PaperlayerView exact revision on arXiv

Source-linked AI summary

UAV Assisted Data Collection for Internet of Things: A Survey

Zhiqing Wei, Mingyue Zhu, Ning Zhang, Lin Wang, Yingying Zou, Zeyang Meng, Huici Wu, Zhiyong Feng

arXiv:2211.09555v1cs.RO

TL;DR

IoT data collection must address massive sensor deployments and requirements for low energy consumption, low delay, and high reliability. This paper surveys UAV-assisted collection through system modeling and reviews of clustering, collection modes, and joint path planning with resource allocation. It concludes by identifying efficient multiple access and joint sensing and data collection as future research trends.

  • Problem

    Massive IoT sensor deployments create data-collection challenges, while existing surveys insufficiently cover collection modes, detailed clustering algorithms, and joint path planning with resource allocation.

  • Method

    The paper conducts a comprehensive literature review covering the system model, sensor clustering, UAV collection modes, and joint path planning and resource allocation.

  • Results

    The survey synthesizes UAV-assisted IoT data-collection scenarios and technologies, including hovering, flying, and hybrid modes and graph-theory, optimization-theory, and AI-based planning approaches.

  • Takeaways & Limitations

    Efficient multiple access and joint sensing and data collection are identified as likely future research trends for UAV-assisted IoT data collection.

Abstract

from arXiv · show

Thanks to the advantages of flexible deployment and high mobility, unmanned aerial vehicles (UAVs) have been widely applied in the areas of disaster management, agricultural plant protection, environment monitoring and so on. With the development of UAV and sensor technologies, UAV assisted data collection for Internet of Things (IoT) has attracted increasing attentions. In this article, the scenarios and key technologies of UAV assisted data collection are comprehensively reviewed. First, we present the system model including the network model and mathematical model of UAV assisted data collection for IoT. Then, we review the key technologies including clustering of sensors, UAV data collection mode as well as joint path planning and resource allocation. Finally, the open problems are discussed from the perspectives of efficient multiple access as well as joint sensing and data collection. This article hopefully provides some guidelines and insights for researchers in the area of UAV assisted data collection for IoT.

I. INTRODUCTION

IoT sensor growth makes data collection demanding in energy, delay, and reliability, while UAVs offer mobile, flexible collection. The survey reviews core technologies and identifies gaps in existing surveys.

  • By 2030, the number of IoT sensors is estimated to exceed 100 trillion, intensifying data-collection requirements for low energy consumption, low delay, and high reliability.
  • UAVs support agile, flexible, and low-cost IoT data collection, including in complex, harsh, or remote environments.Their proximity to sensors can reduce IoT energy consumption and facilitate timely data collection.
  • UAV-assisted collection faces coupled optimization challenges involving sensor clustering, collection-mode selection, and joint UAV path planning with resource allocation.Sensor deployment and clustering affect collection performance, while resource allocation is coupled with UAV path planning.
  • This survey reviews sensor clustering, UAV data-collection modes, and joint UAV path planning and resource allocation.It presents these as three key technologies for UAV-assisted IoT data collection.
  • A. Existing Surveys and Tutorials: Existing surveys provide limited coverage of UAV data-collection modes, detailed sensor-clustering algorithms, and joint path-planning and resource-allocation schemes.

B. Contributions and Organization

The article surveys UAV-assisted IoT data collection by reviewing its system model, key technologies, and future research directions. It frames data collection as a coordinated system involving sensors, UAVs, control stations, base stations or access points, and servers.

  • Contributions: The survey reviews sensor clustering, UAV data collection modes, and joint UAV path planning with resource allocation.It identifies efficient multiple access and joint sensing and data collection as likely future research trends.
  • Network Model: The network includes sensors, UAVs, a control station, a base station or access point, and a server.Sensors send data to UAVs or directly to the BS/AP; the server stores and processes the collected data.
  • Network Model: UAVs collect sensor data while flying above sensors, deliver it to the BS/AP, and return to the control station.The control station plans the flight path, while the UAV operates under mobility, energy, and buffer-capacity constraints.
  • System Assumptions: When sensor populations are large, sensors are clustered into cluster heads and cluster members, with cluster heads communicating with the UAV.For small or scattered deployments, sensors can upload data directly to the UAV.
  • Mathematical Model: The mathematical model represents probabilistic LoS transmission and sensor-to-UAV rate using channel, power, bandwidth, noise, and environmental parameters.The LoS probability depends on the sensor-UAV angle and environmental parameters related to building density and height.
  • Optimization: Path planning jointly considers distance, flight time, and energy consumption, while hovering, flying, and hybrid collection modes affect data collection efficiency.In hovering mode, the UAV collects each cluster head’s data at a designated collection point.

1) Sensor energy consumption constraint:

In hovering mode, sensors transmit under unchanged channel conditions during the UAV’s hovering period. The model constrains sensor energy and collected data requirements.

  • Sensor energy consumption constraint: In hovering mode, unchanged channel conditions allow sensor s_k to transmit with constant power P_k during hovering time t_k.t_k denotes the UAV’s hovering time at data collection point u_k.
  • Sensor energy consumption constraint: Each sensor s_k must satisfy its minimum data collection requirement r_k.The requirement is stated separately from the sensor energy-consumption constraint.

2) Data collection constraint:

The data-collection model constrains UAV communication, energy, trajectory, and data timeliness. It aims to collect sensor data efficiently while respecting route, energy, and AoI requirements.

  • UAV energy constraint: The UAV’s total energy consumption cannot exceed its maximum available energy Emax.The energy constraint includes propulsion and flight-related power terms.
  • UAV trajectory constraint: The trajectory constraint requires the UAV to start at the control station, visit all collection points, and return to the BS.Binary variable xi,j indicates whether the UAV flies from ui to uj.
  • Age of information constraint: AoI measures data timeliness by including transmission delay, source waiting time, and destination residence time.For sequential CH access, collected-packet AoI includes uploading, UAV return, and unloading times.
  • Age of information constraint: Expired packets receive an AoI equal to the overall time required for the UAV to complete the task.Queuing assumptions such as FCFS, LCFS, priority-based, and limited-buffer models modify the relevant delay parameters.
  • Optimization objectives: UAV-assisted collection schemes jointly target collected-data volume, flight time, energy consumption, and AoI.The survey identifies sensor clustering, UAV collection mode, and joint path planning and resource allocation as key issues.

III. CLUSTERING OF SENSORS

Sensor clustering organizes large networks into subnetworks led by cluster heads, reducing routing complexity while balancing sensor and UAV costs. The survey covers clustering classifications, LEACH, energy-aware variants, and UAV-specific trade-offs.

  • Clustering algorithms: Clustering divides large sensor networks into subnetworks and selects one cluster head to aggregate data from each cluster.This reduces the routing cost associated with complicated network topologies.
  • Clustering algorithms: Clustering algorithms are classified by cluster-head selection as deterministic, random, or adaptive, and by control as distributed, hybrid, or centralized.These classifications organize methods according to how cluster heads are selected and how clustering is controlled.
  • LEACH and its variants: LEACH periodically switches randomly selected cluster heads to distribute sensor energy consumption and prolong network lifetime.Sensors join clusters according to received-signal strength and transmit feedback to the cluster head using TDMA.
  • LEACH and its variants: In UAV-assisted collection, random LEACH selection may choose low-energy sensors that cannot complete communication with the UAV.This motivates clustering methods that account for sensor energy and UAV communication requirements.
  • UAV-specific clustering trade-offs: The number of clusters creates a trade-off between intra-cluster sensor energy and UAV flight energy.Many clusters reduce forwarding energy but increase UAV flight energy; few clusters have the opposite effect.
  • UAV-specific clustering methods: Existing methods incorporate sensor energy, UAV flight time, path planning, iterative k-means updates, and decomposition of joint optimization problems.Some approaches select high-power sensors as cluster heads or use value-of-information thresholds to reduce transmissions.

IV. DATA COLLECTION MODE

UAV data collection modes are classified by UAV movement into hovering, flying, and hybrid modes, each presenting different trade-offs among transmission stability, collection speed, energy consumption, and timeliness.

  • Data collection modes: UAV data collection modes comprise hovering, flying, and hybrid modes, classified according to UAV movement during collection.These modes are compared in Table III and surveyed across the literature in Table IV.
  • Hovering mode: Hovering mode provides relatively stable transmission but consumes relatively large energy because the UAV remains over a collection point.The UAV may hover above a cluster head, nearby location, or sub-region while collecting sensor data.
  • Optimization across modes: Joint optimization of cluster-head placement, UAV trajectories, collection mode, scheduling, and energy-related variables targets collection time, reliability, AoI, and sensor lifetime.The literature includes optimizing hovering locations and duration, sensor association, wake-up scheduling, and weighted UAV-sensor energy consumption.
  • Flying mode: Flying mode completes collection quickly while moving, but may fail to meet timely collection requirements when sensors have large data volumes.The UAV slows while passing collection points and collects data within sensor communication ranges.
  • Flying mode: Under flying mode, the optimal UAV speed increases with sensor energy and density but decreases with the amount of data to upload.Flight distance, speed, and sensor transmit power can be jointly optimized to reduce collection flight time.
  • Hybrid mode: Hybrid mode combines hovering and flying, with energy consumption and flight time between those of hovering and flying modes.Its selection depends on the application scenario and performance requirements, and it can dynamically adjust collection behavior according to UAV energy.

V. JOINT PATH PLANNING AND RESOURCE ALLOCATION

UAV path planning seeks feasible trajectories under objectives such as distance, flight time, energy, and data freshness. The survey organizes methods into graph-theory and optimization-theory approaches while reviewing their use in data collection.

  • Path planning overview: UAV path planning calculates a trajectory from source to destination under specified conditions, with goals including shortest path, shortest flight time, and lowest energy consumption.Its design strongly affects UAV-assisted data collection performance.
  • Graph theory based algorithms: Graph-theory methods convert the UAV flight space into a graph and search for paths, including Voronoi diagrams, PRM, and Hilbert curves.Table V compares these algorithms using collected-data, time-efficiency, energy-efficiency, and AoI metrics.
  • Evaluation metrics: The reviewed algorithms are evaluated against amount of collected data, time efficiency, energy efficiency, and age of information.AoI measures the interval from sensor-data generation until the UAV returns and uploads the data to the data center.
  • Graph theory based algorithms: Voronoi diagrams generate paths away from obstacles, while modified versions can select energy-aware hovering points to maximize minimum remaining sensor energy.The modified approach is reported to shorten total UAV flight length relative to prior Voronoi-based planning.
  • Graph theory based algorithms: PRM uses sampled points and collision-free straight-line edges to construct a roadmap for path search.Its sampling-based formulation is applied to feasible, collision-free data-collection trajectories under energy constraints.

3) Hilbert curve:

The Hilbert curve is a space-filling mapping from one dimension to two dimensions that partitions an area into subspaces and supplies a coverage trajectory for UAV data collection.

  • 3) Hilbert curve:: The Hilbert curve maps one-dimensional space to two-dimensional space and passes through all subspaces created by area partitioning.The survey illustrates its construction from first through fourth order.
  • 3) Hilbert curve:: Each Hilbert-curve iteration divides every subspace into four smaller subspaces, improving trajectory granularity while preserving passage through subspace centers.Repeated iteration forms a curve covering the entire collection area.
  • 3) Hilbert curve:: The resulting curve provides a trajectory for UAVs to cover the data-collection area.The cited survey passage describes this as an optimal trajectory for area coverage.

3) Successive convex approximation:

Successive convex approximation transforms nonconvex optimization into iterative convex subproblems, enabling efficient approximate solutions for UAV trajectory and resource-allocation tasks reviewed alongside AI-based methods.

  • 3) Successive convex approximation:: SCA transforms a nonconvex optimization problem into a series of convex problems solved iteratively until a convergence condition is reached.The final solution is approximately treated as a solution to the original problem.
  • 3) Successive convex approximation:: SCA is used with path discretization to optimize fixed-wing UAV trajectories for minimum propulsion energy while satisfying throughput requirements.The trajectory model consists of straight-line and circular segments.
  • 3) Successive convex approximation:: SCA and block coordinate descent jointly optimize UAV trajectory, data-collection time, and sensor wake-up scheduling for a local optimum under UAV-sensor energy trade-offs.The formulation minimizes weighted energy consumption of the UAV and sensors.
  • 3) Successive convex approximation:: SCA supports multi-UAV completion-time optimization by converting the problem through bisection and time discretization before obtaining a KKT solution.The formulation also maintains sufficient sensor energy for uploading data.
  • AI-based algorithms: AI-based path-planning algorithms are classified into machine-learning and intelligent-optimization algorithms and are described as handling path-planning uncertainty more effectively than traditional algorithms.The survey presents this classification in Fig. 9.
  • 3) Successive convex approximation:: SCA-based iterative algorithms jointly optimize UAV three-dimensional paths, time allocation, power control, throughput, and wireless-power-assisted data collection.The reviewed applications include maximizing collected data, average throughput, and power-transmission efficiency.
  • AI-based algorithms: Matrix completion guides UAV sampling-point selection using position/time matrices and dynamically adjusts sampling probabilities to reduce data redundancy.The method uses observed matrix structure to recover missing data and guide collection-point selection.

1) Supervised learning:

The survey organizes learning-based UAV path planning around supervised learning, reinforcement learning, and deep reinforcement learning, linking each to data collection objectives and decision settings.

  • Supervised learning:: Supervised learning classifies new data from labeled training data, while regression predicts continuous target values from extracted sample features.
  • Reinforcement learning:: Reinforcement learning models sequential interaction with the environment to learn actions that maximize cumulative revenue, commonly for UAV obstacle avoidance and path planning.
  • Q-learning algorithms:: Q-learning-based methods formulate UAV collection as decision processes and optimize objectives including collected data, collection time, energy consumption, collision avoidance, and transmission reliability.
  • Deep reinforcement learning algorithms:: Reinforcement-learning optimization is reported to have low time consumption and data loss rate.
  • Deep reinforcement learning algorithms:: Deep reinforcement learning combines deep learning perception with reinforcement-learning decision making to solve joint UAV data-collection optimization problems.

3) Intelligent optimization algorithms:

Intelligent optimization methods combine search and learning mechanisms to approximate strong solutions for UAV path planning and related data-collection objectives.

  • 3) Intelligent optimization algorithms:: Intelligent optimization algorithms combine random and local search with learning strategies to improve approximately optimal solutions.
  • Individual behavior based algorithms:: Individual-behavior methods such as simulated annealing optimize feasible solutions while probabilistically escaping local optima in large solution spaces.
  • Swarm intelligence based algorithms:: Swarm-intelligence methods include GA, ACO, PSO, DE, SFLA, and CS, using populations and evolutionary or information-sharing mechanisms to improve solutions.
  • Genetic algorithm: GA-based studies optimize cluster-head placement, UAV number, and trajectories for objectives including data collection time and energy consumption.
  • Ant colony optimization: ACO-based path planning is used to reduce collection time, approach theoretical per-node capacity, or reduce UAV energy consumption.
  • Swarm intelligence based algorithms:: PSO, DE, SFLA, and CS are applied to flight-time, positioning, energy, throughput, latency, and obstacle-related path-planning problems.

VI. FUTURE TRENDS

The survey identifies efficient multiple access and joint sensing with data collection as underexplored challenges for large-scale and location-uncertain IoT deployments.

  • Most reviewed studies focus on sensor clustering and UAV path planning to achieve efficient, low-energy, low-delay data collection.
  • Large-scale IoT deployments create data congestion when many sensors transmit data to a UAV.
  • Unknown sensor locations can cause transmission conflicts and make advance positioning difficult in ocean, agricultural, and soil monitoring scenarios.
  • The proposed complete solution combines joint sensing and data collection, sensor clustering, efficient access, efficient routing, and joint path planning with resource allocation.

A. Efficient Multiple Access

Future UAV-IoT research emphasizes efficient multiple access and joint sensing and data collection to address massive connectivity, spectrum constraints, and uncertain sensor positions.

  • A. Efficient Multiple Access: MAC protocols must address transmission conflicts, packet delay, throughput, massive sensor access, and UAV–sensor energy balance.
  • A. Efficient Multiple Access: Resource scheduling across time, frequency, and space, along with regional access priorities, is used to reduce transmission conflicts.
  • A. Efficient Multiple Access: Bandwidth limits can bottleneck MAC performance as sensor numbers grow, motivating mmWave and THz technologies for higher data rates and lower latency.
  • A. Efficient Multiple Access: NOMA is identified as promising for improving MAC efficiency, spectral efficiency, and massive IoT connectivity when bandwidth and scheduling are insufficient.
  • B. Joint Sensing and Data Collection: Joint sensing and data collection can locate sensors during collection while meeting low-cost and low-energy IoT requirements.
  • B. Joint Sensing and Data Collection: Existing work jointly optimizes UAV trajectories and sensor transmission schedules for low-energy collection and high-precision 3D positioning.
  • B. Joint Sensing and Data Collection: Future algorithms must handle missing or inaccurate sensor locations and perform low-power sensing and data collection simultaneously.
Loading 2211.09555v1…