I am working in the Network Research Lab (NRL) with Dr. Beichuan Zhang and focusing on application and challenges of Named Data Networking in wireless networks and disruption tolerant networks. Previously I worked on scheduling algorithms and network coding for vehicle-to-infrastructure communication in Vehicular Ad-hoc Networks (VANETs). I am also interested in Database Systems, Natural Language Processing, Machine Learning and Software Engineering.

NDN in wireless ad-hoc, delay-tolerant and challenging networks

Over the past few decades, wireless communication technology has improved manifold, and the number of wireless devices has increased exponentially and will continue to be so in the coming years. However, the backbone of communication still relies upon the IP-based end-to-end protocol stack. As a result, communication in multi-hop wireless environments such as military networks on the battlefield or peer-to-peer data exchange under mobility such as vehicular networks suffers greatly under frequent link-breakage and packet loss from mobility, hidden station problems, to name a few. Application performance in such networks suffers manifold because of mandatory pathfinding and end-to-end communication of IP. Focusing on this "IP" backbone has led to more complex protocol designs, such as complex tree formation in multicast groups.

Named Data Networking or NDN re-imagines the communication stack from the ground up and focuses on the "data first" model. Such data-centric design with "named" content promises scalable and reliable data retrieval by the network through built-in multicast, loop detection, caching, hop-by-hop packet forwarding management. However, how to properly use NDN in wireless networks is still under vigorous investigation.

In my Doctoral research, I am focusing on the different segments of the network stack and NDN protocols to improve application performance. The overall work has the following significant sub-sections,

    • Improve data retrieval rate by improving the network layer performance. We avoid protocol messages altogether and only use Interest-Data exchange to learn path towards any closest data node and re-learn a newer one when loss or path breakage occurs through in-network RTT measurement.

    • Next, we show that adaptive-rate applications in ad-hoc wireless networks suffer from redundant data transmission because of stale PIT entries. Furthermore, out-of-order data retrieval introduces excessive channel contention. Thus, we use the congestion window limit and propose a novel consumer side RTT-based dynamic PIT lifetime setup to reduce redundant data transmission. Together, they significantly increase application throughput.

    • We then focus on the Link-layer to improve single-hop communication and minimize resource usage. This ongoing work promises to improve link bandwidth usage.

    • We are also investigating how we can adequately utilize NDN in delay-tolerant networks with minimal resource usage and high data exchange rate.

Network Coded Data Dissemination in RSU-based Vehicular Ad-hoc Networks (VANETs).

This project focuses on the challenges of efficient packet scheduling in a multi-RSU vehicular ad-hoc network. Application of network coding in packet scheduling for both single item and multi-item query has been investigated to minimize wireless broadcast data transmissions and overall Vehicle-to-RSU communication latency to provide improved road-safety and infotainment. In single item query, a vehicle asks for a single item which is a simple node or vertex in a graph. On the other hand, a vertex in a multi-item query consists of multiple data points. Moreover, data items can also have different sizes (heterogeneous). Our work achieved significant lower latency and wireless broadcast overhead with high data-retrieval rate, compared to the state-of-the-art techniques available at that time.

We focused on solving scheduling and optimization problems primarily in three different cases,

  • We used a weighted moving average-based coding technique to encode heterogeneous data items in single item request scenarios.

  • Next, we proposed a scheduling technique that encodes multi-item queries for homogeneous data items.

  • We also proposed cooperative cache transfer technique where an RSU deduces vehicles' movements to transfer cache information to next possible RSU(s) reducing bandwidth usage for cache information upload required for network coding.