[BMB17] Mobility Prediction in Vehicular Networks: An Approach through Hybrid Neural Network under Uncertainty

Conférence Internationale avec comité de lecture : International Conference on Mobile Secure and Programmable Networking (MSPN 2017), June 2017, pp.195-217, Series Springer LNCS 10566, France,

Auteurs: S. Banerjee , P. MÃ¼hlethaler , S. Bouzefrane

Mots clés: Vehicular Network, Mobility Prediction, Link Failure, Fuzzy Constrained Boltzmann Machine, VANET, Uncertainty.

Résumé: Conventionally, the exposure regarding knowledge of the inter vehicle link duration is a significant parameter in Vehicular Networks to estimate the delay during the failure of a specific link during the transmission. However, the mobility and dynamics of the nodes is considerably higher in a smart city than on highways and thus could emerge a complex random pattern for the investigation of the link duration, referring all sorts of uncertain conditions. There are existing link duration estimation models, which perform linear operations under linear relationships without imprecise conditions. Anticipating, the requirement to tackle the uncertain conditions in Vehicular Networks, this paper presents a hybrid neural network-driven mobility prediction model. The proposed hybrid neural network comprises a Fuzzy Constrained Boltzmann machine (FCBM), which allows the random patterns of several vehicles in a single time stamp to be learned. The several dynamic parameters, which may make the contexts of Vehicular Networks uncertain, could be vehicle speed at the moment of prediction, the number of leading vehicles, the average speed of the leading vehicle, the distance to the subsequent intersection of traffic roadways and the number of lanes in a road segment. In this paper, a novel method of hybrid intelligence is initiated to tackle such uncertainty. Here, the Fuzzy Constrained Boltzmann Machine (FCBM) is a stochastic graph model that can learn joint probability distribution over its visible units (say n) and hidden feature units (say m). It is evident that there must be a prime driving parameter of the holistic network, which will monitor the interconnection of weights and biases of the Vehicular Network for all these features. The highlight of this paper is that the prime driving parameter to control the learning process should be a fuzzy number, as fuzzy logic is used to represent the vague and and uncertain parameters. Therefore, if uncertainty exists due to the random patterns caused by vehicle mobility, the proposed Fuzzy Constrained Boltzmann Machine could remove the noise from the data representation. Thus, the proposed model will be able to predict robustly the mobility in VANET, referring any instance of link failure under Vehicular Network paradigm.

BibTeX

@inproceedings {
	BMB17,
	title	=	"{Mobility Prediction in Vehicular Networks: An Approach through Hybrid Neural Network under Uncertainty}",
	author	=	" S. Banerjee and P. MÃ¼hlethaler and S. Bouzefrane ",
	booktitle	=	"{International Conference on Mobile Secure and Programmable Networking (MSPN 2017)}",
	year	=	2017,
	month	=	"June",
	series	=	"Springer LNCS 10566",
	pages	=	"195-217",
	address	=	" France",
}