Energy Efficient Resource Allocation in UAV-Enabled Mobile Edge Computing Networks
read more
Citations
Terahertz Band: The Last Piece of RF Spectrum Puzzle for Communication Systems
Multi-Agent Reinforcement Learning Based Resource Management in MEC- and UAV-Assisted Vehicular Networks
Completion Time and Energy Optimization in the UAV-Enabled Mobile-Edge Computing System
Multi-Agent Deep Reinforcement Learning-Based Trajectory Planning for Multi-UAV Assisted Mobile Edge Computing
A Vision and Framework for the High Altitude Platform Station (HAPS) Networks of the Future
References
Wireless communications with unmanned aerial vehicles: opportunities and challenges
A Survey on Mobile Edge Computing: The Communication Perspective
Optimal LAP Altitude for Maximum Coverage
Energy-Efficient UAV Communication With Trajectory Optimization
Joint Trajectory and Communication Design for Multi-UAV Enabled Wireless Networks
Related Papers (5)
Joint Trajectory and Communication Design for Multi-UAV Enabled Wireless Networks
Energy Minimization for Wireless Communication With Rotary-Wing UAV
Frequently Asked Questions (19)
Q2. What have the authors stated for future works in "Energy efficient resource allocation in uav-enabled mobile edge computing networks" ?
The optimization problem for UAVenabled MEC network, where UAVs are served as UEs, is left for their future work.
Q3. How many CPU cycles is used to satisfy the latency constraints?
6. It is observed that the sum power of the network increases with the data size for all algorithms since more data needs to be computed and more transmission power of the UEs is used to satisfy the latency constraints.
Q4. How is the complexity of solving Problem (49) with fixed j?
Since Problem (49) with fixed θj is convex and the number of variables of this convex problem is three, the complexity of solving Problem (49) with fixed θj is small and can be neglected.
Q5. What is the sum power of the proposed IACL algorithm?
Since the proposed IACL algorithm can fully utilize the optimization of latitude and beamwidth, the increased power of UEs for high data rate by IACL is smaller than that by SCAFAH.
Q6. How can the authors solve a problem that is decoupled into subproblems?
To solve {fij}j∈M,i∈Nj , Problem (39) can be decoupled into M subproblems since both the objective function and constraints can be decoupled.
Q7. How many UEs can be obtained by solving the KKT conditions of Problem (53)?
(Yl − yi)2 + (Hminl )2) − 1m−1 ,(54) for all i ∈ S1, j ∈ M, which can be obtained by solving the KKT conditions of Problem (53) with fixed Z̄ .
Q8. What is the optimal solution to the problem?
6: Update Lagrange multipliers{βi}i∈N , {γi}i∈N , {λj}j∈M, {µj}j∈M based on (34)-(37).7: until the objective function (29a) converges 8: Denote (A(n+1), f (n+1)) as the optimal solution of Problem (29).
Q9. What is the problem that is hard to solve?
Problem (24) is hard to be solved due to non-smooth `0- norm, which can be approximately solved via a sequence of weighted `1-norm minimizations in compressive sensing2`0-norm is usually used for vectors, and scalar can be viewed as a special case of vector with one dimension.5 according to [47].
Q10. What is the auxiliary vector variable in Problem 29?
Note that f in Problem (29) is an auxiliary vector variable, which helps us designthe Lagrangian dual decomposition method to get integer solutions.
Q11. Why does the sum power increase with the number of CPU cycles?
This is because large number of the CPU cycles requires the UAVs and UEs to allocate high computation capacity to meet the latency constraints, which leads to high power consumption according to (24a).
Q12. What is the algorithm to solve the problem?
Based on [50], an iterative algorithm is proposed to solve Problem (53) via optimizing Ā with fixed Z̄ and updating Z̄ with given Ā. Specifically, given location Z̄ , the optimal association isaij = ((Xj − xi)2 + (Yj − yi)2 + (Hminj )2) − 1m−1∑M l=1((Xl − xi)2 +
Q13. How is the complexity of solving a problem decoupled?
To solve each subproblem (42), the complexity is O(N log2(1/ 1)) log2(1/ 2), where O(1/ 1) is the complexity of obtaining the inverse function h−1ij (·), and O(1/ 2) is the complexity of solving (45) or (47) via the bisection method.
Q14. Why is the sum power of the network reduced with the increase of the maximum latency?
This is because the propulsion power of all the UAVs is the dominant part and the transmission power of the UE is slightly reduced even for high computation capacity of the UAVs according to latency constraints (24d).
Q15. How many iterations of the proposed algorithm are sufficient?
It can be seen that the proposed algorithm converges rapidly, and only three iterations are sufficient to converge, which shows the effectiveness of the proposed algorithm.
Q16. What is the maximum number of UEs for each UAV?
The authors assume equal MEC parameters for all UEs (i.e., Di = D, Fi = F , ∀i ∈ N ), equal maximal number of associated UEs for all UAVs (i.e., Uj = U , ∀j ∈ M), and equal maximal computation capacity for all UAVs (i.e., f uavj,max = f uav max, ∀j ∈ M).
Q17. What is the optimal value of aij for the UE?
In Steps 7-15, the authors associate the UE with the UAV using the maximal value of aij obtained from solving Problem (53) if maximal UE number constraint and computation capacity constraint of this UAV can be satisfied.
Q18. What is the complexity of solving problem (48)?
In8 Algorithm 1, the complexity of optimizing user association A and auxiliary vector f is O(MN) according to (30)- (31), and the complexity of updating Lagrange multipliers ({βi}i∈N , {γi}i∈N , {λj}j∈M, {µj}j∈M) is also O(MN) according to (34)-(37).
Q19. What is the optimal (t) for the UEs?
In Algorithm 4, nj and Nj respectively denote the number and set of UEs associated with UAV j, and9 Algorithm 4: FCM Clustering Based Algorithm1: Set the initial location Z̄ (0), iteration number t = 1, nj = 0, Nj = ∅, Sj = 0, ∀j ∈M. 2: repeat 3: With fixed Z̄ (t−1), obtain the optimal Ā(t) according to (54).