Intelligent Handover using User-Mobility Pattern Analysis for 5G Mobile Networks.

Millimeter Waves (mmWaves) offer gigabit data rates in fifth generation (5G) mobile networks. The challenge however is mmWaves are highly sensitive to user mobility and topographic dynamics. Sensitivity enhances occurrences of irregular Line of Sight (LOS) and non-LOS spots in the network that distort cell patterns causing instances of gradual and abrupt changes to user data rates. To maintain desired quality of service, Handoff (HO) schemes need to be versatile and intelligent to avoid target links likely to abruptly fail. Unfortunately, most HO schemes cannot forecast instances of abrupt changes in targeted links. This leads to too early, delayed or wrong HOs, at worst actual link failure. To mitigate the HO challenges, this paper adopts a Jump-Markov-Linear-System HO model. The scheme learns to forecast the likely deterioration pattern of the target mmWave link using maximum likelihood approximation. This in turn helps switch to links likely to remain in LOS longer/gradually deteriorate between HOs. Thus, the scheme does not just select target links with higher data rate/Signal but considers there reliability too past a HO. Simulation results show selection of more reliable links than in classic schemes: users remain connected longer between HOs.