In the not-so-distant future, many cities will be powered by AI, with smart sensors, cameras, and mobile phones continuously analyzing data to enhance safety, optimize traffic, and improve urban living. While this vision is almost within reach, a major roadblock stands in the way: processing speed.

Powerful AI models require significant computational resources and are challenging to deploy on devices with limited processing power, such as street cameras, sensors, and even our mobile phones.

Innovative AI Solutions for Smart Cities

To overcome these challenges, members of the Sense, Collect, and Move Data Center at the Data Science Institute (DSI) are developing new AI architectures that distribute tasks intelligently across the network with support from the semiconductor company MediaTek.

A large node of the NSF PAWR COSMOS testbed covering part of Hamilton Heights and Manhattanville.

Javad Ghaderi, Associate Professor of Electrical Engineering a faculty affiliate of DSI, and Gil Zussman, Professor and Chair of Electrical Engineering, as well as a DSI member, are using a combination of local devices (such as street cameras), edge servers positioned closer to the data source, and powerful cloud servers to handle the most computationally intensive tasks. This method allows each part of the network to do what it does best, optimizing the use of available resources.

“Our goal is to create a seamless system where AI tasks are automatically distributed based on the current network conditions and the capabilities of each device,” explains Professor Ghaderi. “For instance, a camera might perform initial object detection, while more detailed analysis is offloaded to an edge server or the cloud. This reduces latency and ensures that critical data is processed as quickly as possible.”

By optimizing how and where data is processed, the team aims to improve the overall efficiency and responsiveness of AI applications in urban settings, making them more reliable and effective.

The AI architectures that Ghaderi and Zussman develop will be tested out in the COSMOS testbed, located in West Harlem, which provides a unique and flexible environment to test and develop next-generation wireless systems in a dynamic, real world setting. Mashid Ghasemi, an Electrical Engineering PhD student, has been involved in the early research on this topic and will contribute to the implementation of architectures on the COSMOS testbed.     

Image obtained from one of the COSMOS nodes and processed in real time using the edge-servers.  

If all goes well, a smart city utilizing Ghaderi and Zussman’s distributed AI systems could improve everything from traffic congestion to trash collection to emergency response. 

Laying the Foundation for 6G and Beyond

Beyond the cityscape, this project is crucial for developing future 6G communication networks, the next step in wireless technology needed to support the ever-more-powerful applications people rely on, from new generative AI applications like ChatGPT to AI-powered features in social media and ecommerce platforms. 

“The explosive growth of AI has transformed user experiences across familiar applications and spawned a multitude of novel tools that empower people to enhance their productivity like never before,” said Ghaderi “But these innovations are entirely cloud-based, so how well they work depends on the strength of our network connection and the amount of server traffic we are competing with. The new architectures we are developing can make these services more reliable even when the network connection is weak or the server is slow.” 

CoSMOS
A medium node of the COSMOS testbed that includes sub-6GHz radios, mmWave radios, and cameras. Such nodes need to be commented to the edge-cloud and cloud for processing data collected from the cameras and mmWave radios.

As a leading semiconductor company heavily invested in the future of wireless communications and AI technologies, MediaTek sees significant potential in the collaboration with Columbia researchers. Their involvement in this groundbreaking project not only supports the advancement of smart city infrastructures but also aligns with their vision for 6G communication systems and pioneering the future of AI technologies.

Dr. Doru Calin, AVP, Head of U.S. 6G Wireless Research Center at MediaTek said “As a global leader in semiconductor solutions, MediaTek powers the technology that connects and enriches everyday life. From smartphones, smart homes and autos, to transformative technologies like AI and 5G, MediaTek lays the foundation for a smarter and more connected world. MediaTek is pleased to collaborate with Columbia University researchers to lay the foundation of future 6G communication systems and pioneer the future AI technologies.”