^{1}

^{*}

^{2}

^{3}

^{1}

^{1}

This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution license (http://creativecommons.org/licenses/by/3.0/).

In this paper, we focus on the problem of tracking a moving target in a wireless sensor network (WSN), in which the capability of each sensor is relatively limited, to construct large-scale WSNs at a reasonable cost. We first propose two simple multi-point surveillance schemes for a moving target in a WSN and demonstrate that one of the schemes can achieve high tracking probability with low power consumption. In addition, we examine the relationship between tracking probability and sensor density through simulations, and then derive an approximate expression representing the relationship. As the results, we present guidelines for sensor density, tracking probability, and the number of monitoring sensors that satisfy a variety of application demands.

The recent development of low-cost sensors [

However, there are a number of issues that need to be addressed for moving target tracking in WSNs. First, sensor nodes are subject to severe resource constraints such as low processing speed, small memory size, limited battery power, and a narrow and unreliable communication bandwidth. Most existing schemes for target tracking assume that the capability of individual sensors is very high, such that a sensor can determine the distances and/or directions of nearby sensors (and, thus, to a target) [

In a WSN consisting of a number of low-performance sensors, the information obtained from one sensor node may be of little or no use by itself. Therefore, by combining the data from neighboring sensors, also referred to as sensor fusion, more target information can be obtained, and high-resolution tracking can be achieved [

In the present paper, we discuss two simple and flexible target-tracking schemes for a distributed WSN: a primitive and a sophisticated schemes. We propose the use of the sophisticated scheme. Unlike those in previous studies, this scheme can select the appropriate number of monitoring sensors in a distributed manner with consideration of monitoring resolution and power consumption, and then monitor a target using these multiple low-performance sensors (multi-point surveillance). Moreover, the proposed scheme changes the monitoring sensors in response to the movement of the target. Through simulation experiments, we demonstrate that the proposed scheme can achieve (1) high tracking probability, (2) low power consumption regardless of the target's speed, and (3) flexibility with respect to the change in the number of required monitoring sensors.

Furthermore, when a WSN is actually constructed, a wide range of configurations, such as the state transition period in each sensor and the sensor density, should be evaluated for both the required tracking probability and power consumption, which differ according to the application. To the best of our knowledge, no existing studies mention a procedure to construct the WSN for multi-point surveillance of a moving target. Therefore, we examine the relationship between the tracking probability achieved by the proposed scheme and the sensor density. Then, through simulation experiments, we show the impact of the number of monitoring sensors on the performance. From this relationship, we derive an approximate expression for easy estimation of tracking probability. Finally, through the approximation, we demonstrate the deployment design of a WSN for multi-point surveillance of a moving target.

This paper is organized as follows. Section 2. describes related researches. Section 3. outlines our proposed multi-point surveillance scheme. In Section 4., we examine the performance of our proposed scheme in terms of tracking probability and energy consumption. Section 5. discusses the deployment design of a WSN based on the effects of both sensor density and the number of monitoring sensors. Concluding remarks are presented in Section 6.

Research on WSNs for target tracking has recently attracted considerable attention. Some studies, such as [

As stated in Section 1., (1)

In contrast, the next three schemes (Single-sensor Monitoring, Dynamic Clustering, and Prediction-based) can achieve good tracking probability based on the assumption that all of the sensors in the WSN are managed in a centralized manner (rather than in a distributed manner), meaning that a coordination node, such as a base station or cluster head, controls all of the sensors in its area. However, it is difficult to deploy these powerful nodes in the WSN and to control them in a centralized manner. Moreover, these existing schemes generally assume that the capability of individual sensors is very high, for example: a sensor can detect the location of the target and can issue directions (wake up/shut down) to nearby sensors based on the prediction of the target's movement.

Based on these results, we believe that these centralized schemes do not provide scalability to the WSN. Thus, in the present paper, we focus on a distributed type of scheme in which an individual simple sensor decides its own behavior based on localized and limited information, thereby achieving multi-point surveillance with the appropriate number of monitoring sensors. Recently, several schemes providing the distributed coordination among multiple sensors have been proposed to track a moving target. In [

In the DELTA, an event is detected or tracked by dynamically established groups. The group is managed by group leader, which is dynamically decided in a distributed manner and behaves as the dynamic cluster head. Ref. [

Through these considerations, to the best of our knowledge, none of the existing schemes satisfies all the requirements. That is, no existing scheme selects the appropriate number of monitoring sensors with low-performance in a distributed manner while achieving good performance in terms of tracking probability (monitoring resolution) and power consumption.

As described in Section 1., multi-point surveillance with quick switching of monitoring sensors is essential for obtaining detailed information and achieving continuous monitoring of a moving target. Here, we propose two simple and flexible multi-point surveillance schemes with exactly

A sensor node typically has two functions:

A sensor can sense whether or not the target exists in its sensing area of radius _{s}

The sensor can broadcast messages in the communication area of radius _{c}_{s}

Today's sensor nodes are equipped with a state transition function for power savings. In the present paper, to achieve uniform power savings among all sensors, we assume that each sensor can autonomously and individually change its status at time intervals of

In this section, we compare our two schemes: primitive and sophisticated. In our proposed schemes, since the low-performance sensors are assumed with considering the actual sensor products, we design the simplest algorithm for deciding the monitoring sensors by reducing the load of message exchange as much as possible. More specifically, we firstly assume that the sensors exchange messages with each other in a broadcast manner, thus they don't require the neighboring sensors' address information. Next, to reduce the complexity and the energy consumption of the proposed schemes, only two types of messages, i.e., ALERT and DETECT, are defined. To obtain detailed information of a moving target by low-performance sensors, the proposed sophisticated scheme reliably and quickly decides which multiple monitoring sensors to use by user demand and/or applications based on the ALERT and DETECT message exchanges.

The primitive scheme focuses only on the selection of monitoring sensors; that is, the movement of the target is not considered. The primitive scheme consists of two phases:

In this way, the primitive scheme selects

We enhance the primitive multi-point surveillance scheme to address these problems. By using the flow diagram shown in

Each

Each

In this manner, the sophisticated scheme attempts to keep an individual sensor's state transition sequence unchanged (

Finally, we modify the random delay time, that is, the function of message transmission control. For access to a wireless channel, some types of MAC layer protocols, such as CSMA/CA of IEEE 802.15.4 and CSMA of TinyOS, have been employed in recent sensor products. Moreover, some enhanced protocols have been developed. Therefore, the proposed scheme should be designed to work well with any MAC protocol. Thus, to avoid collisions with other messages, each of the sensors calculates the random delay based on the slot time with consideration of the transmission time.

The random delay time is calculated as follows:

In

We evaluated the proposed multi-point surveillance schemes in terms of tracking probability and power consumption to examine the effect of the moving speed of the target.

The simulation parameters are listed in

Here, we consider the maximum number of monitoring sensors that are necessary to detect and track the target with exactly

Therefore, to track the target with

In our simulation, since we assumed that the sojourn times of each state are the same (

In the present paper, we investigate the effectiveness of the proposed scheme for tracking probability and power consumption. The tracking probability is examined by two performance measures: average tracking probability _{t}_{h}_{s}_{c}

The ratio of the interval time of ALERT messages to the monitoring time of the same _{h}_{h}_{h}_{s}_{i}_{i}_{c}_{j}_{j}

In this section, we evaluate the performance of the proposed schemes (primitive scheme and sophisticated scheme) through simulation experiments. In the present paper, as we would like to investigate the fundamental performance of the proposed scheme, that is, realistic wireless effects such as fading, shadowing, and interference from other devises are not considered at all, we develop a new simple simulator from the scratch to achieve this. Our main concern is enabling the proposed schemes to provide high tracking probability with low power consumption, even when the target moves at high speed. First, we assume

Next, we show the handover index (_{h}_{h}_{h}_{h}

_{h}

Next, we compare the energy consumption in the primitive and sophisticated schemes. In the present paper, we classify the energy consumption into two parts (state and message exchange) and calculate _{s}_{c}_{i}_{j}

_{s}_{t}_{s}_{s}

In the sophisticated scheme, _{s}_{t}

So far, we assumed the number of required monitoring sensors (

Moreover, comparing _{h}

In the previous section, we examined the tracking probability and power consumption under our proposed schemes (primitive and sophisticated schemes). Through simulations, we showed that the sophisticated scheme, in response to the target movement, can achieve (1) high tracking probability, (2) low power consumption, and (3) quick decisions of the monitoring sensors. Although the achievable tracking probability depends dynamically on the sensor density, we have not examined the impact of the density. Therefore, in this section, we first examine the effects on the tracking probability due to (i) sensor density and (ii) number of monitoring sensors. Finally, we consider the deployment design of a WSN that satisfies the user requirements effectively. As shown later in the present paper, the sophisticated scheme becomes the proposed scheme.

This section describes the simulation model employed here to investigate the effects of the sensor density and number of monitoring sensors on the tracking probability by using our proposed scheme. The simulation model and parameters are almost the same as those shown in _{all}_{t}

Obviously, the change in the state transition period of a sensor has a strong influence on the relationship between the sensor density and tracking probability. Hereinafter, we assume a fixed state transition period, and then set the period of

Through simulation experiments, we first examine the relationship between the tracking probability and the sensor density when the number of monitoring sensors (

Here, the minimum sensor density that can satisfy the required tracking probability is examined in detail. Here, we target both the number of _{AL}_{AL}^{2}, i.e., density of _{AL}

Next, we consider the sensor density that can achieve the maximum tracking probability. Although the sensor density that can achieve convergent tracking probability in _{AL}

Next, we use nonlinear regression to derive the approximate numerical formula that represents the relationship shown in this section. We apply _{AL}

From

So far, we examined the relationship between the sensor density and the tracking probability when the number of monitoring sensors _{all}_{AL}

Next, we derive the factors

Finally, we discuss the relationship of the tracking probability and both the number of monitoring sensors and the number of deployment sensors. More specifically, we integrate the approximate formulas derived thus far in order to estimate the required sensor density with consideration of both the tracking probability and the number of monitoring sensors. From this, we discuss the deployment design of WSNs for multi-point surveillance.

We first discuss the integration of the formulas based on

Then, we discuss factor

From these results, we verify that the relationship between parameters, sensor density,

In this way, the number of monitoring sensors

In the present paper, we proposed a simple multi-point surveillance scheme for the moving target in WSNs. The proposed scheme can satisfy the following two requirements to achieve moving target tracking with high resolution: (I) multi-point surveillance with an appropriate number of sensors, and (II) dynamic and quick decision making of monitoring sensors in response to the movement of the target. In the present paper, we proposed two new multi-point surveillance schemes. Basically, a sensor that detects the target broadcasts the ALERT message, and other sensors that can receive this message and detect the target then send the DETECT message to monitor the target with the number of sensors required by the user.

We first proposed a simple distributed scheme, which we call the primitive scheme. However, three problems were found in the primitive scheme. Therefore, based on the primitive scheme, we proposed a second scheme, which we call the sophisticated scheme, to address these problems. Simulation results revealed that the sophisticated scheme can achieve high tracking probability with low power consumption in the state and message exchange even for a high-speed moving target under randomly deployed sensor networks, due to (a) message transmission control and (b) maintenance of the state transition cycle. Since the sophisticated scheme can provide flexibility with respect to the change in number of monitoring sensors, the sophisticated scheme proposed here is considered to be practical for general WSNs.

Then, we examined the relationships among (1) sensor density, (2) number of monitoring sensors, and (3) tracking probability under our proposed multi-point surveillance scheme (sophisticated scheme). Two approximate formulas that express the relationship of sensor density and number

We integrated the two formulas to estimate the appropriate sensor density for achieving both tracking probability and the number of monitoring sensors. As a result, we showed the deployment design of WSNs that can achieve distributed multi-point surveillance for a moving target.

Currently, we are conducting a detailed examination of the relationship with tracking probability. More specifically, we are investigating the relationship in terms of sensor capability (sensing/ communication), state transition period of a sensor, and moving speed of a target, in addition to the sensor density and number of monitoring sensors

This research was supported in part by the Japan Society for the Promotion of Science, Grant-in-Aid for Scientific Research (S)(No. 18100001).

Sensor model.

State transition model.

Monitoring sensor decision flow.

Simulation model.

Number of

Handover index, _{h}

Tracking probability, _{t}

Power consumption (state), _{s}

Power consumption (message exchange), _{c}

Impact of _{h}

Impact of _{t}

Effects of sensor density on tracking probability.

Comparison between simulations and approximate formula representing the relationship between sensor density and tracking probability.

Effect of

Comparison between the approximate formula and simulation results : Effect of the number of monitoring sensors.

Architecture and sensor's capability employed in existing target tracking schemes.

Management | Monitoring sensor | Sensor's capability | ||||
---|---|---|---|---|---|---|

Centralized | Single | High | ||||

Naive | × | × | × | ○ | ○ | × |

Periodic Monitoring | × | × | × | ○ | ○ | × |

Single-sensor Monitoring | ○ | × | ○ | × | × | ○ |

Dynamic Clustering | ○ | × | × | ○ | × | ○ |

Prediction-based | ○ | × | × | ○ | × | ○ |

Distributed Coordination | × | ○ | × | ○ | × | ○ (Δ [ |

Simulation parameters.

Simulation time | 600 [sec] (10 [min]) |

Simulation area | 6,000 [m] × 6,000 [m] (square area) |

Number of sensors in the WSN | |

Required number of sensors | |

Sensing Range | _{s} |

Communication Range | _{c} |

State Transition | |

Speed of target | 0 [m/s] - 10 [m/s] |

Message length | 10 [bytes] |

Time slot | 400 [ |

Slot range | Act_max = 30, Lis_max = 128 |

Bandwidth | 250 [Kb/s] |

Power consumption.

State | Message exchange | ||
---|---|---|---|

30 mW/sec | Transmit | 56.7 mW/message | |

21.6 mW/sec | Receive | 62.91 mW/message |

Simulation parameters.

Density of _{all} |
1.4 ∼ 25 [_{all}/^{2}] |

Density of _{AL} |
0.93 ∼ 16.67 [_{AL}/^{2}] |

Speed of target | 9.7 [m/sec] |

Factors calculated from nonlinear regression.

A | B | Decrease in B | |
---|---|---|---|

2 | 0.9137 | 0.7838 | – |

3 | 0.8407 | 0.5359 | 0.2479 |

4 | 0.8391 | 0.3855 | 0.1504 |

5 | 0.8269 | 0.2984 | 0.0871 |

Parameters of sensor density.

_{all}_{all} |
[0.5/1/2]-million, [1.4, 2.8, 5.6]^{2} |

Density of _{AL} |
[0.93, 1.85, 3.7]^{2} |

Estimated factors by nonlinear regression.

Sensor density [#/100 ^{2}] |
||
---|---|---|

0.93 | 1.2745 | 0.6290 |

1.85 | 1.2035 | 0.7504 |

3.70 | 1.1315 | 0.8616 |