PaperlayerView exact revision on arXiv

Source-linked AI summary

A Survey of Multi-Objective Optimization in Wireless Sensor Networks: Metrics, Algorithms and Open Problems

Zesong Fei, Bin Li, Shaoshi Yang, Chengwen Xing, Hongbin Chen, Lajos Hanzo

arXiv:1609.04069v1cs.NI

TL;DR

WSNs must balance conflicting objectives such as energy, reliability, coverage, and lifetime, while existing MOO research spans varied methods and settings. This paper surveys WSN objectives, MOO techniques, and prior studies, then identifies open problems including multi-hop transmission and emerging network scenarios. Its supported outcome is a structured review of trade-offs, algorithms, research progress, and future directions.

  • Problem

    WSNs face conflicting trade-offs among data rate, reliability, energy consumption, coverage, and lifetime, while relatively few studies jointly address some of these objectives, especially in multi-hop settings.

  • Method

    The paper provides a tutorial and survey covering WSN optimization objectives, scalarization methods, metaheuristics, advanced techniques, and recent MOO studies.

  • Results

    The survey reviews diverse WSN trade-offs and classifies MOO approaches, including scalarization, evolutionary algorithms, swarm intelligence, and hybrid computational-intelligence methods.

  • Takeaways & Limitations

    Future MOO research directions include multi-hop transmission, dynamic deployment, secure multipath routing, 3D networks, CR-WSNs, and smart grids.

Abstract

from arXiv · show

Wireless sensor networks (WSNs) have attracted substantial research interest, especially in the context of performing monitoring and surveillance tasks. However, it is challenging to strike compelling trade-offs amongst the various conflicting optimization criteria, such as the network's energy dissipation, packet-loss rate, coverage and lifetime. This paper provides a tutorial and survey of recent research and development efforts addressing this issue by using the technique of multi-objective optimization (MOO). First, we provide an overview of the main optimization objectives used in WSNs. Then, we elaborate on various prevalent approaches conceived for MOO, such as the family of mathematical programming based scalarization methods, the family of heuristics/metaheuristics based optimization algorithms, and a variety of other advanced optimization techniques. Furthermore, we summarize a range of recent studies of MOO in the context of WSNs, which are intended to provide useful guidelines for researchers to understand the referenced literature. Finally, we discuss a range of open problems to be tackled by future research.

I. INTRODUCTION

WSNs require trade-offs among conflicting objectives, making multi-objective optimization more realistic than single-objective formulations. This survey introduces the relevant metrics and algorithms, reviews prior work, and identifies future research directions.

  • WSNs optimize conflicting criteria including energy efficiency, lifetime, delay, and other application-specific performance metrics.
  • Single-objective designs can overemphasize one metric, whereas MOO treats multiple objectives simultaneously under constraints.
  • Because objectives generally cannot reach their optima simultaneously, MOO seeks Pareto-optimal solutions whose improvements require sacrificing other objectives.
  • The survey covers scalarization methods, nature-inspired metaheuristics, and other prevalent MOO algorithms for WSN problems.
  • It reviews research progress according to trade-off metrics and discusses algorithm strengths, weaknesses, open challenges, and future trends.
  • The paper is organized around WSN models and objectives, MOO techniques, existing studies, and open problems.

A. System Model

WSNs use distributed, low-power sensor nodes that communicate wirelessly, often through clustered architectures and sink nodes. Their applications span monitoring and tracking across environmental, industrial, healthcare, urban, agricultural, ecological, and military settings.

  • A. System Model: WSNs comprise spatially distributed, low-cost, low-power, multifunctional sensor nodes communicating over short distances.Nodes typically include sensing, processing, radio, conversion, memory, and battery components.
  • A. System Model: Nodes are commonly organized into clusters, where cluster heads collect sensed data and forward it hop by hop to sink nodes or base stations.Large-scale deployments may use multiple clusters and base stations.
  • A. System Model: WSNs commonly use ZigBee under IEEE 802.15.4, but coexistence with other wireless systems can create co-channel interference.The shared unlicensed ISM band includes IEEE 802.11 and 802.15.1 systems.
  • B. Applications: WSN applications are broadly classified as monitoring or tracking, including environmental sensing, healthcare, agriculture, industrial control, transportation, and military surveillance.Examples include forest-fire detection, patient monitoring, irrigation management, equipment monitoring, and troop tracking.
  • B. Applications: Newer application scenarios include the Internet of Things, cyber-physical systems, and smart grids, which may support multiple concurrent applications on one WSN.

C. MOO Metrics

WSN optimization must account for multiple QoS and resource objectives that commonly conflict. Coverage measures sensing over a region, while connectivity and coverage formulations translate deployment requirements into optimization functions.

  • C. MOO Metrics: Common WSN QoS metrics include coverage, delay, active-node count, bit-error rate, and network lifetime.Energy conservation and lifetime can also quantify network performance in application-specific settings.
  • C. MOO Metrics: These metrics can conflict: higher energy consumption reduces lifetime, while more active nodes may affect other performance objectives.
  • C. MOO Metrics: Coverage in WSNs denotes sensing rather than radio reach and is classified into area, point, and barrier coverage.Area coverage observes an entire two-dimensional region; point coverage observes specified points; barrier coverage detects movement across a sensor barrier.
  • 1) Coverage:: Coverage optimization models a rectangular area as G = xy equal cells and defines coverage as the percentage of adequately covered cells.A cell is covered when at least one node lies within sensing range Rs.
  • 1) Coverage:: Finding optimal node positions for maximum coverage is generally NP-complete, so practical approaches seek suboptimal solutions that are computationally manageable.

2) Network Connectivity:

Network connectivity requires active nodes to communicate through suitable ranges, while deployment and energy decisions must preserve connectivity, coverage, and service over time. Connectivity is intertwined with lifetime and energy efficiency.

  • 2) Network Connectivity:: The connectivity objective places nodes within communication range while preventing them from becoming too close to one another.For the stated formulation, Rci − Rsi > 0 must hold.
  • 2) Network Connectivity:: Maintaining connectivity ensures messages reach the sink or base station, and connectivity loss is often treated as network-lifetime termination.
  • 2) Network Connectivity:: In dense networks, scheduling some nodes to sleep can save energy while remaining nodes provide continuous service.The design problem is to minimize active nodes while preserving acceptable QoS, sensing coverage, and connectivity.
  • 2) Network Connectivity:: Network lifetime is defined as the interval from application activation until a sensor node fails because its energy source is depleted.The paper typically assumes nodes have equal importance because one node’s death may partition the network or uncover an area.
  • 2) Network Connectivity:: Node lifetime depends inversely on its own generated information and the information it relays, linking lifetime to all nodes’ source rates.

5) Energy Efficiency:

Energy efficiency, delay, deployment quality, and node count represent competing WSN design concerns. The paper relates energy efficiency to transmission rate and power, while latency depends on forwarding hops and deployment constraints vary by application sensitivity.

  • 5) Energy Efficiency:: Energy efficiency is defined here as a node’s transmission rate divided by its power dissipation.The metric is closely related to network lifetime because sensor nodes have limited energy resources.
  • 5) Energy Efficiency:: For ηi = W log2(1 + γi) / pi, W is bandwidth, pi is transmission power, and γi is the destination SINR relative to node i.
  • 5) Energy Efficiency:: Energy conservation can use active-node scheduling with sleep modes or dynamically adjust nodes’ sensing ranges.Both approaches seek to use limited node energy efficiently and extend network lifetime.
  • 6) Network Latency:: With constant per-forwarder delays, minimizing source-to-sink latency corresponds to minimizing the number of intermediate forwarders.The delay includes queue, propagation, and transmission components collectively represented by c.
  • 5) Energy Efficiency:: Differentiated deployment assigns higher detection probabilities to safety-critical areas and lower probabilities to less sensitive regions to reduce node counts.A probabilistic detection model uses distance, sensing range Rs, and device parameter a.
  • 5) Energy Efficiency:: Node count affects both total deployment cost and stealth, particularly when nodes operate in battlegrounds where discovery should be minimized.

8) Number of Nodes:

The number of sensor nodes affects deployment cost, stealth, coverage, fault tolerance, and fair information delivery. Network design therefore balances node count with connectivity, reliability, sensing accuracy, and resource allocation.

  • Number of Nodes: Fault tolerance preserves uninterrupted operation when nodes fail because of physical conditions or energy depletion.k-connectivity provides independent paths, allowing networks with k ≥ 2 to tolerate some node and link failures.
  • Number of Nodes: Fair rate allocation is necessary because maximizing aggregate throughput can reduce some sources’ rates and impair application accuracy.Network utility maximization uses individual utility functions to represent satisfaction with assigned source rates; α = 1 gives proportional fairness and α →∞ gives max–min fairness.
  • Number of Nodes: Coverage improves event-detection probability, but sensing accuracy depends on timely delivery of density and latency information.Coverage is linked to monitoring status, sensing range, and node-to-cell distance, while target-detection accuracy depends on delivered network information.
  • Number of Nodes: Privacy concerns in WSN deployments include data-oriented and context-oriented threats arising from adversarial node capture and eavesdropping.Attackers may retrieve private communication keys and compromise sensor communications.

12) Network Security:

WSN security includes privacy threats, optimization formulations, and general solution concepts. The section distinguishes data compromise from traffic interception and frames network design through constrained multi-objective optimization and Pareto solutions.

  • Network Security: Data-analysis attacks compromise decrypted payloads, whereas traffic-analysis attacks track flows without decrypting the data.A malicious node may exploit decryption ability in data analysis, while traffic analysis proceeds through hop-by-hop eavesdropping.
  • Network Security: Privacy loss can be minimized using mutual information between a node’s position and an attacker-observed variable.The paper also identifies minimizing eavesdropping probability as a security objective.
  • Network Security: Heuristics and metaheuristics provide approximate solutions for difficult non-differentiable nonlinear optimization problems.Bio-mimetic strategies are presented as a broad class alongside finitely terminating and convergent iterative algorithms.
  • Network Security: A general MOP simultaneously optimizes multiple objectives subject to inequality and equality constraints in decision and objective spaces.Conflicting objectives motivate Pareto-optimality, where improving one objective cannot occur without worsening at least one other.
  • Network Security: Pareto-optimal solutions are feasible solutions not dominated by others, and their objective-space images form the Pareto front.The paper illustrates Pareto fronts for constrained and unconstrained MOPs.

B. MOO Algorithms

MOO algorithms for WSNs include scalarization methods and nature-inspired metaheuristics. Their suitability depends on objective and constraint structure, while scalarization methods differ in reachability, interpretability, computational burden, and Pareto-front coverage.

  • MOO Algorithms: The choice of objective function can have greater impact than the particular optimization algorithm, although algorithm selection remains vital for complex MOPs.The survey organizes MOO solution methods into scalarization, nature-inspired metaheuristics, and other approaches.
  • Linear Weighted-Sum Method: Linear weighted-sum scalarization normalizes metrics and combines them using weights into one objective.It is easy to implement but requires appropriate weights and repeated single-objective solves for different weight settings.
  • Linear Weighted-Sum Method: Weighted-sum methods are sensitive to subjective weights, cannot reach non-convex Pareto-front regions, and may cluster solutions unevenly.These limitations can reduce objectivity and leave important trade-off regions underrepresented.
  • ε-Constraints Method: The ε-constraints method optimizes one objective while imposing upper bounds on the remaining objectives as constraints.It can find non-convex-front solutions and weak Pareto-optimal points, but varying bounds is inefficient when more than two objectives exist.
  • Goal Programming: Goal programming assigns target values to objectives and minimizes their non-achievement to obtain a satisfactory solution.It extends linear programming to multiple conflicting objective measures.
  • Nature-Inspired Metaheuristics: Nature-inspired MOEAs and swarm-intelligence algorithms are frequently used because classical methods must fit restrictive objective and constraint structures.MOEAs seek representative Pareto-optimal solutions in a single run.

2) Nature-Inspired Metaheuristic Algorithms:

Nature-inspired metaheuristics provide flexible approximate-search approaches for WSN multi-objective problems, including evolutionary, swarm, immune, and reinforcement-learning methods.

  • Metaheuristics guide subordinate heuristics using concepts from biology, chemistry, physics, social sciences, or artificial intelligence.
  • Evolutionary Algorithms: Evolutionary algorithms evaluate objective or fitness functions through exploration and exploitation while producing multiple Pareto-optimal solutions in one run.
  • Evolutionary Algorithms: Genetic algorithms handle diverse objective functions and support parallel implementation for complex optimization problems.
  • Evolutionary Algorithms: MOGAs evolve populations across generations to obtain Pareto-optimal solution sets representing flexible trade-offs, including WSN layout alternatives.
  • Evolutionary Algorithms: Differential evolution uses randomly initialized candidate vectors, differential mutation, and relatively few parameters compared with traditional evolutionary algorithms.
  • Artificial Immune Systems: AIS uses immune-system-inspired cloning and mutation, while its population development differs from genetic algorithms’ crossover and mutation process.
  • Imperialist Competitive Algorithm: ICA models countries as population individuals, separates imperialists from colonies, and uses assimilation and imperialistic competition to evolve solutions.
  • Swarm Intelligence: Swarm methods exploit collective behavior in decentralized systems; ACO uses pheromone-marked paths and performs well for dynamic, distributed WSN routing.

3) Other Advanced Optimization Techniques:

The survey covers fuzzy logic, game theory, and representative multi-objective algorithms as advanced techniques for addressing WSN optimization problems.

  • Fuzzy Logic: Fuzzy logic represents imprecise inputs and outputs through linguistic descriptions and supports conclusions from ambiguous information.
  • Fuzzy Logic: A fuzzy system contains a fuzzifier, inference engine, fuzzy rule base, and defuzzifier that transform inputs into crisp outputs.
  • Fuzzy Logic: Low-complexity fuzzy logic has been applied to WSN routing by aggregating multiple objectives into a scalar cost reflecting their collective effects.
  • Game Theory: Game theory models conflict and cooperation among rational decision-makers, with pricing mechanisms guiding node behavior toward efficient Nash equilibria.
  • Representative MOO algorithms include MOGA, NPGA, NSGA-II, SPEA, MODE, and MOEA/D, alongside search, swarm, immune, fuzzy, neural, and game-theoretic methods.

C. Software Tools

The survey identifies numerous software tools available for solving multi-objective optimization problems.

  • Representative software packages include BENSOLVE, DEAP, Decisionarium, D-Sight, GUIMOO, IDSS, iSIGHT, jMetal, and MOMHLib++.

V. EXISTING LITERATURE ON USING MOO IN WSNS

Existing WSN research applies MOO to balance conflicting objectives such as coverage, lifetime, energy consumption, and delay, including deployment and power-assignment problems.

  • WSN performance metrics such as coverage versus lifetime and energy consumption versus delay create trade-offs requiring multi-objective optimization.
  • Coverage favors spread-out sensor layouts that reduce sensing overlap, whereas lifetime considerations create a conflicting deployment pressure.
  • Jourdan et al. used a MOGA to optimize WSN node locations for sensing and communication connectivity while maximizing coverage and lifetime.
  • The survey situates these studies alongside representative MOO algorithms, software tools, and coverage-versus-lifetime trade-off resources.
  • The resulting Pareto front lets the network dynamically select a preferred coverage–lifetime solution.
  • MOEA/D decomposed the deployment and power assignment problem into scalar subproblems while optimizing sensor locations and transmit powers.

B. Energy-versus-Latency Trade-offs

WSN studies examine conflicting energy, latency, lifetime, application-performance, coverage, and deployment-cost objectives using multi-objective formulations and optimization methods.

  • Energy-versus-Latency Trade-offs: Energy and delay conflict because shorter hops reduce energy, whereas fewer forwarders reduce delay but can unbalance forwarding load and deplete nodes unevenly.The section motivates jointly optimizing these objectives rather than optimizing them independently.
  • Energy-versus-Latency Trade-offs: Wake-up schemes balance energy savings against delay by overlapping wake-up procedures with packet transmissions or trading sleep efficiency against wake-up latency.Pipelined tone wake-up uses asynchronous overlap and wake-up tones to avoid large per-hop wake-up delays.
  • Energy-versus-Latency Trade-offs: Analytical models and data-aggregation algorithms compare or optimize energy-versus-delay trade-offs across dense, large, and multi-hop WSN scenarios.Reported approaches include latency-oriented and energy-oriented aggregation-tree construction and asymptotic transmission analyses.
  • Energy-versus-Latency Trade-offs: MOO frameworks address energy and collection delay in slot scheduling, routing, clustering, aggregation, and data forwarding, sometimes adding uniform battery depletion.One forwarding approach seeks minimum energy, uniform battery depletion, and minimum delay using concentric communication bands and a weighted scale-uniform-unit sum.
  • Energy-versus-Latency Trade-offs: Higher sensing and communication rates can improve application performance while increasing energy consumption and reducing network lifetime.Related work therefore studies trade-offs among network utility, throughput, rate allocation, energy, lifetime, node count, coverage, and collection duration.

E. Reliability-Related Trade-offs

Reliability-related WSN research balances QoS, energy, latency, data rate, lifetime, security, coverage, and connectivity under challenging routing and deployment conditions.

  • Reliability-Related Trade-offs: Maintaining QoS for picture, video, and other data applications requires routing to account simultaneously for energy, delay, bandwidth, and reliability.These factors create conflicting routing requirements summarized in the reliability-related trade-off literature.
  • Reliability-Related Trade-offs: Existing protocols use probability-based forwarding, QoS-aware routing, and fuzzy-random multiobjective optimization to balance energy, latency, reliability, and related path factors.The fuzzy-random formulation also considers delay jitter, interference, and energy balance.
  • Reliability-Related Trade-offs: Data rate, delivery reliability, energy consumption, and network lifetime form an inherent trade-off under link-capacity, reliability, and energy constraints.A reported formulation was transformed into a separable convex problem before developing a distributed subgradient dual-decomposition algorithm.
  • Reliability-Related Trade-offs: Security-oriented designs seek to maximize security without degrading lifetime, while disjoint-set-cover and fuzzy-dominance methods jointly optimize coverage, energy, lifetime, node count, and connectivity.The cited approaches include MOEA/D, NSGA-II, and MOEA/D with fuzzy dominance.
  • Reliability-Related Trade-offs: Open problems include extending MOO from predominantly single-hop studies to multi-hop, mobile, three-dimensional, hostile, secure, cognitive-radio, and smart-grid WSN settings.The paper specifically notes limited multi-hop coverage, frequent two-dimensional assumptions, mobility effects on connectivity, security attacks, spectrum scarcity, and smart-grid constraints.

VII. CONCLUSIONS

The paper surveys multi-objective optimization in WSNs, covering optimization objectives, MOP-solving algorithms, and design trade-offs. It also identifies future directions including dynamic deployment, secure multipath routing, and optimization in 3D, cognitive-radio, and smart-grid networks.

  • The paper provides a tutorial and survey of multi-objective optimization research in WSNs.
  • It introduces WSN optimization objectives and explains families of algorithms for solving multi-objective optimization problems.
  • Because multiple objectives produce Pareto-optimal solutions rather than one globally optimal solution, algorithms may seek as many Pareto-optimal solutions as possible.
  • Future research directions include multi-hop transmissions, highly dynamic node deployment, secure multipath routing, and optimization in 3D, CR-WSNs, and smart grids.
Loading 1609.04069v1…