Deploying multiple interconnected gateways in heterogeneous wireless sensor networks: An optimization approach
Introduction
Wireless sensor networks (WSN) have recently emerged as an ideal solution to a large number of applications where the goal is collecting measurements of a physical parameters (temperature, humidity, light intensity, etc.) or detecting events in the covered area (intrusion, wild fire, etc.) [3]. Sensors are usually low-cost battery-operated devices geared with sensing, processing and communication functionalities. Obviously, the limited energy availability deeply impacts on the design of WSN and in particular on communication protocols running at different levels [4], [5], [6] as well as on network deployment [7], [8] and topology.
In many scenarios, data collected by sensors must be delivered to application servers for processing. These servers can be reached via a sink node that is connected to a local or geographical network through a communication interface with a different technology with respect to the WSN. Since the transmission range of sensor nodes is often much smaller than the area covered by the WSN, sensors cooperate to deliver information to the sink node through multi-hop paths.
Multi-hop information delivery to sink nodes requires multiple transmissions that consume energy in intermediate nodes and heavily limit lifetime of those nodes that are used more frequently. Even if many routing approaches have been proposed to balance the load among nodes and prolong the overall network lifetime, it is quite evident that nodes closer to the sink are anyway the most critical ones since they are required to relay all traffic generated by the other nodes and headed to the sink. For this reason, several methods have been proposed for optimizing the sink position in WSN design [15], [17].
However, a single sink node collecting all data from the network is not an efficient solution when dealing with large size WSNs, like e.g. in applications for monitoring natural areas. Indeed, as the network grows, the amount of information that must be delivered to the sink by surrounding nodes increases creating traffic and energy consumption bottlenecks. A promising alternative approach is based on the use of additional concentration nodes, gateways, that allow to spread the load and also to keep multi-hop paths within a reasonable length [10], [11], [12]. Also in this case gateway positioning is a key element for designing efficient networks. The entire set of sensor nodes need to be divided into independent subsets that send their traffic to a gateway, giving rise to a clustering problem.
Generally speaking, gateways are more powerful and expensive devices with respect to sensors and a reasonable objective is minimizing their number while matching constraints on energy efficiency and network quality [10], [12], [14]. However, this approach focuses on the WSN only neglecting the problem of interconnecting the gateways to the sink or more in general to the network where application servers are located. According to the specific application scenario, several solutions can be adopted for this problem, including the use of direct geographical links among the gateways (wired links, satellite links, etc.). However, this may be quite expensive and inefficient, since the traffic collected at each gateway is usually much smaller than the link capacity while its cost (interface, installation, and operation) often high.
For this reason, we consider an alternative network scenario where gateways are inter-connected through a wireless mesh network (WMN) and act as wireless routers (WRs) forwarding traffic along multi-hop paths toward a sink node that is the only one equipped with a geographical link (see Fig. 1). The resulting architecture is an heterogenous multi-hop network with different devices (sensors, WRs, and sink points) and wireless links (low-power low-rate links for sensors, and high rate links for mesh routers) that can be adapted to the specific application scenario using available technologies (like e.g. ZigBee, WiFi, WiMax, etc.)
In the aforementioned scenario, we investigate the joint problem of selecting/positioning the gateways (WRs) and designing the WMN that interconnects them. We provide an optimization framework that minimizes installation costs and maximizes the energy efficiency, while considering both multi-hop coverage and connectivity constraints. Coverage considers the availability of a multi-hop path between each sensor node and at least a gateway, while connectivity among gateways must be ensured by properly designing the WMN. The proposed optimization framework is independent on the specific routing strategy in the WSN; indeed, routes in the WSN (from sensors to gateways) are assumed to be given. This allows to use any of the routing mechanisms that have been proposed for WSNs [5] and to plug it into our optimization framework just modifying a separate software module that computes traffic loads in the WSN. We note here that routing could be easily optimized together with gateway positions by including further degrees of freedom (variables) in the optimization framework [10], [14], [26]. However, we believe that our approach is more practical since in real-life wireless sensor networks the network installer can hardly modify/optimize the routing schemes running in the specific WSNs technology.
We propose an integer linear programming (ILP) framework that allows to solve to optimality reasonable size networks and to evaluate the impact of system parameters, including the routing strategy, on the obtained solutions. We also propose an heuristic approach that allows to achieve good quality solutions (within 10% from the optimum in the considered scenarios) in short computing time, tackling also the optimization of large-scale networks.
Furthermore, we introduce an extension to the optimization framework for those scenarios where the amount of data to be managed by the WSN significantly varies during time. This is the case, for example, of WSNs for surveillance and reaction applications. In such cases, the network has two main operation points: in the surveillance state the sensors keep monitoring the area generating minimal coordination and detection traffic; then, the detection of a specific event triggers the reaction state where the WSN infrastructure is used to exchange more refined and higher data traffic on the event which has been detected (e.g., trajectory and speed, in case of a target tracking application) [1], [2].
To this extent, we consider the additional possibility of switching off installed WRs when they are not needed. Indeed, during the surveillance state, the load generated by sensors is much lower than load during the tracking phase, therefore, the collecting capacity of gateways (WRs) is in excess. Some of them can be turned off and traffic from sensor nodes rerouted, until a target comes in. This allows energy-aware strategies and wake-on-demand mechanisms to be applied also to the WMN.
The paper is organized as follows. We review and discuss previous works on this topic in Section 2. In Section 3 we introduce the generic problem of multiple interconnected gateway placement (MIG-P), and further propose four different mathematical programming ILP formulations which differ in their objectives. Section 4 provides numerical results to get insight on the characteristics of the optimal solutions obtained through the ILP framework. The heuristic algorithms proposed for tackling large size instances and results to assess their performance are presented in Section 5. We conclude the paper in Section 6.
Section snippets
Related work
The gateway placement problem has been extensively studied in literature [25]. The whole set of works can be divided into two classes. One is the class of Static Base Station Positioning problems where the WSN is deployed, gateways are installed and their positions never change for the entire network lifetime. The other class, the class of Dynamic Base Station Positioning problems is characterized by the gateway re-positioning during the network operation. The latter type of problems
The multiple interconnected gateway placement problem
We consider a heterogeneous network scenario where clouds of wireless sensor networks (WSNs) for data collection are connected to a sink node through a wireless mesh backbone of gateway devices. We will use gateways and WRs as synonyms throughout the paper. Each WR acts as a traffic concentration point for all the sensor nodes belonging to the corresponding WSN. In such network scenario, the number and the positions of the WRs obviously impact on the wireless sensor network lifetime. In fact,
Optimal gateway placement
We have tested the quality of the planned heterogeneous WSNs under two different routing algorithms: a classical min-hop shortest path, and an ideal energy-aware routing. In the latter case, routing paths are obtained solving a Minimum Cost Flow problem towards each candidate site, where the best routing paths from every sensor to a given candidate site are determined according to the following objective function: , where is the set of wireless links among sensor
Heuristic algorithms
It is easy to show that the MIG-P problem is NP-hard since it contains the Minimum Cardinality Set Covering problem as sub-problem. Even if we have shown that the problem can be solved for reasonable size instances, we have also designed heuristics to cope with large networks. We present first an algorithm for the basic formulation of the MIG-P problem, and then we extend it to solve the energy-aware formulation.
Our heuristic approach is based on a continuous relaxation of the corresponding ILP
Conclusion
In this paper we have proposed an optimization framework for the design of WSNs where gateways must be placed in the area and interconnected with the sink through the wireless links of a WMN. The resulting network architecture is an heterogeneous multi-hop wireless networks where short-range low-rate links interconnect sensor nodes, while long-rage high-rate links create a wireless backbone among gateways.
The objective is that of optimizing a tradeoff between the network installation cost (the
Acknowledgments
This work has been partially supported by research projects PRIN SESAME and FIRB INSYEME.
References (26)
- et al.
Wireless sensor networks: a survey
Computer Networks, Elsevier
(2002) - et al.
A survey on routing protocols for wireless sensor networks
Ad Hoc Networks, Elsevier
(2005) - et al.
Energy-efficient coverage problems in wireless ad-hoc sensor networks
Computer Communications, Elsevier
(2006) - et al.
Sink repositioning for enhanced performance in wireless sensor networks
Computer Networks, Elsevier
(2005) - et al.
Relay node placement in large scale wireless sensor networks
Computer Communications, Elsevier
(2006) - et al.
Strategies and techniques for node placement in wireless sensor networks: a survey
Ad Hoc Networks, Elsevier
(2008) - C. Gui, P. Mohapatra, Power conservation and quality of surveillance in target tracking sensor networks, ACM MobiCom...
- N.A. Vasanthi, S. Annadurai, Energy Saving Schedule for Target Tracking Sensor Networks to Maximize the Network...
- et al.
MAC protocols for wireless sensor networks: a survey
IEEE Communications Magazine
(2006) - L. Campelli, A. Capone, M. Cesana, E. Ekici, A Receiver Oriented MAC Protocol for Wireless Sensor Networks, IEEE MASS...