Fusion-based surveillance WSN deployment using Dempster–Shafer theory
Introduction
A Wireless Sensor Network (WSN) is a network of wirelessly connected small-scale, low-power, low-cost devices called sensor nodes (or sensors). These tiny sensors are equipped with sensing, computation, storage, communication, and power.
Critical applications such as military target detection impose stringent requirements for event detection accuracy such as a low false alarm rate coupled with a high detection rate. To detect a target, the sensors have to make local observations of their surrounding environment and collaborate to produce a global decision that reflects the status of the Region of Interest (RoI). It is well known that collaboration among sensors improve the sensing quality by jointly considering noisy measurements of multiple unreliable sensors. For example, He et al. (2006) show that the false alarm rate of a real-world Mica2 WSN can be reduced from 60% (when sensors decide independently) to near zero by adopting a fusion scheme.
The deployment is a mandatory and critical step in the process of developing WSNs solutions for real-life applications. Even if many optimizations can be done once the WSN is deployed to enhance its performance, we believe that the most important optimization steps are those performed at the design step in order to build the best possible topology that meets specific user requirements.
Deploying WSNs into the real-world could be a very challenging task. Actually, in addition to the intrinsic properties of WSNs, many challenges stem from the close interactions between the WSN and the physical environment. Examples of such challenges include:
- •
How to consider uncertainty related to sensors readings?
- •
How to consider sensor reliability?
- •
How to combine sensory data from multiple sensors that can vary in their reliability?
- •
How to consider harsh deployment environments?
By considering such fundamental challenges in the early design steps of the WSN, one can avoid unpleasant surprises and save effort, time, and money. In an attempt to handle the challenges above, the present paper brings the following contributions:
- 1.
A formal definition of an evidence-based sensing model is provided on the basis of Dempster–Shafer theory (Smets and Kennes, 1994). The proposed model captures several characteristics of real-world applications.
- 2.
To build belief functions from raw sensing data, a two-step method is described.
- 3.
The fusion-based uncertainty-aware sensor networks deployment (denoted by FUSD) problem is formulated as a binary non-linear and non-convex optimization problem. Several characteristics of real-world applications including sensor׳s spatial distribution, uncertain sensor measurements, challenging environments, and sensor reliability can be captured in this framework.
- 4.
An efficient heuristic for the FUSD problem is developed. The effectiveness and efficiency of the proposed approach are validated through extensive experiments carried out on both synthetic dataset, data traces collected from a military vehicle detection experiment, as well as a lab testbed.
The remainder of this paper is organized as follows. Section 2 reviews related work. The background of the present work is described in Section 3. Section 4 defines our evidence fusion model. In Section 5, the FUSD problem is formalized. Section 6 addresses our sensor placement algorithm. The evaluation of our proposed approach via numerical experiments and trace-driven simulations is presented in Section 7. Section 8 discusses the results obtained by deploying an experimental testbed for motion detection. In Section 9, some possible extensions are presented. Section 10 concludes the paper and discusses some future directions for our work.
Section snippets
Related work
In our previous work (Senouci et al., 2015), we have investigated the belief functions theory (Smets and Kennes, 1994) to design a unified approach for the uncertainty-aware WSNs deterministic deployment. However, the devised approach considers only the rate of coverage and ignores completely the issue of false alarm. Hence, generated topologies do not guarantee the false alarm rate. In contrast to our previous work (Senouci et al., 2015), in this work we consider both false alarm and detection
Preliminaries
In this section, we describe the background of our work, which includes an overview of the Transferable Belief Model (TBM) (Smets and Kennes, 1994), a subjective interpretation of the Dempster–Shafer theory (Shafer, 1976).
The TBM is a model for representing the quantified beliefs held by an agent at a given time on a given frame of discernment. It involves the same concepts as those considered by the Bayesian model, except it does not rely on probabilistic quantification, but on a more general
Evidence fusion model
In this section, we describe our first contribution, which includes a formal definition of the evidence-based sensing model and a description of how to build evidence from raw sensing data.
Sensor placement problem
The previous section formalizes the evidence-based sensing model and describes how to build evidence from raw sensing data. In this section, we describe our second contribution that includes a formal definition of the fusion-based uncertainty-aware sensor networks deployment (FUSD) problem.
Sensor placement algorithm
The two combinatorial non-linear and non-convex optimization problems formalized in Section 5 are NP-hard (Ke et al., 2007). A non-linear programming solver can be applied to both FUSD problems as done in Chang et al. (2011) where the authors used the CSA solver (Wah et al., 2007). The limitations of this technique are as follows: first, only a local optimum is obtained, and second, there is a limitation on the number of constraints and hence the size of the RoI that can be handled.
In the
Simulation results
To evaluate the effectiveness and efficiency of the proposed approach, both simulations and experiments have been conducted. In this section, we present the numerical results and the trace-driven simulations.
Testbed deployment
To further quantify the real benefit of our approach, we develop ArduiNet, a real WSN testbed for motion detection comprising more than 30 sensor nodes. Our testbed is based on the Arduino (1991) platform, which is an open-source fast prototyping platform based on flexible, easy-to-use hardware and software.
Discussion
Our proposed approach can be extended in several ways. In this section, we discuss how correction mechanisms for weakening or reinforcing belief functions (Smets and Kennes, 1994, Elouedi et al., 2004, Mercier et al., 2006) can be included in our framework in order to handle deployment-related issues such as sensor reliability and challenging environments.
Conclusion
In the present work, a unified approach for robust deployment of fusion-based WSNs is proposed. Our approach is based on the Dempster–Shafer theory that captures several characteristics of real-world applications and allows an efficient management of the uncertainty related to sensor readings. This distinguishes our effort from existing sensor deployment approaches that are either based on ideal detection models or rely on simple fusion schemes. The effectiveness and efficiency of the proposed
References (32)
- et al.
Constructing consonant belief functions from sample data using confidence sets of pignistic probabilities
Int J Approx Reason
(2008) - et al.
Vehicle classification in distributed sensor networks
J Parallel Distrib Comput
(2004) - et al.
Wsns deployment framework based on the theory of belief functions
Comput Netw
(2015) - et al.
The transferable belief model
Artif Intell
(1994) Decision making in the TBMthe necessity of the pignistic transformation
Int J Approx Reason
(2005)- Ababnah A, Natarajan B. Optimal sensor deployment for value-fusion based detection. In: IEEE global telecommunications...
- Ababnah A, Natarajan B. Lqr formulation of sensor deployment for decision fusion based detection. In: IEEE global...
- Ababnah A. Sensor deployment in detection networks—a control theoretic approach [PhD thesis]. Kansas State University;...
- Appriou A. Formulation et traitement de l׳incertain en analyse multisenseurs. In: GRETSI, 1991. p....
- Arduino,...
Sensor placement algorithms for fusion-based surveillance networks
IEEE Trans Parallel Distrib Syst
Sensor deployment strategy for detection of targets traversing a region
Mob Netw Appl
Fault tolerance in collaborative sensor networks for target detection
IEEE Trans Comput
Theoretical and numerical constraint-handling techniques used with evolutionary algorithmsa survey of the state of the art
Comput Methods Appl Mech Eng
Assessing sensor reliability for multisensor data fusion within the transferable belief model
IEEE Trans Syst Man Cybern
Cited by (25)
Integration of type-2 fuzzy logic and Dempster–Shafer Theory for accurate inference of IoT-based health-care system
2021, Future Generation Computer SystemsA robust uncertainty-aware cluster-based deployment approach for WSNs: Coverage, connectivity, and lifespan
2019, Journal of Network and Computer ApplicationsCitation Excerpt :This has incited the development of new proposals (Ababnah and Natarajan, 2009, 2010; Chang et al., 2011; Senouci et al., 2016) that adopt data fusion techniques. These proposals consider tiered architecture and use different tools such as optimal control theory (Ababnah and Natarajan, 2009, 2010), probability theory (Chang et al., 2011; Yang et al., 2013), fuzzy theory (Collotta et al., 2017), and Dempster-Shafer evidence theory (Senouci et al., 2016). Most of these approaches enhance the performance of the WSN by taking advantage of the collaboration between sensors.
Bayesian fuzzy hypothesis test in wireless sensor networks with noise uncertainty
2019, Applied Soft Computing JournalCitation Excerpt :In recent years, wireless sensor networks (WSNs) have received much attention due to their numerous applications such as environmental monitoring and surveillance [1,2], precision agriculture [3,4], smart home [5,6], and health care [7,8].
In-network context inference in IoT sensory environment for efficient network resource utilization
2019, Journal of Network and Computer ApplicationsCitation Excerpt :These LLCs are further transmitted to the location of inference to infer the high level contexts (HLCs), i.e., ‘occurrence of fire’ in this scenario. Such applications can be extrapolated in different domains including agriculture (Kumar and Ilango, 2018), surveillance (Senouci et al., 2016), civil infrastructure monitoring (Alves et al., 2017), and health-care (Majumder et al., 2017; Abreu et al., 2017) etc. Recent advancements in this area suggest the HLCs inference mechanism to be executed at the gateway node stationed outside the network and hence known as ‘out-network’ context inference mechanisms (Gu et al., 2005; Salber et al., 1999).
Performance evaluation of advanced manufacturing technologies: A De novo approach
2017, Computers and Industrial EngineeringCitation Excerpt :An approach using the Dempster-Shafer theory was proposed by Wang and Jing (Tiffany) (2012) for ranking and selection of wireless network in a very much complicated scenario. A fusion based Wireless Sensor Networks (WSN) surveillance application was deployed by Senouci, Mellouk, Aitsaadi, and Oukhellou (2016) to achieve high detection rate coupled with a low false alarm rate in WSN. Gruyer, Demmel, Magnier, and Belaroussi (2016) applied the Dempster-Shafer theory in their Multi-Hypotheses Tracking (MHT) approach that allowed for solving ambiguities arising with the methods of associating targets and tracks within a highly volatile vehicular environment.