mallesham Dasari was awarded a $623K NSF CAREER award for “A Networked 4D Architecture and Systems Foundation for Spatial Intelligence,” which he will use to advance the systems and networking foundations of spatial intelligence, an emerging computing paradigm that enables machines to perceive, understand, and interact with the physical world.
Spatial intelligence technologies, including smart glasses, augmented reality systems, digital twins, and intelligent robotics, are rapidly moving from research laboratories into everyday life. These systems promise to transform industries ranging from healthcare and education to manufacturing and logistics by overlaying digital information onto physical environments and enabling new forms of human-computer interaction.
However, today’s Internet was built primarily for images and video rather than dynamic 3D environments. As a result, current networks struggle to efficiently transmit the rich spatial information required by future immersive applications.
Dasari’s CAREER project seeks to address this challenge by developing the fundamental infrastructure needed to represent, compress, transmit, and process dynamic 3D environments in real time. The research explores how resource-constrained devices such as smart glasses can efficiently communicate and understand the physical world while operating over existing wireless networks.
“Just as the Internet evolved from carrying documents to carrying video, the next major evolution will be carrying real-time representations of the physical world,” said Dasari. “Our goal is to develop the systems and networking foundations that make spatial intelligence practical, scalable, and accessible.”
The project will develop new methods for representing dynamic 3D scenes, adaptive streaming techniques for spatial data, and experimental smart-glasses platforms that allow researchers to study trade-offs among latency, energy consumption, visual quality, and network performance. The resulting open platforms and software tools will support reproducible research and accelerate innovation across academia and industry.
In addition to its research contributions, the project includes a strong educational component. Students will gain hands-on experience with emerging spatial computing technologies through open-source platforms, virtual laboratories, and experiential learning activities that bring networking and distributed systems concepts to life.
Abstract Source: NSF
Spatial intelligence enables computing systems to perceive and interact with the physical world. Emerging spatial intelligence technologies such as smart glasses and extended reality systems are beginning to bring digital information directly into everyday physical spaces over the consumer Internet, helping people perform complex tasks in education, healthcare, manufacturing, and logistics more efficiently. Strategic adoption of these technologies could yield billions of dollars in savings. However, today’s Internet is designed primarily for 2D images and video rather than for rich, dynamic 3D, i.e., 4D spatial information. As a result, current systems struggle to efficiently send 4D spatial data on the Internet, limiting the reliability and scalability of these technologies. This research will enable smart glasses to understand and communicate 4D environments in real time, making immersive technologies more practical, accessible, and reliable. The project also expands educational opportunities by introducing students to spatial computing through hands-on experiments, open-source tools, and outreach programs that engage learners from high school through graduate school. The project contributes to workforce development and strengthens national leadership in emerging computing technologies by building accessible platforms and training the next generation of engineers and researchers.
This award develops a systems and networking foundation for spatial intelligence by advancing methods for representing, transmitting, and programming 4D spatial data in resource-constrained devices such as smart glasses. The research introduces a hybrid spatial representation that combines structured-mesh geometry with Gaussian-splat-based appearance models to enable real-time spatial perception, rendering, and analytics such as localization and tracking. Building on this representation, the project develops compression and streaming techniques that exploit temporal redundancy in dynamic meshes and allocate network bandwidth efficiently between geometry and texture components. These techniques include 4D mesh compression methods, predictive bitrate control models, and adaptive streaming strategies that jointly optimize visual fidelity and network performance. The project further investigates networked spatial systems through an open hardware and software platform for experimental smart glasses. The platform exposes tunable components across the sensing, networking, and computational layers, enabling researchers and developers to study cross-layer trade-offs among latency, energy consumption, and visual quality. The resulting system will support reproducible research in networked spatial computing and provide new tools for evaluating real-time perception, streaming, and multi-user spatial applications. Additionally, the award will lead to educational activities, including the development of hands-on courses and virtual laboratory environments that use spatial computing systems to illustrate complex systems concepts. The research outcomes will allow students to observe and interact with abstract computing processes in real time, creating new opportunities for experiential learning in networking and distributed systems.