裝置對裝置通訊(Device-to-Device communications)允許D2D pairs重複使
用CUE(cellular user equipment)之頻譜資源，鄰近的裝置不須透過基地台(Base
Station, BS)即可直接通訊，能提高的頻譜使用效率、服務裝置數量，但重複使
用相同資源的裝置間會相互干擾，所以藉由無線通訊資源分配以及傳輸功率控
制，避免反而降低頻譜使用效率。
D2D常應用在行動裝置和的通訊，這類裝置多是由電池提供電源，但電池容
量有限，若電量消耗太快，則裝置運行時間會變短。因此，這篇論文將探討在
滿足所有裝置的最小速率要求之下，使得系統能源效率最大化的問題。
能源效率問題是一個分式非線性問題，沒有辦法直接得到最佳解，我們提
出一個迭代式的功率控制演算法，利用Dinkelbach's method，並且在每一次迭
代中把問題連續近似成幾何規劃(geometric programming, GP)問題。因為前述
的方法花費較多時間，我們嘗試使用機器學習(Machine Learning)來協助解決問
題， 並且改寫一個現存的能源效率最佳化演算法，來快速產生大量的訓練資料
集(Training Data Set)。
根據實驗的結果，我們的方法不僅在主要目標能源效率上表現良好，整體的
系統速率總和也有很大的改進。我們也可以看到加入機器學習之後， 對於運算
時間和效果的影響。

D2D (Device-to-Device) communications allow D2D pairs to reuse cellular user
equipment's (CUE) radio resources, and to communicate directly without routing
through the base station (BS). The UEs that reuses the same radio resources will
cause mutual interference, so we need better radio resources allocations and transmit
power control. Additionally, D2D communications usually apply to mobile
devices, if the power consumption is too high, the lifetime of devices and system
will be short.
In this thesis, we propose an iterative algorithm that utilizes optimization
techniques to maximize energy efficiency (EE). The EE maximization problem
is a fractional nonlinear programming problem and cannot get the optimal solution
directly, so we use the Dinkelbach's method, and successively approximate the
problem to a geometric programming (GP) problem. However, the aforementioned
optimization algorithm is time-consuming, so we try to use machine learning to
help solve the problem quickly. We modify an existing EE optimization algorithm
to generate large training data set rapidly. Simulation results show that our
method not only improves the system EE, but also reaches high system sum rate.
We also see the impact of machine learning on running time and performance.

Abstract i
Acknowledgements iii
Contents iv
List of Figures vi
List of Tables vii
1 Introduction 1
2 Related Work 3
3 System Model 6
3.1 System model 6
3.2 Problem formulation 8
4 Pure Optimization 12
4.1 Geometric programming 12
4.2 Objective function transformation 14
4.3 Successive GP approximation 16
4.4 Initial power allocation generation 18
4.5 Pure optimization algorithm 19
5 Machine Learning Assistance 21
5.1 Machine learning features creation 22
5.2 Training set creation 23
5.3 Machine learning phases and fast algorithm 26
6 Simulations 29
6.1 Compared algorithm 29
6.1.1 Optimal resource allocation algorithm 29
6.1.2 Greedy throughput maximization algorithm 30
6.1.3 Maximum independent set based and Stackelberg power based algorithm 31
6.1.4 Two-Stage resource sharing optimization 31
6.1.5 Energy efficient stable matching algorithm 31
6.2 Simulation settings 32
6.3 Simulation results 34
7 Conclusion 41
Reference 42

[1] D. Feng, L. Lu, Y. Yuan-Wu, G. Y. Li, G. Feng, and S. Li, “Device-to-device communications underlaying cellular networks,” IEEE Transactions on Communications, vol. 61, no. 8, pp. 3541-3551, Aug. 2013.
[2] F. Hajiaghajani, R. Davoudi, and M. Rasti, “A joint channel and power allocation scheme for device-to-device communications underlaying uplink cellular networks,” in 2016 IEEE Conference on Computer Communications Workshops (INFOCOM WKSHPS), IEEE, Apr. 2016, pp. 768-773.
[3] S. Shen and T.-M. Lok, “Power control for underlay in-band spectrum sharing in device-to-device communications,” in 2014 IEEE Wireless Communications and Networking Conference (WCNC), IEEE, Apr. 2014, pp. 1350-1355.
[4] H. Sun, M. Sheng, X. Wang, Y. Zhang, J. Liu, and K. Wang, “Resource allocation for maximizing the device-to-device communications underlaying lte-advanced networks,” in 2013 IEEE/CIC International Conference on Communications in China-Workshops (CIC/ICCC), IEEE, Aug. 2013, pp. 60-64.
[5] S. A. Ciou, J.-C. Kao, C. Y. Lee, and K.-Y. Chen, “Multi-sharing resource allocation for device-to-device communication underlaying 5g mobile networks,” in IEEE Annual International Symposium on Personal, Indoor, and Mobile Radio Communications (PIMRC), IEEE, Aug. 2015, pp. 1509-1514.
[6] S.-H. Lin, K.-Y. Chen, J.-C. Kao, and Y.-F. Hsiao, “Fast spectrum reuse and power control for device-to-device communication,” in IEEE Vehicular Technology Conference (VTC Spring), IEEE, Jun. 2017, pp. 1-5.
[7] S.-H. Lin, “Two-stage resource sharing optimization for d2d communication underlying cellular networks,” Master’s thesis, National Tsing Hua University, Jul. 2017.
[8] H.-C. Lin, “Joint channel allocation and power control for device-to-device communication underlaying downlink cellular networks: Clique-based approaches,” Master’s thesis, National Tsing Hua University, Jul. 2017.
[9] D. Wen, G. Yu, and L. Xu, “Energy-efficient mode selection and power control for device-to-device communications,” in 2016 IEEE Wireless Communications and Networking Conference (WCNC), IEEE, Apr. 2016, pp. 1-7.
[10] Z. Zhou, G. Ma, C. Xu, Z. Chang, and T. Ristaniemi, “Energy-efficient resource allocation in cognitive d2d communications: A game-theoretical and matching approach,” in 2016 IEEE International Conference on Communications (ICC), IEEE, May 2016, pp. 1-6.
[11] W. Wei, Q. Wang, L. Yang, and X. Hu, “Auction based energy-efficient resource allocation and power control for device-to-device underlay communication,” in 2016 IEEE Vehicular Technology Conference (VTC-Fall), IEEE, Sep. 2016, pp. 1-6.
[12] Z. Zhou, M. Dong, Z. Chang, and B. Gu, “Combined centralized and dis-
tributed resource allocation for green d2d communications,” in 2015 IEEE/CIC International Conference on Communications in China (ICCC), IEEE, Nov. 2015, pp. 1-6.
[13] K. Balachandran, J. H. Kang, K. Karakayali, and K. M. Rege, “Nice: A network interference cancellation engine for opportunistic uplink cooperation in wireless networks,” IEEE Transactions on Wireless Communications, vol. 10, no. 2, pp. 540-549, Feb. 2011.
[14] R. S. Sutton and A. G. Barto, Reinforcement learning: An introduction. MIT press, Feb. 1998.
[15] S. Nie, Z. Fan, M. Zhao, X. Gu, and L. Zhang, “Q-learning based power control algorithm for d2d communication,” in 2016 IEEE 27th Annual International Symposium on Personal, Indoor, and Mobile Radio Communications (PIMRC), IEEE, Sep. 2016, pp. 1-6.
[16] Z. Fan, X. Gu, S. Nie, and M. Chen, “D2d power control based on supervised and unsupervised learning,” in 2017 3rd IEEE International Conference on Computer and Communications (ICCC), IEEE, Dec. 2017, pp. 558-563.
[17] F. Wang, C. Xu, L. Song, Z. Han, and B. Zhang, “Energy-efficient radio resource and power allocation for device-to-device communication underlaying cellular networks,” in 2012 International Conference on Wireless Communications & Signal Processing (WCSP), IEEE, Oct. 2012, pp. 1-6.
[18] W. Dinkelbach, “On nonlinear fractional programming,” Management science, vol. 13, no. 7, pp. 492-498, Mar. 1967.
[19] S. Boyd and L. Vandenberghe, Convex optimization. Cambridge university press, Mar. 2004.
[20] M. Grant and S. Boyd, CVX: Matlab software for disciplined convex programming, version 2.1, http://cvxr.com/cvx, Dec. 2017.
[21] C.-Y. Chi, W.-C. Li, and C.-H. Lin, Convex optimization for signal processing and communications: from fundamentals to applications. CRC Press, Feb. 2017.
[22] M. Klugel and W. Kellerer, “Determining frequency reuse feasibility in device-to-device cellular networks,” in 2015 IEEE 26th Annual International Symposium on Personal, Indoor, and Mobile Radio Communications (PIMRC), IEEE, Aug. 2015, pp. 1503-1508.
[23] E. Alpaydin, Introduction to machine learning. MIT press, Sep. 2014.
[24] F. Pedregosa, G. Varoquaux, A. Gramfort, V. Michel, B. Thirion, O. Grisel,
M. Blondel, P. Prettenhofer, R. Weiss, V. Dubourg, J. Vanderplas, A. Passos,
D. Cournapeau, M. Brucher, M. Perrot, and E. Duchesnay, “Scikit-learn: Machine learning in Python,” Journal of Machine Learning Research, vol. 12, pp. 2825-2830, 2011.
[25] L. Breiman, “Random forests,” Machine learning, vol. 45, no. 1, pp. 5-32, Oct. 2001.