Using clustering for target tracking in vehicular ad hoc networks

Abstract

Vehicular Ad-hoc Networks (VANETs) have recently drawn the attention of academic and industry researchers due to their potential applications in enabling various Intelligent Transportation Systems (ITS) applications for safety, entertainment, emergency response, and content sharing. Another potential application for VANETs lies in vehicle tracking, where a tracking system is used to visually track a specific vehicle or to monitor a particular area. For such applications, a large volume of information is required to be transferred between certain vehicles and a command and control centre, which can easily congest the wireless network in a VANET if not designed properly. The development of low-delay, low-overhead, and precise tracking systems in VANET is a major challenge requiring novel techniques to guarantee performance and to reduce network congestion. Among the several proposed data dissemination and management methods implemented in VANETs, clustering has been used to reduce data propagation traffic and to facilitate network management. However, clustering for target tracking in VANETs is still a challenge due to the dynamic nature of such networks. We have proposed two cluster-based algorithms for target tracking in VANETs in our previous works [1], [2]. These algorithms provide a reliable and stable platform for tracking a vehicle based on its visual features. In this paper, we have demonstrated the performance evaluation and testing results of both of our algorithms in the context of vehicular tracking under various scenarios. We have also compared the performance of both our algorithms to assess the performance of distributed algorithms as compared to centralized cluster-based target tracking algorithms. Besides, we have tested two data dissemination techniques for information delivery. Performance evaluation results demonstrate clearly that the proposed clustering schemes provide better performance for target tracking applications as compared to other cluster-based algorithms.

Introduction

Vehicular Ad hoc Networks (VANETs) play an important role in Intelligent Transportation Systems (ITS) by providing critical information about roads and traffic condition, sending safety messages, and providing entertainment for passengers. In VANETs, vehicles can connect to each other for many purposes such as exchanging safety and infotainment messages. A special characteristic of VANET nodes, compared to nodes of other ad hoc networks such as MANETs, is the abundant on-board processing, storage and energy resources of the vehicles, which makes them a suitable platform for processing complex algorithms for a number of applications. Over the last few years, a number of research works have been conducted on VANETs, mainly focusing on routing techniques and data dissemination under various road and traffic conditions [3], [4], [5], localization of nodes [6], [7], location privacy protection [8], communication security [9], social networking and advertisement [10], [11].

While VANET is still in its infancy, a number of non-safety applications have been proposed in the literature. One of the envisioned applications is target tracking, where a target vehicle is located and tracked using on-board vehicle sensors such as cameras. Such applications may be used by police agencies to locate a specific vehicle with particular visual features such as license plate information, color, model, etc. Even though police agencies may use pre-installed security camera infrastructure across the city, the cost of installing cameras to cover all roads can be very high. In addition, there is the probability of losing the target in non-monitored areas, i.e., “blind spots”. However, most new models of vehicles are being equipped with front and rear cameras, proximity sensors, and on-board communication capabilities that can be used as enabling components for a distributed mobile tracking system. Another application of such system is passive monitoring to collect pictures or video footage of incidents that happened in areas where security camera systems are unavailable, using only the cameras of nearby vehicles.

One of the challenges in continuous monitoring systems in VANETs is bandwidth availability, which can be a limiting factor especially when there are multiple data sources in close proximity, and are streaming video data simultaneously [12], [13]. A traditional solution to control bandwidth usage in ad hoc networks is to segment the network into clusters and select one representative, i.e., a cluster head, for each cluster to act as a connection point to the cluster [14]. However in a highly dynamic environment such as VANETs, the selection of appropriate metrics for cluster head election and cluster membership can be a challenging problem as vehicles constantly enter and leave the clusters.

In previous work, we have proposed two cluster-based protocols for vehicle tracking in VANETs [1], [2]. Target tracking can be a simple task when the target vehicle has a Global Positioning System (GPS), with the location data communicated to external entities. However, we assume that GPS devices are not available or have been turned off on target vehicles. In order to solve this issue, we can rely on cluster formation around a target and visual identification of targets using the on-board cameras of neighboring vehicles and reporting the location and visual information of the target to a control center. The control center could be a police station or cruiser looking for a special vehicle based on its visual description. Therefore, in the absence of a proper data dissemination mechanism, every vehicle that detects the target will broadcast location information of the target towards the control center. In VANETs, nodes communicate with each other in a multi-hop fashion. Assuming the control center is located within a multi-hop distance from the target, there is a high probability of packet collision and packet loss due to concurrent transmission of location information by all the neighboring vehicles [15], [16] resulting in a significant drop in data delivery ratio. Also, the control center might receive duplicate messages which are unnecessary and redundant. This problem is due to the unavailability of a central aggregation node to collect, process and aggregate information from neighboring vehicles. Another concern in such a system is the data overload in the control center due to direct transmission of position information to the central entity by all vehicles that detected the same target. In order to address these problems, we proposed to use a clustering approach to coordinate data transmission from vehicles around the target. Therefore, neighboring vehicles that can detect the target join a cluster and select a cluster head (CH). The neighbor vehicles send their location information to the CH. The CH is responsible for aggregating the information and sending it to the control center. Therefore, instead of every node sending its information to the control center separately, only one node is responsible for delivering the information to the control center.

The challenges toward designing a high-performance and efficient clustering algorithm mostly include clustering stability and overhead. Due to high speed of vehicles in VANETs, the cluster topology changes frequently, which induces a high control overhead. Also, the CH role may change too quickly among eligible vehicles, which causes a high number of CH changes. Any change in the cluster topology requires the dissemination of control messages within the cluster to inform other cluster member vehicles about the change. The studies conducted in [17], [18] show the clustering overhead induced by constant broadcasting of control messages. Thus, it is critical to use appropriate cluster membership and CH selection rules in order to extend the lifetime of cluster members and cluster heads as much as possible. Control packets can congest the cluster if not managed properly. Therefore, reducing cluster control overhead is a necessary step toward an efficient clustering protocol. In this paper we demonstrate performance evaluation results of our proposed cluster-based target tracking algorithms in various scenarios. Also, a comparative study of our proposed protocols to an existing VANET clustering algorithms is provided to show the improved performance of both our proposed algorithms.

The rest of this paper is organized into six sections. Section 2 provides a literature review of VANET's features and applications, cluster-based techniques for VANET environment, and target tracking in these networks. In Section 3 we provide a brief review of our proposed cluster-based target tracking algorithms for VANETs, the definition of the functions and techniques used in the proposed algorithms, and the information routing techniques. Section 4 provides the simulation results and evaluation of the proposed protocols. Finally, conclusion and future works are represented in Section 5.