“My focus is on designing algorithms and network protocols to deliver fresh information”
Get to know better our Research Assistant Professor Jaya Champati in this second part of the interview given by the leader of the Edge Networks Group
24 June 2021
In this second interview, Jaya Champati, Research Assistant Professor at IMDEA Networks, speaks about his main lines of research: the freshness of information in networked systems and learning at the Edge. Furthermore, he makes a prediction about what innovations will be seen in the near future.
What do you hope to accomplish with your research?
My current research topics are mainly two. The first one is the Freshness of Information in networked systems. So far, communication in networked systems has been designed to carry voice traffic, video traffic, and the protocols are designed for it and they are doing pretty well now. But if you come to emerging applications like autonomous driving and automation in industry 4.0, there the information itself that is received might not be useful if it is too stale or if it is generated a long time ago. Here comes the freshness of the information into the limelight. And my focus is on designing algorithms and network protocols to deliver fresh information so that future networks can better support these emerging applications.
And the second research topic I am working on is Learning at the Edge. So, there is a huge thrust for moving machine learning applications from data centers to the edge for obvious reasons: because of all the massive data that is being collected from mobile phones and IoT devices and the first ports of entry for them is the cellular network base stations and WiFi access points. If we can do training at the edge, then you can also save bandwidth that is required to move this massive data to data centers and also delay that is required for the inference to come back to these edge devices.
I was mainly focusing on which data to collect from these devices because machine learning models are as good as the data that is fed into them. We are investigating which data to select and when to transmit them. These are important both from an energy perspective because if you just collect random data and send some of them might be redundant, and also you can save bandwidth for the wireless transmissions. My focus is primarily on designing algorithms for such kinds of selections.
What advances do you expect to see in edge networks?
We already saw a big jump from 4G LTE to 5G in terms of the wide variety of applications that 5G can support. And two of the key enablers are edge computing and network slicing. Apart from supporting these applications, we need support from the network to deploy these applications at a large scale. For example, now we see there are very few autonomous cars but if we want to think of a city full of autonomous cars, network support is required for ensuring their safety, coordinating them, and also managing the traffic in the city.
So, I think the key innovations will be made at multiple levels: one is in the network stack (new protocols coming into place) and in the evolved packet core (how the processing is done, from the data that is coming from different devices) and at the edge computing level (you have to look at the load balancing, machine learning models that are being deployed and so on).