Abstract—Next-generation wireless network systems and technologies provide a new paradigm for achieving the fastest access to any network. But one of the significant design concerns is the support of handoff, irrespective of the services. The key objective of this work is to enable a node to make appropriate decisions for performing handoff through Reinforcement learning. The work concentrates on the handoff decision phase for choosing the best network with a minimum delay during the handoff process. The reduction in decision delay has been achieved by minimizing the number of handoffs. The environment is modeled as a Markov decision process with the aim of increasing the total anticipated reward per link. The network resources that are used by the link is taken by a reward function and network switching cost that is utilized to model the signaling and processing load incurred on the network during handoff. It has been shown that the total number of unnecessary handoffs can be decreased enhancing the performance of heterogeneous networks. Also, an assessment of the proposed scheme with the existing Vertical handoff decision algorithm like the Simple Additive Weighting method (SAW) has been made and the results show an improved performance over SAW.
 

Index Terms—Reinforcement learning, expected reward, vertical handoff, access point, value iteration algorithm, MDP.

Cite: Hemavathi and S. Akhila, "Reinforcement Learning Based Vertical Handoff Decision Algorithm for Next Generation Wireless Network," Journal of Communications vol. 16, no. 12, pp. 566-575, December 2021. Doi: 10.12720/jcm.16.12.566-575