Presenting a dynamic target tracking algorithm based on prediction in wireless sensor network

Computer Engineering-Computer Architecture Master Thesis

Abstract

With the advancement of electronic device manufacturing technology and the cost-effectiveness of large-scale sensor networks, wireless sensor networks provide research fields with rapid growth and great interest, which have attracted a lot of attention in recent years. Large-scale wireless sensor networks containing several hundred to several tens of thousands of sensors bring a wide range of applications and challenges. The special features of these networks provide the possibility of using them in applications such as control and investigation of accident-prone areas, border protection and security and military care. One of the most important applications envisioned for these networks is target tracking. In this application, wireless sensor networks use the sensors that make up this network to sense and detect a specific target and follow it in the area monitored by the network. Due to the fact that the sensors in this type of networks have energy limitations and the communication between sensors is done wirelessly, it is very important to pay attention to the problem of power consumption and error-free tracking of several moving targets simultaneously in these networks. Target tracking algorithms in sensor networks, in terms of their application and performance, are divided into four categories: message-based, tree-based, prediction-based, and clustering-based protocols. Meanwhile, protocols based on clustering are optimal in terms of energy consumption. So far, many methods have been designed to solve the energy problem, such as fast target tracking algorithms, DPT and CDTA. The fast target tracking algorithm has the ability to track fast targets, but one of its disadvantages is the high amount of communication in the network due to the small size of the clusters. DPT algorithm has a predictive algorithm with low complexity, but one of its disadvantages is its inability to track multiple targets simultaneously. The disadvantages of the CDTA algorithm include the lack of an error correction procedure to re-identify the lost target, network segmentation based on the network model, and its inability to track multiple targets at the same time. In the proposed algorithm, a clustering point of view based on prediction has been used in order to be scalable in the network and optimal energy consumption, so that it is resistant to possible failures of sensors and wrong predictions of the target location. In this algorithm, an error correction procedure has been presented so that the algorithm is able to re-identify the target when the target is out of range of the sensors due to its high speed or a sudden change of direction. The results obtained by the simulator show that the proposed algorithm is able to track several targets simultaneously and also the proposed algorithm reduces the energy consumption in sensor networks as much as possible by reducing the communication between clusters and the possibility of losing the target. 

Key words: sensor networks, tracking moving targets, network scalability, network model, predictive algorithm, error correction algorithm

Chapter One

Introduction

1-1- Description and importance of the topic

One of the networks that has attracted a lot of attention in recent years is wireless sensor networks [1]. (WSN) are Sensor networks are composed of a large number of sensors that after being distributed in the area, the sensors that are located near an event collect information about the desired event in the environment after detecting that event and send the information obtained from the event to the well sensor. A well sensor is a sensor that is connected to a base station[2] that is located outside of sensor networks[1]. The sensors of these networks have a wireless interface, which has caused these networks to be set up on the ground, underwater and other dangerous or inaccessible places. Therefore, sensor networks are able to cover areas that other networks cannot cover, and in fact, sensor networks provide the possibility of interaction between humans, the environment, and machines. The expansion of wireless sensor networks began with military applications, but today, with the rapid expansion of sensor network applications, wireless sensor networks are used in the fields of disaster relief, environmental control and biodiversity mapping, smart structures, facility management, agriculture, medicine and health, transportation, remote processing, and target tracking.For this reason, today many advances have been made in the field of electromechanical subsystems to enable the development of smart sensors [1]. One of the applications mentioned for sensor networks is the tracking of moving targets, the purpose of which is to follow a specific object in a predetermined space called the sensor field and detect the path of that object. This application can be made more complete with the ability to identify a specific target among various targets. For this purpose, sensors with different technologies that can measure various characteristics of a phenomenon are used in target tracking. These sensors consist of four units: power unit, information processing unit, communication unit and sensing unit. These sensors can be of the presence, vibration, light, sound, laser and image sensors, among which image sensors because they carry a lot of information are of high importance in target tracking applications to identify a specific target in battlefields or buildings and public places [2]. Due to the limitation of the power unit of sensors and the high energy consumption of image sensors compared to other types of sensors, optimizing energy consumption is considered one of the important challenges of sensor networks. becomes In this regard, the energy consumption of sensor components including micro sensors, analog to digital converter, signal processor, transmitter and receiver should be reduced as much as possible. Researches have shown that the energy required for communication is higher than other energy-consuming units of sensors due to the high volume of audio and video data sent by video sensors and as a result of imposing a large overhead on the data transmission system [2].

Since target tracking applications require sending information to the user in real-time and therefore many calculations are done in real-time in each sensor, a lot of power is always being consumed in the sensor network and for this reason Target tracking is one of the applications that consumes a lot of power. Due to the fact that optimal power consumption ensures the stability and reliability of sensor networks in difficult conditions and the high energy consumption in sensor networks doubles the importance of providing target tracking algorithms with low power consumption. In central approaches, only one sensor is responsible for identifying the target at any time, and therefore the accuracy of target tracking will decrease and the energy of the sensors will not be optimally consumed due to the imposition of heavy calculations. In these methods, by increasing the number of sensor nodes in the network, more messages are sent to the well sensor, which causes a lot of network bandwidth usage, and therefore, these approaches are not error-resistant. In the new target tracking paradigms, the sensor nodes that can detect the target are kept in the active state, and the rest of the sensors go to the passive state to save power. In order to track the target continuously, a group of sensors must be activated before the target reaches them. This group of sensors is determined according to the speed and direction of the target. Therefore, most of the researches in the field of target tracking have been done to obtain a suitable algorithm for the optimal selection of this group of sensors. In this research, by using the near-optimal guess of this group of sensors, they minimize the amount of information exchange between sensors, and therefore the telecommunication subsystem, which is the main source of power consumption of sensors, becomes less active, and as a result, energy consumption is significantly reduced. But another group of target tracking algorithms have focused on the power consumption within a sensor considering that not taking into account the reduction of energy consumption in the sensing and processing subsystems of the sensors keeps us away from the possibility of further reducing the power consumption of the network. In these algorithms, the power consumption of sensor subsystems is reduced by providing algorithms whose goal is to track the target with minimum processing overhead and the appropriate sampling method with the appropriate timing and frequency of activation of the sensor subsystem [3]. These approaches include message-based, tree-based, prediction-based, and clustering-based approaches.