PaperlayerView exact revision on arXiv

Source-linked AI summary

Enhancing UAV Trajectory and Communications Through Vision-Inertial Tracking

Abdallah S. Ghazy, Hussein A. Ammar, James Bayes, Francois Chan

arXiv:2609.03099v1eess.SP

TL;DR

In GPS-denied wireless environments, UAV communication needs to address localization alongside energy efficiency, reliability, and physical-layer security. VIT-UAVCom uses camera–IMU tracking to jointly optimize UAV trajectory and communication, reporting up to 144% higher energy efficiency than K-means and faster computation with binary search than linear search.

  • Problem

    The paper identifies a gap in real-time computer-vision-based frameworks jointly optimizing UAV trajectory, outage probability, and channel secrecy, against a background of GPS denial and potential eavesdropping.

    Source: §I · p. 2, A. Overview of the Proposed Framework · p. 2

  • Method

    VIT-UAVCom uses onboard camera–IMU tracking to localize moving users and jointly optimize UAV trajectory and adaptive RF beamforming for energy-efficient, secure, and reliable communication.

    Source: §I · p. 2

  • Results

    Up to 144% energy efficiency improvement over K-means is reported; BS reduces computational time by about 50% versus LS, with a slight efficiency trade-off.

    Source: §I · p. 2

  • Takeaways & Limitations

    The framework supports lightweight real-time optimization for green non-terrestrial-network communications, with LS yielding the highest energy efficiency and BS the lowest computational time.

    Source: §VI · p. 12

    The model neglects pointing errors on the BS-to-UAV link, assuming a wide RF beam and no nearby eavesdroppers.

    Source: C. Wind-Induced Pointing Errors · p. 7

Abstract

from arXiv · show

In this paper, we propose an energy-efficient and reliable communication system for non-terrestrial networks deployed in dynamic GPS-denied wireless environments, enabled by a Vision--Inertial Tracking-Assisted UAV Communication (VIT-UAVCom) system. To the best of our knowledge, this is the first work to exploit onboard UAV cameras and IMU sensors for UAV-assisted communications. We consider a complete VIT-UAVCom system that incorporates the key design parameters while explicitly accounting for system noise and residual tracking inaccuracies. Building on this framework, we formulate an optimization problem for jointly designing the UAV trajectory and communication performance to improve propulsion energy efficiency, reduce outage probability, and enhance physical-layer security. We then develop a dedicated solution framework to efficiently compute near-optimal trajectory and communication control actions in dynamic scenarios. Furthermore, to enable real-time implementation, we propose and evaluate three optimizers, namely linear search (LS), binary search (BS), and genetic search. Our numerical results demonstrate that our proposed VIT-UAVCom framework significantly outperforms the K-means benchmark in terms of energy consumption while maintaining robust secrecy performance and reliable user coverage. Specifically, our proposed framework improves the energy efficiency by 144% compared to the benchmark. Interestingly, our results also show that, compared with the LS, the BS reduces the computational time by approximately 50%.

I. INTRODUCTION

Prior UAV deployment and vision-assisted communication studies leave a gap in real-time joint optimization of trajectory, outage, and secrecy. VIT-UAVCom combines camera–IMU tracking with trajectory and communication control for GPS-denied networks.

I. INTRODUCTION

  • Prior dynamic deployment studies often assume known user locations or infer mobility indirectly, while this work targets real-time sensing-based localization.The paper identifies the lack of a computer-vision-based unified framework jointly optimizing trajectory, outage probability, and channel secrecy in real time.Source: §I · pp. 1–2
  • VIT-UAVCom uses onboard cameras and an IMU to localize moving ground users and jointly optimize UAV trajectory and adaptive RF beamforming.Its stated objectives are lower UAV energy consumption and secure, reliable communication.Source: §I · p. 2
  • Up to 144% energy efficiency improvement over K-means is reported for the proposed framework.The introduction also reports that BS reduces computational time by about 50% versus LS, with a slight efficiency trade-off.Source: §I · p. 2

A. Overview of the Proposed Framework

  • The overview links user tracking and motion detection to trajectory and beam adjustments, while directional beams concentrate energy toward legitimate users and mitigate leakage.The camera footprint depends on UAV altitude, field of view, and transmit power; terrain obstruction, GPS denial, and potential eavesdropping motivate the system setting.Source: A. Overview of the Proposed Framework · pp. 2–3

B. VIT-UAVCom System Model

The system combines camera–IMU localization with communication control: estimated user positions guide UAV motion and adaptive RF beamforming. A camera footprint is selected to cover the user group with a safety margin.

B. VIT-UAVCom System Model

1) Relative Position Estimation:

  • The camera altitude is selected so its footprint covers the user group’s maximum extent plus a safety margin for localization uncertainty and abrupt motion.The model assumes a conical camera projection; footprint radius depends on altitude and camera field of view.Source: 1) Relative Position Estimation · p. 4
  • Estimated user positions and detected mismatches with current RF beam footprints trigger optimization of UAV displacement and beamforming parameters.The resulting displacement guides propulsion control, while beam orientation and beamwidth are adjusted dynamically for users.Source: 1) Relative Position Estimation · p. 5
  • The VIT and communication modules form a continuous closed loop for tracking legitimate users and providing real-time secure and reliable service.The transmitted beamformed signal combines the beamforming vector and communication symbol with system noise.Source: 1) Relative Position Estimation · p. 5

2) Residual Localization Errors:

The residual localization error model combines sensing and calibration uncertainty while treating instantaneous vision-aided updates as bounded rather than drifting without bound. Its assumptions and optional temporal-correlation model delimit how the error is represented.

2) Residual Localization Errors:

  • Instantaneous vision-aided localization updates keep error bounded without unbounded drift, unlike long-horizon SLAM-based state propagation.Short-term temporal correlation may still arise from slowly varying IMU biases.Source: 2) Residual Localization Errors · p. 5
  • The error model assumes locally planar users, near-nadir camera geometry, fixed pre-calibrated camera parameters, inertially obtained altitude and attitude, and locally linear sensing.Under these conditions, horizontal errors dominate and first-order uncertainty propagation is used.Source: 2) Residual Localization Errors · p. 5
  • The model represents residual localization error as a zero-mean Gaussian vector with covariance capturing visual, inertial, and calibration uncertainties.Linearization propagates component covariances through Jacobians; the camera contribution is approximated for horizontal error under near-nadir geometry.Source: 2) Residual Localization Errors · p. 5
  • IMU-induced variance accounts for error accumulation between camera updates, and the model optionally represents slowly varying bias effects with first-order Gauss–Markov correlation.Covariance can be inflated to represent harsher sensing conditions.Source: 2) Residual Localization Errors · pp. 5–6

III. COMMUNICATION SYSTEM MODELING

The communication model considers a terrestrial base station backhauling to a UAV that serves ground users over the downlink while potential eavesdroppers attempt interception.

III. COMMUNICATION SYSTEM MODELING

  • A terrestrial base station provides the wireless backhaul link to the UAV.The communication system includes a BS-to-UAV link and a UAV-to-user downlink.Source: §III · p. 6
  • The UAV acts as transmitter for K legitimate ground-user receivers over the downlink.The passage specifies the UAV as Tx and the ground users as Rxs.Source: §III · p. 6
  • K potential eavesdroppers attempt to intercept the transmitted signals, and the section evaluates downlink communication performance.Source: §III · p. 6

A. Network Mobility Model

The model represents users moving around a shared, time-varying group center. It captures changing cluster size with a bounded radius process.

A. Network Mobility Model

  • Users move around a common, time-varying group center under a point-reference group mobility model.The center is denoted c[n].Source: A. Network Mobility Model · p. 6
  • Users are distributed within a disk whose radius R[n] evolves to represent group expansion and contraction.User radius and angle are modeled as uniform within the disk.Source: A. Network Mobility Model · p. 6
  • The group radius updates by adding ν[n] to its previous value, then clipping it to [Rmin, Rmax].Values outside the allowed range are set to the nearest bound.Source: A. Network Mobility Model · p. 6

B. Path Loss

The path-loss model specifies the UAV, base-station, user, and eavesdropper geometry, then characterizes the links using line-of-sight conditions and air-to-ground path loss.

B. Path Loss

  • The UAV is the coordinate origin, and safety zones of radius ds surround legitimate receivers to restrict eavesdropper proximity.The section also defines relative locations for the base station, receivers, and eavesdroppers.Source: B. Path Loss · p. 6
  • The base-station-to-UAV link assumes strong line of sight, while the UAV-to-user link uses an air-to-ground model accounting for LoS probability and LoS/NLoS losses.The model also describes a Gaussian RF beam directed toward each receiver with adjustable beamwidth.Source: B. Path Loss · pp. 6–7
  • The UAV-to-user model uses environment-dependent excessive-loss parameters and the UAV-user distance and elevation angle.The path-loss passage identifies ηLoS, ηNLoS, a, and b as environment-dependent quantities.Source: B. Path Loss · p. 7

C. Wind-Induced Pointing Errors

The model accounts for wind-driven UAV displacement and orientation changes that can cause pointing errors. It then characterizes link capacity, outage probability, and secrecy rate for communication evaluation.

C. Wind-Induced Pointing Errors

  • Wind-induced displacement and orientation fluctuations cause pointing errors; the displacement components are modeled as independent Gaussian variables, with lateral fluctuations dominating vertical ones.Only UAV disorientation is considered because its effects are described as more pronounced than receiver dynamics.Source: C. Wind-Induced Pointing Errors · p. 7
  • The UAV-to-user link is especially sensitive to beam misalignment because beamwidth must balance energy efficiency and information leakage.Pointing errors on the wide-beam base-station-to-UAV link are neglected under the stated assumption of no nearby eavesdroppers.Source: C. Wind-Induced Pointing Errors · p. 7

D. Channel Secrecy and Outage Probability

  • Channel capacity denotes the maximum achievable bit rate, and the model relates required average transmit power to a target capacity over bandwidth.The base-station-to-UAV capacity is constrained by the number of bits per channel use.Source: D. Channel Secrecy and Outage… · p. 7
  • Outage probability is evaluated against a channel-capacity threshold, while channel secrecy rate is the nonnegative difference between legitimate-link and eavesdropper capacities.The threshold is denoted cth.Source: D. Channel Secrecy and Outage… · p. 7

E. Energy Consumption Model

The energy model totals propulsion, hovering, and communication energy over a mission, with each term tied to UAV motion, hover duration, or communication activity.

E. Energy Consumption Model

  • Total mission energy is the sum of propulsion, hovering, and communication energy.The terms are denoted Epro, Ehov, and Ecom, respectively.Source: E. Energy Consumption Model · p. 8
  • Communication energy depends on transmit power and downlink bit rate for each user.The model also specifies circuit power and the number of bits received per user among its parameters.Source: E. Energy Consumption Model · p. 8

IV. PROBLEM FORMULATION AND SOLUTION

The section formulates UAV energy minimization under secure-communication, outage, coverage, and operational constraints, then introduces a real-time optimizer. Its design balances energy-saving motion against the tracking and beamforming needed for secure, reliable service.

IV. PROBLEM FORMULATION AND SOLUTION

  • The real-time optimizer is designed to minimize UAV energy while meeting network-security and outage-capacity constraints.The solution framework seeks a suboptimal solution within short processing time.Source: §IV · p. 8

A. Problem Formulation

  • Propulsion energy dominates total UAV energy, so the formulation minimizes total energy by reducing propulsion energy and flight-path length.Shorter travel means more hovering; unnecessary stops also increase acceleration and deceleration losses.Source: A. Problem Formulation · p. 8
  • Security and reliability impose competing beamwidth choices: narrower beams favor secrecy, while wider beams improve coverage robustness under outage constraints.The formulation also balances shorter energy-saving paths against closer user tracking, which can reduce path loss.Source: A. Problem Formulation · p. 8
  • The multi-objective design is converted to a constrained single-objective problem using the ϵ-constraint method.The objective minimizes total UAV energy while secure communication and outage probability must meet specified requirements.Source: A. Problem Formulation · p. 8
  • The formulation defines a 3D UAV position and mission duration, with propulsion, hovering, and communication energy terms and secrecy-rate and outage constraints.It also includes capacity, transmit-power, beamwidth, steering-angle, and dimensional constraints.Source: A. Problem Formulation · p. 8
  • The solution workflow evaluates FoV margins and QoS, then either hovers and updates beamforming or estimates motion and applies a displacement.The algorithm checks VIT updates and user feedback while selecting whether to move.Source: A. Problem Formulation · p. 9
  • At each time slot, the UAV jointly optimizes its position and beamforming to minimize per-slot energy while satisfying communication constraints.Position is constrained by user-cluster coverage; altitude must meet line-of-sight, footprint, and regulatory requirements.Source: A. Problem Formulation · pp. 8–9
  • The UAV’s footprint must contain the user cluster, while altitude is bounded by minimum operational and footprint requirements and a regulatory limit.The beam’s polar orientation is limited to half the UAV field-of-view angle.Source: A. Problem Formulation · pp. 8–9

B. Solution Approach

The solution uses a low-complexity policy to handle a nonlinear, dynamically coupled optimization problem. It triggers hovering or repositioning from visibility and QoS conditions, with motion directed toward the user-cluster centroid.

B. Solution Approach

  • The optimization is nonconvex and dynamically coupled, making closed-form solutions challenging and generally requiring iterative numerical methods for efficient suboptimal solutions.Nonlinear rate expressions and mobility-driven channel variation create nonconvex feasible regions; trajectory-dependent propulsion energy couples time slots.Source: B. Solution Approach · p. 9
  • The policy fixes UAV altitude and does not predict future user trajectories, instead using motion toward the user-cluster centroid as a heuristic.Fixed altitude removes vertical maneuvers; altitude selection must account for obstacles and flight regulations, with lower altitude trading smaller footprint and collision risk against lower transmission power and turbulence.Source: B. Solution Approach · p. 9
  • The UAV hovers when users remain within the FoV and satisfy QoS, updating beam direction and width; otherwise it computes a repositioning action that maintains coverage and QoS.VIT proximity-to-boundary signals and user feedback inform the decision.Source: B. Solution Approach · p. 9
  • The event trigger keeps the UAV stationary when the worst-case visibility margin meets its threshold and all users satisfy QoS constraints.Otherwise, a motion policy is activated.Source: B. Solution Approach · p. 10
  • When motion is needed, the UAV estimates the dominant user-cluster direction and optimizes displacement along that direction to compute its next position.The cluster centroid and displacement update define the repositioning step.Source: B. Solution Approach · p. 10

C. Searching-Based Optimizers

The section compares linear, binary, and genetic search for selecting a feasible UAV displacement. Binary search exploits monotonic feasibility to reduce the search interval, while the included figures show optimizer trajectories and performance-analysis dimensions.

C. Searching-Based Optimizers

  • Binary search tests midpoints, linear search checks displacements sequentially by step size, and genetic search refines a population through selection, crossover, mutation, and elitism.All three methods seek a near-optimal displacement under the same optimization framework.Source: C. Searching-Based Optimizers · p. 10
  • All three optimizers have computational time O(J_x O_f), where O_f is the cost of one evaluation and J_x is the optimizer’s iteration count.The search procedures are presented as real-time alternatives for determining displacement.Source: C. Searching-Based Optimizers · p. 10
  • Binary search reduces the displacement interval logarithmically and typically needs fewer iterations than linear or genetic search to achieve near-optimal performance.Its per-iteration midpoint test updates the interval bounds according to feasibility.Source: C. Searching-Based Optimizers · p. 10
  • Figure 4 compares UAV paths for BS versus KM and LS versus GS, while Figure 3 covers energy, CSR, outage, and transmit-power/FoV relationships.The simulation-parameter table is identified as Table III.Source: Table III · p. 10, Figure 4 · p. 11, Figure 3 · p. 11

V. NUMERICAL RESULTS

The evaluation compares the proposed optimizers with centroid-following K-means across mission durations and examines trade-offs among energy, outage, and secrecy performance.

V. NUMERICAL RESULTS

  • The evaluation compares BS, LS, and GS against a K-means benchmark that follows the user-cluster centroid without the proposed optimization framework.Source: §V · p. 11

A. System Characteristics

  • Narrow beams generally improve CSR but increase misalignment-related outage risk; wider beams improve robustness, while lower altitude improves CSR and outage performance at extra propulsion-energy cost.Source: A. System Characteristics · p. 11
  • For a given footprint, increasing FoV angle lowers required transmit power by allowing lower altitude, while larger footprints require more power and gains become marginal at moderate-to-large angles.Source: A. System Characteristics · p. 11

B. System Performance

VI. CONCLUSIONS AND DISCUSSIONS

  • The conclusion presents VIT-UAVCom as improving energy efficiency while maintaining reliable communication in GPS-denied environments using onboard camera and IMU sensors.Source: §VI · p. 12
Loading 2609.03099v1…