The fifth (5G) generation mobile network adopted the LTE-mmWave architecture to provide a Heterogeneous mobile network environment. The LTE-mmWave architecture is flexible, and often reliable. It provides sufficient bandwidth and works over a diverse topography. It is also cost effective for mobile companies transitioning from 4G to 5G. This architecture, however, faces a lot of operational and optimization challenges. Particularly, despite its abundant bandwidth hence high data rate, it sometimes provides patchy coverage owing to mmWave propagation characteristics. mmWave propagation characteristics exhibit high propagation losses and blockage sensitivity. Further, users located just a couple of hundred meters away from mmWave gNBs (BSs) and in NLOS scenarios can be noise limited. In fact, for certain communication scenarios, using more bandwidths of mmWave bands has proven to be counterproductive. It is found out that channel estimation penalties end up exceeding the gains of using more bandwidth. This behavior of mmWaves is in contrast with that of legacy cellular systems using only LTE links. Usually, using more microwave bandwidth, for instance in LTE networks favors longer transmission range and high data rates. Given such conflicting LTE and mmWave performance factors, and as mobile networks transition from 4G to 5G, smart Handoffs (HO) are needed in the LTE and mmWave HetNet to ensure seamless, uniform and continuous communication for 5G’s mobile users are attained. Unfortunately, most classic HO schemes are unable to take into consideration the vulnerability or the outage unpredictability of mmWaves. Moreover, in spite the 4G to 5G technology transition, a lot of HO schemes were optimized for old generation mobile networks using microwaves and other radio bands. Furthermore, while recent studies show that smart HO algorithms using AI are being introduced in 5G communication and beyond, they hardly consider the coexistence and competition of multiple performance deterioration factors and parameters in microwave and mmWave links. This leads to HO failures including too-early or too-late handoff, a handoff to the wrong cell or Ping-Pong handoff decisions. The after-effects of these HO decisions include low data rates, high latency, underutilization of mmWave Bandwidth and energy inefficiency of 5G among other. This research thus explored the suitability of using intelligent/ self-learning schemes to sustain 5G mobile network connectivity and performance for mobile users. The study thus reflects on the following:

(i) The prospects of using distance-dependent and mobility Pattern HO schemes to sustain 5G mmWave mobile network connectivity longer. (ii) The prospects of using Deep Reinforcement Learning with Jump Markov modeling to predicting not just the immediate behavior but also the abrupt and gradual performance changes of the mmWave links before HO execution. (iii) Using mean field Game Theory and Jump Markov Modeling to predict connectivity sustainability of a target link after HO execution. The scheme extends Markov principles into Game Theory to prediction of random and abrupt mmWave link changes. (iv) The prospects of route selection schemes in autonomous cars using not just the shortest distance when selecting the best route to destination but also considering 5G mmWave network connectivity guarantees. This ensures mmWave network availability on chosen routes is continuous and improve safety/infotainment data transmission.

Various network simulators are examined to build realistic scenario but for this research, the NS-3 simulator was picked due to its versatility. The NS-3 simulator can integrate with external AI and route mobility applications, e.g., Open AI Gym and Google Maps. This made the simulations closer to more realistic scenarios encountered in mobile network operation. The performance parameters studied are throughput, latency, HO failure rate, energy and spectral efficiency, etc. From the results, we show that the rate of convergence of our learning schemes and making of HO decisions is faster as a result of utilizing, for instance, jump Markov modelling. This is deduced from the increase in dwell time, reduction of HO failure rate, improvement of energy and spectral efficiency and data rate. In conclusion, the reliability of mmWave connectivity hence 5G mobile access can be improved using smart HO schemes. Furthermore, smart algorithms that can seamlessly be integrated together to be smarter in terms of data rate, HO and energy efficiency if optimized.