چكيده به لاتين
In order to meet the requirements of different user applications, new technologies are presented in next generation communications. Each of these technologies, in addition to the benefits they provide to meet the limitations of applications, also create challenges for these communications. For example, in millimeter waves, by sending at a higher radio frequency, a higher capacity is allocated to the air interface to meet the requirements of low delay and high throughput applications; But, as a challenge, these technologies are very sensitive to the propagation environment and the blocking effect that the transmitter can be in different communication states (even communication interruption). In other words, the effect of blocking and directional sending can lead to rapid changes in the conditions of the communication link. Dual connections are one of the techniques used in fourth and fifth generation cellular networks to increase throughput, where the source node connects to both stations simultaneously. Using this method will be very effective for real-time applications. But just connecting to two stations and increasing throughput will not be a suitable criterion for evaluating and using dual connection technology in real-time applications. In such applications, we need information with low latency, and having information with high latency is practically useless for us and even in critical decision-making times, such as self-driving cars, remote surgery, smart factories, etc., is considered fatal.
In this thesis, a scenario of dual connections based on the TCP multipath transport layer is presented in fourth and fifth generation mobile phone networks so that the self-driving car or the source node can connect to two antennas simultaneously. In order to measure the freshness of information, we used the concept of AoI in this type of connections, because of the random nature of AoI, we used random optimization. Also, due to the change in channel conditions in the fourth generation and especially in the fifth generation of cellular networks, the channel is also modeled randomly. For this reason, we have used stochastic optimization to model the system. Using the difference learning technique, we present an algorithm that can optimize and minimize the information delay in the absence of statistical information of the environment and channel, such as: packet arrival rate, channel condition changes, packet exit rate, etc. That is, the scheduling unit can decide for each generated packet to place it in which of the LTE or mmWave queues so that we have the lowest amount of information delay and deliver new packets to the destination.
