A Nonparametric Bayesian Approach for Opportunistic Data Transfer in Cellular Networks
The number of mobile Internet users is growing rapidly, as well as the capability of mobile Internet devices. As a result, the enormous amount of traffic generated everyday on mobile Internet is pushing cellular services to their limits. We see great potential in the idea of scheduling the transmission of delay tolerant data towards times when the network condition is better. However, such scheduling requires good network condition prediction, which has not been effectively tackled in previous research. In this paper, we propose a Dynamic Hidden Markov Model (DHMM) to model the time dependent and location dependent network conditions observed by individual users. The model is dynamic since transition matrix and states are updated when new observations are available. On the other hand, it has all the properties of a Hidden Markov Model. DHMM can predict precisely the next state given the current state, hence can provide a good prediction of network condition. DHMM has two layers, the top layer is Received Signal Strength (RSS) and the bottom layer consists of states, defined as a mixture of location, time and the signal strength itself. Since the state is defined as a mixture, it is hidden and the number of states is also not known a priori. Thus, the Nonparametric Bayesian Classification is applied to determine the hidden states. We show through simulations that when combined with a Markov decision process, the opportunistic scheduling can reduce transmission costs up to 50.34% compared with a naive approach.
KeywordsTime Slot Signal Strength Cognitive Radio Network Condition Gaussian Mixture Model
Unable to display preview. Download preview PDF.
- 1.Cisco visual networking index: Global mobile data traffic forecast update, 20102015 (2010), http://www.cisco.com/en/US/solutions/collateral/ns341/ns525/ns537/ns705/ns827/white_paper_c11-520862.html
- 2.Trestian, A.K.I., Ranjan, S., Nucci, A.: Taming user-generated content in mobile networks via drop zones. In: Proc. of IEEE INFOCOM 2011 (April 2011)Google Scholar
- 3.Wang, Y., Liu, X.: UPDATE: User-Profile-Driven Adaptive Transfer preprint, http://www.cs.ucdavis.edu/~liu/preprint/Update.pdf
- 6.Teh, Y.W.: Dirichlet processes. Encyclopedia of Machine Learning. Springer, New York (2007)Google Scholar
- 7.Ra, M.-R., Paek, J., Sharma, A., Govindan, R., Krieger, M.H., Neely, M.J.: Energy-delay tradeoffs in smartphone applications. In: Banerjee, S., Keshav, S., Wolman, A. (eds.) MobiSys, pp. 255–270. ACM (2010)Google Scholar
- 9.Mangold, S., Kyriazakos, S.: Applying pattern recognition techniques based on hidden Markov models for vehicular position location in cellular networks. In: IEEE Vehicular Technology Conference (1999)Google Scholar
- 12.Ding, N., Ou, Z.: Variational nonparametric Bayesian Hidden Markov Model. In: The Proceedings of IEEE International Conference on Acoustics, Speech, and Signal Processing (ICASSP), Dallas, TX, USA (2010)Google Scholar
- 14.Saad, W., Han, Z., Poor, H.V., Basar, T., Song, J.B.: A cooperative Bayesian nonparametric framework for primary user activity monitoring in cognitive radio Networks. IEEE Journal on Selected Areas in Communications, special issue on Cooperative Network (accepted)Google Scholar
- 16.Nguyen, N.T., Zheng, G., Han, Z., Zheng, R.: Device fingerprinting to enhance wireless security using nonparametric Bayesian method. In: IEEE Annual IEEE Conference on Computer Communications INFOCOM, Shanghai (April 2011)Google Scholar