Overview
Mobile users autonomously share data by exploiting unused communication opportunities between the short-range radios (e.g., Bluetooth, WiFi) of their mobile devices. Since users may move randomly and their devices are subject to power depletion or functional failure, only intermittent network connectivity is available among users. Two users only communicate when they opportunistically contact, i.e., move into the communication range of their devices. These networks are hence called Opportunistic Mobile Networks, also known as Delay/Disruption Tolerant Networks (DTNs).
My research seeks to percolate various aspects of mobile users' social behavior patterns, including social contact, social interest and social relation, for realizing synergistic and efficient network operation. An overarching strategy in my research is to explore the correspondence of sociology concepts in opportunistic communication among mobile users. These sociology concepts include:
- Centrality: the social importance of a user in facilitating the communication among other users.
- Community: users are formed into groups according to their social relations
- Homophily: users with common interests are usually friends and tend to contact each other more frequently.
Being opposite to online social networks (e.g., Facebook, Twitter) which comprise explicit social links among users, users' social behavior patterns in opportunistic mobile
networks are implicitly exhibited by their contact process. Each contact in such process only provides partial observability to the underlying social link between two users.
I devised analytical measures of sociology concepts to fully and accurately extract users’ social behavior patterns from stochastic modeling of their contact process, and
exploited sociology concepts for efficient network decision on data access.
My work imposes broad research impact on academic, industry and military aspects. In the academic aspect, my work has motivated researchers worldwide to further explore the essentials of users' social behavior patterns on affecting mobile network decisions. It has been incorporated into university courses and seminars (UC Berkeley, University of Texas at Arlington, Lehigh University, Aalto University, National TsingHua University, etc) for graduate teaching or academic discussion. In the industry aspect, I have been in close collaboration with IBM Research, BBN and ArtisTech, and my work has been planned to be incorporated into their commercial product. In the military aspect, my research has been established as a mainstream topic in the 10-year Network Science Collaborative Technology Alliance (NS CTA) project funded by US ARL. It is involved in developing the military testbed (based on the AlgoLink and NS-3 simulation platforms) for in-network caching experiments.
Social Contact
To transmit data in opportunistic mobile networks, researchers adopt the idea of "carry-and-forward". A node carries data as relay when no route to the destination exists, and later forwards data to another relay upon opportunistic contact. Therefore, appropriate data forwarding decision is vital for data being carried by relays with the best chance to contact the destination. Estimation of such chance is based on nodes' contact patterns in the past, which can be inaccurate due to node dynamics. Data delivery is hence uncertain and probabilistic.
I observe that these contact patterns are determined by users' social behaviors, based on which I developed solutions for improving the performance of opportunistic communication. I investigated the skewness of transient social contact distribution of mobile users over realistic mobile networks in university campus, conference site and suburban areas [ICNP10, TMC12]. Based on the experimental investigation results, I developed centrality metric which accurately estimates nodes’ capabilities of contacting others during the specific short time period for data forwarding. Moreover, my work [MobiHoc09, TON12] is the first to study multicast relay selection from a social perspective. I developed a centrality metric which quantitatively calculates the cumulative probability for a relay to deliver data to multiple destinations, and furthermore suggested community-based relay selection strategies for hierarchical multicasting. My proposed solutions analytically improve data delivery ratio and delay, while incurring controlled data transmission overhead. I also studied eliminating the randomness of user mobility from a social perspective, based on accurate characterization of user mobility pattern [MobiHoc10]. I proposed to formulate user mobility with Hidden Markov Model (HMM), based on which I developed fine-grained mobility characterization techniques.
Related Publications:
[TON12] Social-aware Multicast in Disruption Tolerant Networks [pdf]
Wei Gao, Qinghua Li, Bo Zhao, and Guohong Cao, IEEE/ACM Transactions on Networking, to appear.
[TMC12] On Exploiting Transient Social Contact Patterns for Data Forwarding in Delay Tolerant Networks [pdf]
Wei Gao, Guohong Cao, Thomas F. La Porta, and Jiawei Han, IEEE Transactions on Mobile Computing, to appear.
[ICNP10] On Exploiting Transient Contact Patterns for Data Forwarding in Delay Tolerant Networks [pdf]
Wei Gao and Guohong Cao, in Proceedings of the 18th IEEE Int'l Conference on Network Protocols (ICNP), 2010.
(Acceptance Ratio: 31/170=18.2%)
[MobiHoc10] Fine-Grained Mobility Characterization: Steady and Transient State Behaviors [pdf]
Wei Gao and Guohong Cao, in Proceedings of the 11th ACM International Symposium on Mobile Ad Hoc Networking and Computing (MobiHoc), 2010.
(Acceptance Ratio: 26/154=16.9%)
[MobiHoc09] Multicasting in Delay Tolerant Networks: A Social Network Perspective [pdf]
Wei Gao, Qinghua Li, Bo Zhao, and Guohong Cao, in Proceedings of the 10th ACM International Symposium on Mobile Ad Hoc Networking and Computing (MobiHoc), 2009.
(Acceptance Ratio: 31/175=17.7%) (110+ citations)
Social Interest
My research focuses on providing users with prompt access to data that they are interested in. User interests are generally heterogeneous but socially correlated. However, it is difficult for mobile users to maintain the global knowledge about others' interests via opportunistic contacts. It is hence hard to determine where data should be placed among users. In my proposed solution [INFOCOM11], data is actively disseminated to users that are interested in the data. My work measures users' interest in data as probability, maintains information about user interest within an appropriate scope, and incoporates social interests of users into measuring user centrality for data dissemination. It efficiently utilizes the limited network resources and analytically ensures cost-effectiveness of data access.
I also studied to replicate and cache data at specific network locations based on user query history, so that user queries in the future are responded with less delay [ICDCS11, ICNP11, ICDCS12]. During my summer internships at IBM Research, I developed techniques to cache data at users with high centrality, so that data can be easily accessed by other users. Such decision of caching location is also affected by users’ homophily properties so that data is shared more efficiently among users with similar interests. I also developed efficient techniques to maintain the freshness of cached data at individual users in a completely distributed manner, in cases that data is periodically updated by the sources.
Related Publications:
[ICDCS12] Distributed Maintenance of Cache Freshness in Opportunistic Mobile Networks [pdf]
Wei Gao, Guohong Cao, Mudhakar Srivatsa, and Arun Iyengar, in Proceedings of the 32nd Int'l Conference on Distributed Computing Systems (ICDCS), 2012.
(Acceptance Ratio: 71/515=13.7%)
[ICDCS11] Supporting Cooperative Caching in Disruption Tolerant Networks [pdf]
Wei Gao, Guohong Cao, Arun Iyengar, and Mudhakar Srivatsa, in Proceedings of the 31st Int'l Conference on Distributed Computing Systems (ICDCS), 2011.
(Acceptance Ratio: 15%)
[ICNP11] Contact Duration Aware Data Replication in Delay Tolerant Networks [pdf]
Xuejun Zhuo, Qinghua Li, Wei Gao, Guohong Cao, and Yiqi Dai, in Proceedings of the 19th IEEE Int'l Conference on Network Protocols (ICNP), 2011.
(Acceptance Ratio: 31/189=16.4%)
[INFOCOM11] User-Centric Data Dissemination in Disruption Tolerant Networks [pdf]
Wei Gao and Guohong Cao, in Proceedings of the 30th IEEE Conference on Computer Communications (INFOCOM), 2011.
(Acceptance Ratio: 291/1823=15.9%)
Social Relation
Efficient data access requires users to cooperate and share data with each other. However, in practice users are socially selfish, such that a user may be conservative when contributing resources for others. This selfishness is affected by users' social relations. For example, users are more willing to cooperate with friends, but are prudent to cooperate with strangers. We developed incentive schemes to stimulate users' cooperation while enforcing their social selfishness. Our work [AdHoc12] is the first to quantitatively analyze the correlation between users' packet dropping behavior and social selfishness, and highlights the idea of incorporating users' willingness for transmitting data into relay selection. We also proposed an incentive-based framework for 3G traffic offloading [ICNP11], in which we investigate the tradeoff between the amount of 3G traffic being offloaded and users' satisfaction. We demonstrated that the incentive cost with a given offloading target can be minimized via reverse auctions. The trustfulness and rationality of auctions are formally proved, and the potential of traffic offloading is stochastically predicted.
Related Publications:
[ICNP11] Win-Coupon: An Incentive Framework for Cellular Traffic Offloading [pdf]
Xuejun Zhuo, Wei Gao, Guohong Cao, and Yiqi Dai, in Proceedings of the 19th IEEE Int'l Conference on Network Protocols (ICNP), 2011.
(Acceptance Ratio: 31/189=16.4%)
[AdHoc12] A Routing Protocol for Socially Selfish Delay Tolerant Networks [pdf]
Qinghua Li, Wei Gao, Sencun Zhu, and Guohong Cao, Ad Hoc Networks, Elsevier, to appear.
Social-Aware Data Diffusion in Delay Tolerant MANETs [pdf]
Yang Zhang, Wei Gao, Guohong Cao, Tom La Porta, Bhaskar Krishnamachari, and Arun Iyengar, Handbook of Optimization
in Complex Networks: Communication and Social Networks, Springer Publisher, 2010.