Abstract
The next generation hyper-reliable and low-latency communication (HRLLC) aims to surpass the capabilities of current fifth-generation (5G) wireless communication networks, striving for higher spectral efficiency, energy efficiency, security, reliability, coverage, and lower latency.In this thesis, I propose several methods to meet the requirements of high-reliable and low-delay wireless communication. Firstly, I delve into the realm of hybrid automatic repeat request (HARQ) with erroneous feedback. I introduce an asymmetric feedback detection (AFD)-HARQ approach, which involves quantizing the feedback with a certain bias. This results in extra protection for negative feedback (NACK), thereby improving HARQ performance. I formulate optimization problems to determine optimal quantization thresholds and transmission rates for each transmission. Additionally, a deep Q-network (DQN) is employed to decide whether to re-transmit and the transmission rate for subse- quent transmissions based on the current state and the received feedback. Results indicate that AFD-HARQ achieves high throughput and low delay compared to conventional HARQ processes, and it can significantly lower the outage probability. Furthermore, I propose a redundancy version (RV)-adaptive HARQ scheme that dynamically selects the RV for the following transmission based on current decoding states, demonstrating improvements over fixed RV sequence systems.
In 6G networks, the increased deployment of relays is expected to enhance connectivity and network coverage. To ensure the reliability of the relay network, HARQ remains essential. I present a de-centralized relay selection scheme that efficiently chooses the RN for assisting transmission without excessive overhead. Additionally, I introduce a novel partial compress-and-forward (PCF) scheme by combining existing compress-and- forward (CF) and amplify-and-forward (AF) schemes in a diamond relay network with both RNs being noisy. This novel approach exhibits superior performance when the channels connected to one RN are of high quality.
HRLLC is ideal for mission-critical Internet-of-things (IoT) applications. In IoT, both sensing and communication functionalities are crucial, leading to the consideration of integrating them into one single signal, referred to as integrated sensing and communication (ISAC). I focus on ISAC-based vehicular networks. A digital twin (DT) of the ground system is designed for tracking and predicting the movement of the vehicles and further optimizing the system performance. In a scenario featuring two roadside units (RSUs), I propose algorithms designed to concurrently assign vehicles to RSUs and de- sign predictive beamforming, aiming to maximize the overall throughput. Subsequently, in a single-RSU network with vehicles located in the near-field region, I leverage deep reinforcement learning (DRL) algorithms to strike a balance between sensing and communication functionalities. The agents autonomously decide whether to prioritize sensing or communication, and based on this decision, optimal beamforming is designed. Simulation results demonstrate the effectiveness of the proposed algorithms, showing significant enhancements compared to existing techniques. To be specific, the vehicle-assigning algorithm outperforms conventional distance-based methods. Moreover, the DRL-based beamforming design achieves substantially higher long-term throughput when a vehicle remains in the near-field region of the RSU, alleviating concerns about losing track.
| Date of Award | 1 Oct 2024 |
|---|---|
| Original language | English |
| Awarding Institution |
|
| Supervisor | Mohammed Shikh-Bahaei (Supervisor) |