Energy Efficient Coverage with Variable Sensing ... - Semantic Scholar
Energy Efficient Coverage with Variable Sensing Radii in Wireless Sensor Networks Jiong Wang
Sirisha Medidi∗
School of Electrical Engineering and Computer science Washington State University, Pullman, Washington 99164-2752 Email: {wangj,smedidi}@eecs.wsu.edu
Abstract— Wireless Sensor Networks (WSN) consist of spatially distributed autonomous devices using sensors to cooperatively monitor physical or environmental conditions. Generally, sensor nodes are resource-constraint in terms of size, power and bandwidth, therefore energy-efficient design becomes one of the fundamental goals in WSN. To provide reliable service in surveillance as well as energy efficiency, significant attention has recently been devoted to dynamic coverage algorithms based on sensors which are capable of adjusting its transmission and sensing range. In this paper, we propose two local sensing radii optimization schemes based on one-hop approximation of Delaunay Triangulation to minimize the energy consumption and extend the lifetime of networks. Furthermore, we prove that our approximation of Delaunay Triangulation guarantees complete coverage and generates the same radii assignment as traditional Delaunay Triangulation. Our ns-2 based simulation shows performance improvements in many aspects such as coverage ratio, sensing energy consumption, node failure rate, and lifetime of networks. Key words: Wireless Sensor Networks, coverage, Delaunay Triangulation
energy consumption according to WINS Rockwell seismic sensor [2], it becomes more critical for energy-efficient design of WSN because surveillance/monitoring requires a continuous coverage while data communication is only periodic and eventdriven. In the question of energy-efficient coverage with variable sensing radii, centralized optimization algorithms are proposed in [4], [6] based on Linear Programming. Due to the resource constraints and wide-spread deployment of sensor networks, a distributed algorithm with global optimality and minimal overhead is preferred. To address this issue, we propose two Delaunay-Triangulation-based heuristics for complete coverage and energy-efficiency. Furthermore, we present a simple distributed algorithm to approximate the Delaunay Triangulation with one-hop neighborhood information and prove that it can achieve the same results as the traditional one for sensing radii assignments. The goal of our technique includes: •
I. I NTRODUCTION Wireless Sensor Networks (WSN) are collections of a large number of minuscule devices with low-power processing, sensing, and radio communication capabilities. Generally deployed in hostile environments for military surveillance, emergency response, and habitat monitoring, WSN is required to provide long-term reliable services [1], and hence presents non-trivial challenges due to its resource constraints in term of size, power, bandwidth, etc [19]. In WSN, coverage and connectivity are foremost for reliable surveillance and data gathering. For wireless communication, the energy consumption is proportional to rn where r is the transmission range and n is the path loss exponent, hence the communication energy consumption can be minimized by varying the transmission range while keeping the network connected. For sensing and monitoring, having a large sensing radius increases energy consumption because more sophisticated filtering and signal processing methods are required to improve the signal-to-noise ratio and achieve the desired confidence level [27]. Therefore, with adjustable sensing radii, the sensing energy can be optimized by reducing the redundant coverage. Additionally, although the sensing energy consumption is about 102 times less than communication ∗ This research was supported in part by National Science Foundation grants CNS-0454416 and ECS-0646215.
• • •
A distributed sensing radii assignment scheme which requires low overhead and a local view of the topology, complete coverage for reliable surveillance, minimal energy consumption in sensing, and energy balancing for extending the lifetime of networks.
The rest of the paper is organized as follows. In Section 2, we provide a summary of related work about coverage and connectivity in sensor networks. In Section 3, we introduce an algorithm for localized approximation of Delaunay Triangulation and discuss two heuristics of sensing radii assignments for minimal sensing energy consumption and energybalancing. The proof of complete coverage using our heuristics and the correctness of one-hop approximation of Delaunay Triangulation is presented in Section 4. Finally, we evaluate the performance of our technique based on the ns-2 simulator in Section 5 and conclusions are shown in Section 6. II. R ELATED W ORK Topology control for energy-efficiency and connectivity is fundamental for wireless ad hoc and sensor networks. In Span [3], each node maintains its two-hop topology information, and redundant nodes are turned off according to the connectivity among the neighbors. GAF (Geographical Adaptive Fidelity) [23] is a grid based approach in which each grid cell has a cluster head elected for communication while the rest of the nodes are turned off to save energy. In AFECA (Adaptive
Fidelity Energy Conservation Algorithm) [24], a node dutycycle schedule is designed based on the number of neighbors for each node. As for ASCENT (Adaptive Self-Configuring sEnsor Networks Topologies) [5], each node can independently decide its sleeping period according to the number of active nodes and per-link data loss rate through data traffic. The coverage issue is unique to sensor networks. It can be considered as “How well an area of interest can be monitored” [21]. To provide complete coverage while turning off redundant nodes for energy efficiency, Slijepcevic and Potkonjak [17] propose an NP-complete problem called Set K-Cover, where mutually exclusive sets of sensor nodes are chosen and each set has complete coverage of the area. Tian and Georganas [18] propose a node scheduling scheme in which a node will be active only if its “sponsored area” has been covered by neighboring nodes. TGim [8] extended [18] by considering the realistic signal propagation model. Hsin and Liu [9] propose a low duty-cycle scheduling scheme based on random sleep and coordinated sleep. Additionally, a problem of k-coverage is discussed in [10], where authors find a sufficient and necessary condition to achieve multiple levels of sensing reliability; and authors in [25] address the similar problem using a round scheduling scheme. Coverage has been approached differently by Meguerdichian et. al. [15], where they try to find the maximal breach path and maximal support path based on Voronoi Diagram. Additionally, the related problem of finding a minimal exposure path and least energy consumption path have also been addressed in [14] and in [22] respectively. An integrated approach to achieve complete coverage and connectivity is essential for sensor networks. PEAS (Probing Environment and Adaptive Sleeping) [26] first addresses this combined problem by a probabilistic probing, however complete coverage is not guaranteed. Shakkottai, Srikant, and Shroff [16] provide theoretical bounds on the probability that both coverage and connectivity are achieved given unreliable sensor nodes deployed on a grid. Furthermore, an important result about the relationship between coverage and connectivity has been discussed by Wang et. al. [20] and Zhang and Hou [11]. Both show that connectivity can be implied when there is complete coverage and the ratio of transmission radius to sensing radius is larger than 2. Recently, sensors with variable sensing radii were used for energy efficient coverage. Wu and Yang [21] propose a coverage algorithm using sensors with maximum, medium, and small sensing radii according to the network topology. Cardei et. al. [4] propose a model with continuous sensing radii and present an algorithm to find mutually exclusive sensor covers and optimal sensing radii assignment. A similar problem has also been addressed in [6] and [27] using Linear Programming and Voronoi-Diagram-based heuristics. Our work presented here differs from aforementioned approaches in several respects. First, we emphasize simplicity and scalability in our distributed technique, because sensor networks are generally resource-constrained and are large-scale deployments. Second, we consider not only total energy consumption, but also energy
Fig. 1.
Sensing Radius Assignments with Different Triangulations
balancing in order to extend the lifetime of networks. Third, since we utilize all available sensors, our technique does not require dense sensor deployment as in [21] and can be used jointly with other scheduling schemes such as in [17], [18] to further reduce the redundant coverage within each mutually exclusive set of sensor nodes. III. E NERGY E FFICIENT C OVERAGE BASED ON D ELAUNAY T RIANGULATION Coverage and connectivity are two fundamental issues in sensor networks. We assume the transmission range is at least two times larger than the maximum sensing range as discussed in [11], [20]. Based on this result, we are focusing on optimizing sensing coverage for reliable surveillance and energy-efficiency, and providing a distributed algorithm with low communication and computational overhead. A. A local optimization scheme based on Delaunay Triangulation In order to optimize sensing coverage locally, our strategy triangulates sensor nodes into a planar graph, where each triangle is a Responsible Area that needs to be covered by its three adjacent sensors. The sensing radius is first locally optimized by three adjacent vertices of each Responsible Area. Then, for each sensor, it chooses the maximum one that can satisfy all of its adjacent Responsible Areas. To achieve the global optimality of sensing coverage with local decisions, triangulation is essential. For example, in Fig. 1.(a), because the triangles are equilateral, each sensor will have similar optimal radii for all of its adjacent Responsible Areas, and hence sensing radii can be assigned efficiently with minimum overlapping; in Fig. 1.(b), with the same topology but different triangulation, the local optimization introduces significant overlapping coverage such as among nodes A, C, and B. Zhang and Hou [11] proved that an √ equilateral triangulation with each edge as 3rs has the minimal redundant coverage, where sensors have a uniform sensing radius as rs . However, with random sensor deployment and variable sensing radii, the ideal equilateral triangulation with the specific distance between two adjacent nodes may not exist. Therefore, we choose Delaunay Triangulation in order to achieve a better global optimality based on local decisions. The Delaunay Triangulation is the dual of the Voronoi Diagram. Given a set of points S on a two dimensional space
Sirisha Medidi∗
School of Electrical Engineering and Computer science Washington State University, Pullman, Washington 99164-2752 Email: {wangj,smedidi}@eecs.wsu.edu
Abstract— Wireless Sensor Networks (WSN) consist of spatially distributed autonomous devices using sensors to cooperatively monitor physical or environmental conditions. Generally, sensor nodes are resource-constraint in terms of size, power and bandwidth, therefore energy-efficient design becomes one of the fundamental goals in WSN. To provide reliable service in surveillance as well as energy efficiency, significant attention has recently been devoted to dynamic coverage algorithms based on sensors which are capable of adjusting its transmission and sensing range. In this paper, we propose two local sensing radii optimization schemes based on one-hop approximation of Delaunay Triangulation to minimize the energy consumption and extend the lifetime of networks. Furthermore, we prove that our approximation of Delaunay Triangulation guarantees complete coverage and generates the same radii assignment as traditional Delaunay Triangulation. Our ns-2 based simulation shows performance improvements in many aspects such as coverage ratio, sensing energy consumption, node failure rate, and lifetime of networks. Key words: Wireless Sensor Networks, coverage, Delaunay Triangulation
energy consumption according to WINS Rockwell seismic sensor [2], it becomes more critical for energy-efficient design of WSN because surveillance/monitoring requires a continuous coverage while data communication is only periodic and eventdriven. In the question of energy-efficient coverage with variable sensing radii, centralized optimization algorithms are proposed in [4], [6] based on Linear Programming. Due to the resource constraints and wide-spread deployment of sensor networks, a distributed algorithm with global optimality and minimal overhead is preferred. To address this issue, we propose two Delaunay-Triangulation-based heuristics for complete coverage and energy-efficiency. Furthermore, we present a simple distributed algorithm to approximate the Delaunay Triangulation with one-hop neighborhood information and prove that it can achieve the same results as the traditional one for sensing radii assignments. The goal of our technique includes: •
I. I NTRODUCTION Wireless Sensor Networks (WSN) are collections of a large number of minuscule devices with low-power processing, sensing, and radio communication capabilities. Generally deployed in hostile environments for military surveillance, emergency response, and habitat monitoring, WSN is required to provide long-term reliable services [1], and hence presents non-trivial challenges due to its resource constraints in term of size, power, bandwidth, etc [19]. In WSN, coverage and connectivity are foremost for reliable surveillance and data gathering. For wireless communication, the energy consumption is proportional to rn where r is the transmission range and n is the path loss exponent, hence the communication energy consumption can be minimized by varying the transmission range while keeping the network connected. For sensing and monitoring, having a large sensing radius increases energy consumption because more sophisticated filtering and signal processing methods are required to improve the signal-to-noise ratio and achieve the desired confidence level [27]. Therefore, with adjustable sensing radii, the sensing energy can be optimized by reducing the redundant coverage. Additionally, although the sensing energy consumption is about 102 times less than communication ∗ This research was supported in part by National Science Foundation grants CNS-0454416 and ECS-0646215.
• • •
A distributed sensing radii assignment scheme which requires low overhead and a local view of the topology, complete coverage for reliable surveillance, minimal energy consumption in sensing, and energy balancing for extending the lifetime of networks.
The rest of the paper is organized as follows. In Section 2, we provide a summary of related work about coverage and connectivity in sensor networks. In Section 3, we introduce an algorithm for localized approximation of Delaunay Triangulation and discuss two heuristics of sensing radii assignments for minimal sensing energy consumption and energybalancing. The proof of complete coverage using our heuristics and the correctness of one-hop approximation of Delaunay Triangulation is presented in Section 4. Finally, we evaluate the performance of our technique based on the ns-2 simulator in Section 5 and conclusions are shown in Section 6. II. R ELATED W ORK Topology control for energy-efficiency and connectivity is fundamental for wireless ad hoc and sensor networks. In Span [3], each node maintains its two-hop topology information, and redundant nodes are turned off according to the connectivity among the neighbors. GAF (Geographical Adaptive Fidelity) [23] is a grid based approach in which each grid cell has a cluster head elected for communication while the rest of the nodes are turned off to save energy. In AFECA (Adaptive
Fidelity Energy Conservation Algorithm) [24], a node dutycycle schedule is designed based on the number of neighbors for each node. As for ASCENT (Adaptive Self-Configuring sEnsor Networks Topologies) [5], each node can independently decide its sleeping period according to the number of active nodes and per-link data loss rate through data traffic. The coverage issue is unique to sensor networks. It can be considered as “How well an area of interest can be monitored” [21]. To provide complete coverage while turning off redundant nodes for energy efficiency, Slijepcevic and Potkonjak [17] propose an NP-complete problem called Set K-Cover, where mutually exclusive sets of sensor nodes are chosen and each set has complete coverage of the area. Tian and Georganas [18] propose a node scheduling scheme in which a node will be active only if its “sponsored area” has been covered by neighboring nodes. TGim [8] extended [18] by considering the realistic signal propagation model. Hsin and Liu [9] propose a low duty-cycle scheduling scheme based on random sleep and coordinated sleep. Additionally, a problem of k-coverage is discussed in [10], where authors find a sufficient and necessary condition to achieve multiple levels of sensing reliability; and authors in [25] address the similar problem using a round scheduling scheme. Coverage has been approached differently by Meguerdichian et. al. [15], where they try to find the maximal breach path and maximal support path based on Voronoi Diagram. Additionally, the related problem of finding a minimal exposure path and least energy consumption path have also been addressed in [14] and in [22] respectively. An integrated approach to achieve complete coverage and connectivity is essential for sensor networks. PEAS (Probing Environment and Adaptive Sleeping) [26] first addresses this combined problem by a probabilistic probing, however complete coverage is not guaranteed. Shakkottai, Srikant, and Shroff [16] provide theoretical bounds on the probability that both coverage and connectivity are achieved given unreliable sensor nodes deployed on a grid. Furthermore, an important result about the relationship between coverage and connectivity has been discussed by Wang et. al. [20] and Zhang and Hou [11]. Both show that connectivity can be implied when there is complete coverage and the ratio of transmission radius to sensing radius is larger than 2. Recently, sensors with variable sensing radii were used for energy efficient coverage. Wu and Yang [21] propose a coverage algorithm using sensors with maximum, medium, and small sensing radii according to the network topology. Cardei et. al. [4] propose a model with continuous sensing radii and present an algorithm to find mutually exclusive sensor covers and optimal sensing radii assignment. A similar problem has also been addressed in [6] and [27] using Linear Programming and Voronoi-Diagram-based heuristics. Our work presented here differs from aforementioned approaches in several respects. First, we emphasize simplicity and scalability in our distributed technique, because sensor networks are generally resource-constrained and are large-scale deployments. Second, we consider not only total energy consumption, but also energy
Fig. 1.
Sensing Radius Assignments with Different Triangulations
balancing in order to extend the lifetime of networks. Third, since we utilize all available sensors, our technique does not require dense sensor deployment as in [21] and can be used jointly with other scheduling schemes such as in [17], [18] to further reduce the redundant coverage within each mutually exclusive set of sensor nodes. III. E NERGY E FFICIENT C OVERAGE BASED ON D ELAUNAY T RIANGULATION Coverage and connectivity are two fundamental issues in sensor networks. We assume the transmission range is at least two times larger than the maximum sensing range as discussed in [11], [20]. Based on this result, we are focusing on optimizing sensing coverage for reliable surveillance and energy-efficiency, and providing a distributed algorithm with low communication and computational overhead. A. A local optimization scheme based on Delaunay Triangulation In order to optimize sensing coverage locally, our strategy triangulates sensor nodes into a planar graph, where each triangle is a Responsible Area that needs to be covered by its three adjacent sensors. The sensing radius is first locally optimized by three adjacent vertices of each Responsible Area. Then, for each sensor, it chooses the maximum one that can satisfy all of its adjacent Responsible Areas. To achieve the global optimality of sensing coverage with local decisions, triangulation is essential. For example, in Fig. 1.(a), because the triangles are equilateral, each sensor will have similar optimal radii for all of its adjacent Responsible Areas, and hence sensing radii can be assigned efficiently with minimum overlapping; in Fig. 1.(b), with the same topology but different triangulation, the local optimization introduces significant overlapping coverage such as among nodes A, C, and B. Zhang and Hou [11] proved that an √ equilateral triangulation with each edge as 3rs has the minimal redundant coverage, where sensors have a uniform sensing radius as rs . However, with random sensor deployment and variable sensing radii, the ideal equilateral triangulation with the specific distance between two adjacent nodes may not exist. Therefore, we choose Delaunay Triangulation in order to achieve a better global optimality based on local decisions. The Delaunay Triangulation is the dual of the Voronoi Diagram. Given a set of points S on a two dimensional space
