Machine Learnig-based Spatial Reuse for Wireless Networks

Saurabh Kumar Verma


This work studies multi-armed bandits formulation based Spatial Reuse to improve the performance of wireless network in dense deployed areas. Due to immense popularity of wireless network, Dense and uncoordinated deployments of wireless networks are typical in next-generation, resulting in inefficient resource utilization and poor performance of WNs. This work propose a method to enable Spatial Reuse (SR) to solve this problem using entirely decentralized systems. This work concentrate on dynamic channel allocation (DCA) and Transmission Power Control (TPC) to allow networks to learn their ideal configuration. I use Reinforcement Learning (RL), specifically the Multi-Armed Bandits (MABs)[1] algorithm, to solve this problem. In this research, I examine the performance of the epsilon-greedy, EXP3[2], UCB[3], and Thompson sampling[4] action selections to study explorationexploitation trade-offs. Furthermore, I investigate the consequences of selecting actions concurrently in an adversarial context (i.e., concurrently) and compare it to a sequential method. My results show that even when learners have no information about surrounding networks and Wireless Networks (WNs) act selfishly, the network can achieve optimal proportional fairness. Individual networks’ throughput is variable over time, particularly for ϵ-greedy and EXP3. Unlike UCB and Thompson sampling, these techniques operate based on the total experienced reward rather than its distribution. Contributions to CNI I have attended various sessions organized by the CNI department IISc Banglore. I have demonstrated MTech thesis work on “ Machine Learning-Based Spatial Reuse in Wireless network” in Mid review. References

[1] Sébastien Bubeck and Nicolo Cesa-Bianchi. Regret analysis of stochastic and nonstochastic multi-armed bandit problems. arXiv preprint arXiv:1204.5721, 2012.

[2] Peter Auer, Nicolo Cesa-Bianchi, Yoav Freund, and Robert E Schapire. Gambling in a rigged casino: The adversarial multi-armed bandit problem. In Proceedings of IEEE 36th Annual Foundations of Computer Science, pages 322–331. IEEE, 1995.

[3] Peter Auer, Nicolo Cesa-Bianchi, and Paul Fischer. Finite-time analysis of the multiarmed bandit problem. Machine learning, 47(2):235–256, 2002.

[4] William R Thompson. On the likelihood that one unknown probability exceeds another in view of the evidence of two samples. Biometrika, 25(3-4):285–294, 1933.