PaperlayerView exact revision on arXiv

Source-linked AI summary

Mobile Edge Computing for Cellular-Connected UAV: Computation Offloading and Trajectory Optimization

Xiaowen Cao, Jie Xu, Rui Zhang

arXiv:1803.03733v1cs.ITmath.IT

TL;DR

Unlike prior work that uses a UAV as a mobile edge server, this paper studies a UAV offloading computation to multiple ground base stations while jointly optimizing its trajectory and offloading schedule. It proposes an efficient algorithm using alternating optimization and successive convex approximation. Numerical results show significant mission-completion-time gains over benchmark schemes.

  • Problem

    Unlike prior work using a UAV as a mobile edge server, this paper studies a cellular-connected UAV offloading tasks to multiple ground base stations, whose distributed resources it can use simultaneously.

    Source: §I · pp. 1–2

  • Method

    The paper jointly optimizes the UAV trajectory and offloading schedule to minimize mission completion time, using alternating optimization and successive convex approximation.

    Source: §V · p. 5

  • Results

    Numerical results show significant mission-completion-time gains for the proposed design over benchmark schemes.

    Source: §V · p. 5

  • Takeaways & Limitations

    Cellular base stations can provide UAVs seamless communication and ubiquitous computation services, helping increase their operation range and enlarge their application horizon.

    Source: §I · p. 1

    The solution is generally suboptimal because, for a given mission duration, the algorithm obtains only a locally optimal solution.

    Source: 3) Complete Algorithm to Solve (P3) · p. 4

Abstract

from arXiv · show

This paper studies a new mobile edge computing (MEC) setup where an unmanned aerial vehicle (UAV) is served by cellular ground base stations (GBSs) for computation offloading. The UAV flies between a give pair of initial and final locations, during which it needs to accomplish certain computation tasks by offloading them to some selected GBSs along its trajectory for parallel execution. Under this setup, we aim to minimize the UAV's mission completion time by optimizing its trajectory jointly with the computation offloading scheduling, subject to the maximum speed constraint of the UAV, and the computation capacity constraints at GBSs. The joint UAV trajectory and computation offloading optimization problem is, however, non-convex and thus difficult to be solved optimally. To tackle this problem, we propose an efficient algorithm to obtain a high-quality suboptimal solution. Numerical results show that the proposed design significantly reduces the UAV's mission completion time, as compared to benchmark schemes.

I. INTRODUCTION

Cellular-connected UAVs need greater computation capability for latency-critical tasks despite limited onboard resources. The paper proposes offloading to multiple GBSs while jointly optimizing the UAV’s route and offloading schedule to reduce mission time.

I. INTRODUCTION

  • UAVs face latency-critical computation tasks but have limited onboard resources because of size, weight, and power constraints.The introduction identifies enhancing UAV computation capability as an open problem.Source: §I · p. 1
  • MEC lets cellular-connected UAVs offload intensive tasks to GBSs, which can provide ubiquitous computation services alongside seamless communication.The authors state this may help increase UAV operation range and enlarge its application horizon.Source: §I · p. 1
  • UAVs’ LoS links to many GBSs and controllable 3D mobility support simultaneous multi-GBS offloading with jointly designed trajectories and schedules.This differs from prior fixed-terrestrial-user MEC studies –.Source: §I · p. 1
  • The UAV partitions tasks into subtasks for parallel execution at selected GBSs along its route, while minimizing mission time under speed, endpoint, and GBS-capacity constraints.Source: §I · p. 1
  • Unlike prior UAV-MEC work where the UAV serves as a moving edge server for ground users, this paper treats the UAV as the mobile user offloading to multiple GBSs.Source: §I · p. 2

II. SYSTEM MODEL

The system models one cellular-connected UAV offloading partitionable tasks to multiple GBSs during a fixed-altitude flight between specified endpoints. It discretizes the mission and jointly optimizes duration, trajectory, and offloading times subject to communication, execution, and flight constraints.

II. SYSTEM MODEL

  • The model has one cellular-connected UAV and K MEC-enabled GBSs; task-input bits are partitionable for parallel execution, and download time is omitted as negligible.The model also assumes the UAV does not compute locally for exposition, given its limited resources.Source: §II · p. 2
  • The UAV flies at fixed altitude H between projected horizontal endpoints, with GBSs at fixed horizontal locations and mission duration T to be optimized.Source: §II · p. 2
  • The mission is divided into N sufficiently short slots of duration δt, so UAV location is approximately constant per slot and displacement is bounded by δtVmax.Source: §II · p. 2
  • TDMA divides each slot into GBS-specific subslots for offloading, with the UAV transmitting at constant power P.Source: §II · p. 2
  • Each GBS has per-slot execution capacity fkδt, can start processing received bits in the next slot, and must finish all offloaded bits before mission completion.The cumulative constraints enforce that earlier arrivals fit within the remaining execution capacity; no bits can be offloaded in the final slot.Source: §II · p. 3
  • The objective minimizes N by optimizing trajectory and offloading times under speed, endpoint, TDMA, and task-execution constraints; the formulation is non-convex and has an a-priori unknown number of constraints.Source: §II · p. 3

III. PROPOSED SOLUTION TO (P1)

The solution converts mission-time minimization into fixed-duration feasibility checks and searches over duration by bisection. For each candidate duration, it maximizes executable task input and alternates offloading-time and trajectory optimization.

III. PROPOSED SOLUTION TO (P1)

  • For a given N, the method checks feasibility and uses bisection over N to locate the minimum feasible mission duration.A feasible fixed-N problem implies the optimal N is no larger than N; infeasibility implies it is larger.Source: §III · p. 3
  • For fixed N, feasibility is tested by maximizing the number of task-input bits that can be completed and checking whether the maximum reaches L.Source: §III · p. 3
  • The fixed-duration task maximization remains non-convex, so the algorithm alternates between optimizing offloading times and the UAV trajectory.The trajectory and time-allocation subproblems are handled in alternating manner.Source: §III · p. 3

1) Time Allocation for (P3) Under Given UAV Trajectory:

With the UAV trajectory fixed, the time-allocation subproblem becomes a linear program that maximizes the executable task-input bits.

2) UAV Trajectory Optimization for (P3) Under Given Time Allocation:

For a fixed time allocation, the trajectory subproblem is handled with successive convex approximations, then incorporated into an alternating procedure that updates trajectory and offloading allocation.

2) UAV Trajectory Optimization for (P3) Under Given Time Allocation:

3) Complete Algorithm to Solve (P3):

The complete procedure checks feasibility for each candidate mission duration and uses bisection to solve the original problem, but is generally suboptimal.

3) Complete Algorithm to Solve (P3):

  • The bisection-based solution to (P1) is generally suboptimal because each fixed-duration solve obtains only a locally optimal solution to (P3).The authors report that this suboptimal solution performs well in practice in the numerical results.Source: 3) Complete Algorithm to Solve (P3) · p. 4

IV. NUMERICAL RESULTS

Numerical comparisons show that the joint design performs best across task sizes, with its advantage depending on whether flight distance or computation execution limits mission time.

IV. NUMERICAL RESULTS

  • The simulation uses five GBSs in a 1 × 1 km2 area, 1 MHz bandwidth, 50 m altitude, 50 m/s maximum speed, and 2.5 GHz maximum CPU frequency per GBS.Source: §IV · p. 4
  • 100 Mbits: straight flight matches the proposed design; above 200 Mbits, the proposed design outperforms both benchmarks and is best across all tested L values.Source: Figure 3 · p. 5, §IV · p. 5
  • At 100 Mbits the UAV flies straight at maximum speed; at 200 Mbits it approaches GBSs 1, 4, and 5, and at 500 Mbits it visits all five and shuttles between GBSs 4 and 5.Source: Figure 3 · p. 5
  • At small task sizes, flight distance limits mission time, whereas larger tasks make computation execution the main constraint; the proposed path can time-share GBS resources by flying between them.Source: Figure 3 · p. 5, §IV · p. 5
  • The comparison uses straight flight and successive hover-and-fly as benchmarks; the latter visits each GBS at maximum speed and hovers for offloading.Source: §IV · p. 5

V. CONCLUSION

The proposed joint trajectory and offloading design significantly outperforms benchmark schemes in mission-completion time, using an efficient suboptimal algorithm under UAV and GBS constraints.

  • The proposed design achieves a significant performance gain over benchmark schemes in numerical results.Source: §V · p. 5
  • Alternating optimization and successive convex approximation (SCA) yield an efficient algorithm for a suboptimal solution.Source: §V · p. 5
  • The design jointly optimizes UAV trajectory and offloading scheduling to minimize mission completion time, subject to UAV speed and location constraints and GBS computation capacities.Source: §V · p. 5

Related papers

Loading 1803.03733v1…