PaperlayerView exact revision on arXiv

Source-linked AI summary

Dynamic Task and Resource Scheduling Towards Space-Air-Ground-Sea Integrated Network

Yufei Ye, Shijian Gao, Xinhu Zheng, Liuqing Yang

arXiv:2609.31011v1cs.NI

TL;DR

Maritime vessels face limited onboard computing and scarce ocean infrastructure, complicating computing support across a dynamic multi-layer network. DASH adapts task offloading to real-time states, anticipates satellite handovers, and jointly schedules communication, trajectories, and computing resources. Simulations report at least a 23% reduction in average task delay over benchmarks.

  • Problem

    Vessels have limited onboard computing and operate in ocean areas with scarce infrastructure, while task scheduling across a dynamic heterogeneous network is complex.

    Source: §I · p. 1, §IV · p. 6

  • Method

    DASH uses low-complexity layer-wise offloading based on real-time system states, anticipatory satellite handover, joint resource optimization, and UAV residual-energy constraints.

    Source: §I · p. 2, 2) UAV Energy Model · p. 5, A. Anticipatory Satellite Handover… · p. 6

  • Results

    At least 23% reduction in average task delay over benchmarks, alongside adaptation to varying system resources and reduced post-handover congestion.

    Source: §I · p. 2

  • Takeaways & Limitations

    DASH balances workloads across heterogeneous servers while supporting resource utilization and task backlog handover during satellite transitions.

    Source: §I · p. 2, A. Anticipatory Satellite Handover… · p. 6

Abstract

from arXiv · show

In the context of 6G ubiquitous connectivity, the space-air-ground-sea integrated network (SAGSIN) emerges as a new paradigm for pervasive service provisioning. To support expanding maritime activities in infrastructure-scarce ocean areas, we propose an innovative dynamic task and resource scheduling approach for SAGSIN to deliver computing services for vessels. It integrates broad-coverage satellites, relay-capable high-altitude platform (HAP), energy-sufficient coastal base station (BS), and flexibly deployed uncrewed aerial vehicles (UAVs) to accommodate wide-area, highly mobile, and sustained maritime services. To address the challenge of task scheduling across four layers, a dynamic task offloading algorithm is developed. It steers task flows toward servers with light loads, strong computing capabilities, and high-rate links based on real-time system states to reduce task execution delay, integrating an anticipatory satellite handover strategy to mitigate post-handover congestion and improving satellite resource utilization. Considering the limited endurance of UAVs, we impose residual energy constraints to ensure task backlog handover and safe return. Furthermore, the UAV-BS bandwidth allocation, UAV trajectories, and computing resource allocation are jointly optimized to enhance the connectivity among low-altitude devices and accelerate task completion. Simulation results validate the proposed method's superior adaptability to system resource variations during task execution in complex maritime environments, achieving at least a 23% reduction in average task delay over benchmarks.

I. INTRODUCTION

Maritime services face limited vessel computing and scarce ocean infrastructure, while existing scheduling approaches can lack flexibility under changing network conditions. DASH combines dynamic multi-layer offloading with joint resource optimization; its experiments report lower delay and reduced post-handover congestion.

I. INTRODUCTION

  • DASH dynamically routes tasks using real-time server, congestion, link, and topology states, and anticipates satellite handover using current and incoming satellite conditions.It also jointly optimizes UAV-BS bandwidth, UAV trajectories, and computing resources, while reserving UAV energy for backlog handover and safe return.Source: §I · p. 2
  • At least 23% reduction in average task delay over benchmarks, with workload balancing and reduced post-handover satellite congestion.The experiments also report adaptation to varying system resources.Source: §I · p. 2

A. System Overview

  • The system model organizes vessels, a coastal BS, low-altitude UAVs, a HAP, and current and incoming LEO satellites across sea, ground, air, and space layers.A UAV charging station is included because maritime emergency landings are impractical under low-battery conditions.Source: A. System Overview · p. 2
  • Vessels offload portions of task data to a UAV or coastal BS, while UAVs and the HAP forward buffered data upward and network servers compute buffered tasks.Task size is Dv bits and computational density Cv is measured in CPU cycles per bit.Source: A. System Overview · p. 3

1) Vessel-to-UAV and Coastal BS Communication:

  • Vessel-to-UAV and vessel-to-BS links are modeled with large-scale path loss and small-scale Rician fading, reflecting wave-induced multipath and weather conditions.The cited model uses OFDMA for these links.Source: 1) Vessel-to-UAV and Coastal BS… · p. 3

2) UAV-to-HAP Communication:

The communication model specifies the UAV-to-HAP and HAP-to-satellite links, then treats inter-satellite forwarding as part of handover while focusing on adaptive satellite backlog control.

2) UAV-to-HAP Communication:

  • The UAV-to-HAP link model accounts for atmospheric attenuation, scintillation, free-space path loss, and distance between the UAV and HAP.The passage then introduces the link transmission rate and transmission energy for data offloaded by a UAV.Source: 2) UAV-to-HAP Communication · pp. 3–4

3) HAP-to-Satellite Communication:

  • The HAP-to-satellite model uses large-scale fading with a Gaussian fitting deviation whose standard deviation is ω = 0.1.The passage also specifies the HAP-to-satellite transmission rate and channel gain.Source: 3) HAP-to-Satellite Communication · p. 4

4) Inter-Satellite Communication:

  • At handover, unfinished tasks are forwarded over the inter-satellite link; the study assumes propagation-dominated latency and leaves ISL study for future work.The focus is adaptive satellite backlog control to avoid post-handover congestion or overly conservative offloading.Source: 4) Inter-Satellite Communication · p. 4

C. Task Computation and Queue Models

The section models server resources and task backlogs across time slots, then specifies UAV motion and energy requirements. UAVs must retain enough energy to hand off queued tasks and return to their charging station.

C. Task Computation and Queue Models

  • Server resources are the maximum processable CPU cycles per time slot, while queues track vessel task data locally and at each server.Queue updates distinguish data received, computed, and forwarded between servers.Source: C. Task Computation and Queue Models · p. 4

1) UAV Motion Model:

  • UAV displacement between consecutive time slots is constrained by maximum speed, and UAV pairs must maintain at least the minimum safety distance.The section defines displacement using consecutive UAV positions and uses the safety-distance condition to avoid collisions.Source: 1) UAV Motion Model · pp. 4–5

2) UAV Energy Model:

  • UAVs must reserve residual energy for uploading remaining task backlog to the HAP and returning to the charging station.Return energy depends on the UAV's previous position and autonomous return speed; the model also accounts for flight and computation energy consumption.Source: 2) UAV Energy Model · p. 5
  • A vessel task is counted as complete when its data has been cleared from all system queues; because slot indexing starts at zero, completion takes Nv + 1 slots.Nv is defined by two simultaneous formulas in the section.Source: 2) UAV Energy Model · p. 5

III. PROBLEM FORMULATION

The formulation minimizes total vessel task execution delay by jointly selecting offloading, bandwidth, computing resources, and UAV trajectories. Its coupled variables and operational constraints make global optimization intractable, motivating a low-complexity offloading method followed by joint resource refinement.

III. PROBLEM FORMULATION

  • The objective is to minimize total task execution delay across vessels by jointly optimizing offloading, bandwidth, computing resources, and UAV trajectories.The decision sets include vessel and server offloading choices, UAV and BS bandwidth, server computing allocations, and UAV paths.Source: §III · p. 5
  • The constraints enforce binary offloading choices, available bandwidth and computing capacities, feasible data processing, and UAV residual energy for backlog handover and return.UAV maximum-velocity and collision-avoidance conditions are also imposed.Source: §III · pp. 5–6
  • Global polynomial-time optimization is intractable because four-layer offloading decisions are coupled with bandwidth, computation, UAV mobility, and energy constraints.The proposed solution uses low-complexity task offloading, then refines bandwidth, trajectories, and computing allocations through block coordinate descent.Source: §III · p. 6

IV. DYNAMIC TASK OFFLOADING ALGORITHM FOR SAGSIN

The offloading algorithm anticipates satellite handovers by combining current and incoming satellite states, then schedules task data layer by layer using real-time congestion and link information. This reduces decision complexity while adapting to changing network conditions.

A. Anticipatory Satellite Handover Strategy

  • During a pre-handover period, weighted joint states of current and incoming satellites guide offloading, with the incoming satellite's weight increasing as handover approaches.A larger computing-capability gap between satellites calls for a longer preparation period, and the system enters it immediately if remaining service time is insufficient.Source: A. Anticipatory Satellite Handover… · p. 6

B. Dynamic Layer-Wise Task Scheduling for SAGSIN

  • The algorithm models SAGSIN as a graph with communication and virtual computation links, using a virtual sink for each vessel to represent its task destination.Link rates are expressed as bits transmitted or computed per time slot; the Pressure Index incorporates congestion, computing capability, link rates, and topology, with lower values indicating lighter load or better access to capable servers.Source: B. Dynamic Layer-Wise Task Scheduling… · pp. 6–7
  • Vessels choose a UAV or BS using the strongest pressure differential and link rate, while each UAV and the HAP forward the vessel data with the most urgent offloading demand.This favors low-load, high-throughput links and limits forwarding from servers with low backlogs; pressure indices update each slot, and the stated complexity is approximately O(V U).Source: B. Dynamic Layer-Wise Task Scheduling… · p. 7

V. JOINT UAV-BS BANDWIDTH, UAV TRAJECTORY, AND COMPUTING RESOURCE OPTIMIZATION

The section jointly optimizes UAV/BS bandwidth, UAV trajectories, and server computing resources through three iteratively solved subproblems. Its bandwidth subproblem maximizes transmission rates subject to resource and energy constraints.

V. JOINT UAV-BS BANDWIDTH, UAV TRAJECTORY, AND COMPUTING RESOURCE OPTIMIZATION

  • The method jointly optimizes UAV/BS bandwidth, UAV trajectories, and server computing resources by decomposing the coupled problem into three tractable subproblems solved iteratively using block coordinate descent.The iterations continue until the overall solution converges.Source: §V · p. 8

A. Bandwidth Allocation Policy of UAVs and BS

  • Bandwidth allocation maximizes the total transmission rate of UAV- and BS-to-vessel links, while respecting available spectrum and UAV energy constraints.For the BS, the optimal allocation is characterized through KKT conditions; the UAV allocation is adjusted if needed to satisfy the energy constraint.Source: A. Bandwidth Allocation Policy of UAVs… · p. 8
  • The bandwidth subproblem is convex because its objective is strictly concave in allocated bandwidth and its constraints are convex.A monotonic equation in the allocation can be solved efficiently by bisection.Source: A. Bandwidth Allocation Policy of UAVs… · p. 8

B. UAV Trajectory Optimization

UAV trajectories are optimized to improve aggregate data rate and throughput for associated vessels while respecting kinematic, safety, and residual-energy constraints. The non-convex trajectory problem is handled through successive convex approximations solved iteratively.

B. UAV Trajectory Optimization

  • UAV trajectories position aircraft to maximize aggregate data rate and throughput for associated vessels under kinematic, safety, and residual-energy constraints.Source: B. UAV Trajectory Optimization · p. 8

P2: UAV Trajectory Optimization

C. Computing Resource Allocation

The allocation scheme distributes each server’s computing resources according to vessel backlogs and demands, with an additional energy-constrained scaling stage for UAV servers.

C. Computing Resource Allocation

D. Complexity Analysis

The overall method alternates task offloading and resource-optimization updates, and the analysis characterizes its complexity as polynomial despite the inherently exponential multilayer offloading problem.

D. Complexity Analysis

  • Algorithm 3 obtains offloading decisions, then iteratively updates UAV-BS bandwidth, UAV trajectories, and server computing resources until convergence or the iteration limit.Source: D. Complexity Analysis · p. 10
  • Algorithm 1’s overall complexity is O(VU), according to the analysis of link calculations and offloading decisions.Source: D. Complexity Analysis · p. 10
  • The analysis claims polynomial complexity for the overall method despite the inherently exponential complexity of multilayer task offloading.Source: D. Complexity Analysis · p. 10

VI. SIMULATION RESULTS

The simulation section introduces benchmark comparisons and further evaluations of load balancing, UAV energy management, and handover backlog control.

VI. SIMULATION RESULTS

  • The simulations compare the proposed method with benchmarks and evaluate load balancing, UAV energy management, and handover backlog control.Source: §VI · p. 10

A. Scenario

The simulations compare DASH with four benchmark schemes in a maritime scenario and show lower average task delay as computing-resource fluctuations intensify.

A. Scenario

  • The 2 km×2 km ocean scenario deploys six UAVs, with vessels following random trajectories and baseline bandwidth and computing resources varied around Table II settings.The coastal BS and UAV charging station are near the lower-left corner, and the HAP remains quasi-stationary at the region center.Source: A. Scenario · pp. 10–11, Table II · p. 11

B. Benchmarks

  • DASH is evaluated against its no-handover ablation and HACO, FLEC, and SHAR benchmark schemes.HACO targets task execution delay, FLEC targets system task costs, and SHAR uses reactive handover that reoffloads unfinished task data.Source: B. Benchmarks · p. 11

1) Average task delay under varying computing resources:

2) Average task delay under varying bandwidth resources:

Across bandwidth variation, handover, network scale, server backlogs, and UAV energy, DASH adapts scheduling to changing conditions; the conclusion reports reduced delay and controlled satellite backlog.

2) Average task delay under varying bandwidth resources:

  • Around 27%, 25%, and 23% lower average task delay than SHAR, HACO, and FLEC, respectively, under bandwidth-resource fluctuations.DASH and its no-handover variant adjust offloading and bandwidth to real-time link states, while one-shot schemes cannot adapt to changing channel conditions or bandwidth availability.Source: 2) Average task delay under varying… · pp. 11–12

3) Handover backlog with different computing resource gap of incoming satellite:

  • DASH restricts satellite offloading as incoming-satellite computing capacity falls, reducing handover backlog; SHAR reoffloads unfinished data, including already processed portions.The figure compares handover backlog across methods as the incoming satellite’s computing-resource deficit changes.Source: Figure 6 · p. 12, 3) Handover backlog with different… · p. 12

4) Average task delay with varying number of vessels:

  • As vessel numbers rise, all methods incur higher delay, but DASH maintains the lowest delay and slowest growth trend.Its real-time server-state routing balances heterogeneous-server workloads; binary offloading in HACO and SHAR may overload servers in dense scenarios.Source: Figure 7 · p. 12, 4) Average task delay with varying… · p. 12

1) Task backlog among heterogeneous servers:

  • Satellite backlogs are relatively large, HAP and BS backlogs moderate, and UAV backlogs lighter; queues across layers remain dynamically stable with staggered peaks and valleys.DASH adapts task routing, bandwidth, and computing allocation to balance workloads and process buffered tasks according to demand.Source: Figure 8 · p. 12, 1) Task backlog among heterogeneous… · pp. 12–13

3) Impact of pre-handover period:

  • As a UAV approaches its handover-and-return energy threshold, it withdraws and a replacement takes over; ξ from 10 to 100 reduces delay, while 100 to 1000 adds little benefit and may slightly increase it.Residual energy remains above the required handover-and-return energy, supporting UAV operational safety and service continuity.Source: Figure 10 · p. 13, Figure 9 · p. 13, 3) Impact of pre-handover period · p. 13

VII. CONCLUSIONS

  • The conclusion reports that DASH reduces task delay under fluctuating resources and adaptively controls satellite backlog at handover compared with benchmarks.The method jointly optimizes UAV-BS bandwidth, UAV trajectories, and computing resources while accounting for UAV energy needed for backlog handover and safe return.Source: §VII · p. 13
Loading 2609.31011v1…