Loading...

Using Reinforcement Learning (RL)for Rate Splitting Multi Access (RSMA) assisted Integrated Sensing and Communication (ISAC)System

Ayoubi, Bahare | 2024

0 Viewed
  1. Type of Document: M.Sc. Thesis
  2. Language: Farsi
  3. Document No: 57832 (05)
  4. University: Sharif University of Technology
  5. Department: Electrical Engineering
  6. Advisor(s): Behrouzi, Hamid; Hossein Khalaj, Babak
  7. Abstract:
  8. Rate splitting multiple access (RSMA) is an efficient approach that divides massages in space and power domain in specific time and frequency in order to control interference. Therefore, in this approach, the goal is linear precoding and power allocation on the transmitter side and successful interference cancellation (SIC) on the receiver side. On the other hand, integrated sensing and communication (ISAC) is an emerging technology in the new generation of communications (six generation) in which the process of sensing and communication can be done by same hardware and in a same frequency band so it can be more cost-effective. we can use RSMA in ISAC system in the applications which we aim to provide common and private services for users and sensing targets in the same time. Generally, the achievement of optimal communication and sensing in this system is ultimately achieved by solving a non-convex optimization problem which is generally done using mathematical relationships and complex simplifications. On the other hand, in the dynamic environment conditions, changing channel conditions over time makes solving these problems more complicated. So, in this project, we investigate Deep Reinforcement Learning (DRL) methods to solve such problems. we first dealt with the introduction chapter and in the second chapter we discuss about the model of ISAC system and the application of the RSMA approach in these systems which results in the relevant optimization problem based on the metric considered for evaluating the quality of communication and sensing. In the third chapter, we will look into important algorithms for policy optimization such as Deep Deterministic Policy Gradient (DDPG) and Trust Region Policy Optimization (TRPO) and examine the structure of these methods for solving problems. Finally, in the fourth chapter, we will use these methods to solve the optimization problem raised in the second chapter. At first, we show that the DDPG method is not able to solve these problems due to the complexity of the environment and the large number of hyperparameters, then we use TRPO algorithm to solve this problem and shows the function of this method to solve the proposed problem in a dynamic environment for variable channel conditions. In the following, we consider another scenario by assuming a higher bound for the sum of the square errors of the beam pattern and examine the performance of this method compared to the previous case
  9. Keywords:
  10. Rate Splitting Multiple Access ; Deep Reinforcement Learning ; Deep Deterministic Policy Gradient ; Integrated Sensing and Communication (ISAC) ; Trust Region Policy Optimization (TRPO)

 Digital Object List

 Bookmark

No TOC