ALCA: agent learning–based clustering algorithm in vehicular ad hoc networks

Abstract

Vehicular ad hoc network (VANET) is an emerging technology which can be used in various applications such as intelligent transport technology, safety applications, etc. But one of the major issues in VANETs is how to cluster the vehicles on the road for efficient operations such as routing, mobility management and generating safety alarms. Clustering of vehicles has been widely used for routing and data dissemination in VANETs. But due to the high mobility of the vehicles/nodes on the road, it is quite difficult to find the exact route in VANETs. Keeping in view of the above issue, in this paper, we propose a new agent learning–based clustering and routing in VANETs. Agents learn from the environment in which they are deployed, and accordingly, their action performed is rewarded or penalized with certain values. Each agent performs its task in collaboration with the other agents, i.e. agents communicate with each other in collaborative manner for information sharing. The deployed agents estimate the mobility of the vehicles, and based upon their learning, clustering of vehicles is performed. An Agent Learning–based Algorithm for Clustering is proposed. The performance of the proposed scheme is evaluated using extensive simulation with respect to the various metrics such as message transmission ratio, percentage of connectivity, node participation, cluster head duration, and connectivity preservation ratio. The results obtained show that the proposed scheme is effective in performing fast clustering and converges quickly to the final solution.

Appendix

We can formulate the probability density function (PDF) for the vehicles arrival at any junction of the road as follows:

$$ f = \int\limits_{ - \infty }^{ + \infty } {f_{x} (t)e^{ - \eta t} } $$

(8)

where η is the request arrival rate and

$$ f_{x} (t) = \left| \begin{array}{l} M^{*} {\text{TFV}}\mu_{\text{CH}} e^{ - \mu t} \hfill \\ M^{*} {\text{TFV}}\mu_{\text{CH}} e^{ - \mu t} \hfill \\ M^{*} {\text{TFV}}\mu_{\text{CH}} e^{ - \mu t} \hfill \\ \end{array} \right\rangle \left. \begin{array}{l} avg\_value < M < peak\_value \hfill \\ peak\_value < M < avg\_value \hfill \\ avg\_value < M < peak\_value \hfill \\ \end{array} \right| $$

(9)

where M is the variable controls the average and peak values of the traffic flow

$$ f(\mu_{\text{CH}} ) = \int\limits_{0}^{\infty } {f_{x} (t)e^{{ - \mu_{\text{CH}} t}} dt = E(e^{{ - \mu_{\text{CH}} t}} )} $$

(10)

Then, the average duration in the cluster can be calculated as follows:

$$ E(duration + \Updelta t)^{0} = \int\limits_{{Z_{1} }}^{{}} {{\text{TFV}}\mu_{\text{CH}} e^{{ - \mu_{\text{CH}} t}} dx} + \int\limits_{{Z_{2} }}^{{}} {{\text{TFV}}\mu_{\text{CH}} e^{{ - \mu_{\text{CH}} t}} dx} + \int\limits_{{Z_{3} }}^{{}} {{\text{TFV}}\mu_{\text{CH}} e^{{ - \mu_{\text{CH}} t}} dx} $$

(11)

$$ E(duration + \Updelta t)^{1} = \int\limits_{{Z_{1} }}^{{}} {TFV\mu_{\text{CH}} e^{{ - \mu_{\text{CH}} t}} dx} + \int\limits_{{Z_{2} }}^{{}} {{\text{TFV}}\mu_{\text{CH}} e^{{ - \mu_{\text{CH}} t}} dx + \int\limits_{{Z_{3} }}^{{}} {{\text{TFV}}\mu_{\text{CH}} e^{{ - \mu_{\text{CH}} t}} dx} } $$

(12)

$$ E(duration + \Updelta t)^{n} = \int\limits_{{Z_{1} }}^{{}} {{\text{TFV}}\mu_{\text{CH}} e^{{ - \mu_{\text{CH}} t}} dx + \int\limits_{{Z_{2} }}^{{}} {{\text{TFV}}\mu_{\text{CH}} e^{{ - \mu_{\text{CH}} t}} dx} } + \int\limits_{{Z_{3} }}^{{}} {{\text{TFV}}\mu_{\text{CH}} e^{{ - \mu_{\text{CH}} t}} dx} $$

$$ \begin{gathered} E(mean\_duration) = E(duration)^{0} + E(duration + \Updelta t)^{1} + \cdots = \frac{E(duration + \Updelta t)}{{1 - E(duration + \Updelta t)^{n} }}\end{gathered} $$

Hence, the theorem is proved.