Socially- and Geographically-Aware Modeling Framework for User Mobility in Wireless Networks

The "Socially- and Geographically-Aware Modeling Framework for User Mobility in Wireless Networks" is a project sponsored by the NSF under the NeTS Program. One of its main goals is to develop novel, socially-inspired human mobility models that also account for geographic diversity of the region of interest. The proposed models consider features observed in real human networks such as differential popularity, transitivity and clustering as well as geographical features and preferences from users.

Project Goals
One of the project’s main goal is to develop novel, socially-inspired human mobility models that also account for geographic diversity of the region of interest. The proposed models consider features observed in real human networks such as differential popularity, transitivity and clustering as well as geographical features and preferences from users. Another important deliverable of our work is a suite of tools that implement the proposed models and that can be used by other researchers and practitioners in the evaluation of mobile systems and protocols.

Accomplishments
In addition to contributions to user mobility characterization and modeling in wireless networks, this project  also developed: (1) terrain-aware IoT deployment algorithms for outdoor applications, (2) computational intelligence techniques to estimate near-future network conditions, and (3) congestion control for disruption-tolerant networks (DTNs) that can autonomically adjust to network dynamics. More specifically, our accomplishments to-date include:

  • We have developed socially-inspired human mobility models that also account for geographic diversity of the region of interest. As we continue the development of our socially- and geographically-inspired user mobility models, we have also been exploring applications in other disciplines, e.g., urban planning, smart transit- and transportation systems, etc.
  • Under the research activity focusing on IoT and sensor network deployment, we introduced a novel distributed greedy deployment algorithm, we call grid partition, and compared with two other kinds of distributed optimization of algorithms, namely simulated annealing and evolutionary strategy using an objective function that maximizes sensor coverage while keeping the network connected. We have also developed a terrain model, called TerrainLOS, for the Cooja/Contiki wireless sensor network simulator/emulator. We have contributed our TerrainLOS implementation to the Cooja/Contiki software distribution.
  • Another research trust under this project focuses on the use of machine learning techniques to estimate network performance. We have proposed a technique called Smart Experts for Network State Estimation, or SENSE, which uses a simple, yet effective algorithm that combines a machine-learning method known as FixedShare Experts and Exponentially Weighted Moving Average (EWMA). SENSE can be used to improve the performance of a wide range of network protocols and applications. We have applied SENSE to improve the performance of the IEEE 802.11. More specifically, we employed SENSE to decide whether to use congestion avoidance (RTS/CTS) or pure carrier sensing in 802.11. We have also introduced a simple, yet effective machine learning approach to adjust the value of IEEE 802.11’s Contention Window based on present- as well as recent past network contention conditions.
  • Under the DTN congestion control research thrust, we have introduced the first reinforcement learning based congestion control algorithm which is able to adapt a wide range of DTN scenarios and conditions.

Outcomes
To-date, this project has produced a number of publications in peer-reviewed conferences and journals. Some of our key results to-date include: (1) development of the first framework to model “way-point based” mobility; (2) development of the first sensor network deployment algorithm for outdoor applications that considers terrain; (3) development of the first terrain-based communication model for sensor network simulation platforms; (4) development of the first computational-intelligence based congestion control framework for Disruption-Tolerant Networks (DTNs), which can deliver adequate performance in a wide range of DTN scenarios and applications; (5) development of a simple, yet efficient machine learning mechanism to improve network protocol performance. Our work was the first to employ a machine learning based approach to dynamically control the use of the IEEE 802.11’s RTS/CTS handshake as well as dynamically set 802.11’s Contention Window to adjust to current network contention conditions.

For further information about this project, please follow the links below.

Project Links
Personnel
Publications
Collaborators
Broader Impact