Research On Position Optimization Of Mobile Relay Based On One-bit Feedback

Cooperative communication technology, such as relay technology, is considered one of key technologies in the future wireless communication system, because it takes low cost to improve the communication range and the transmission rate. After years of research and development, the traditional relay technology with fixed position has become gradually mature. However, the research on mobile relay technology is rare and it becomes a new research topic. Although the mobile relay technology has many advantages, there are many problems required to be solved in practice due to its dynamic property, such as channel modeling and estimation, power and bandwidth allocation, optimization of relay position. This thesis researches on position optimization of mobile relay.The existing algorithms include two main categories. The first one is ESC(Extremum Seeking Control). This kind of algorithm is suitable for the scenario in which the destination node has enough power to feedback signal frequently, e.g., Base Station(BS), but its obvious defect is there is no any error in the feedback signal. The more complex the feedback signal is, the more the network resource is required to support its accuracy. The second one is multiple on-board antennas. Although there is no feedback signal required in the second kind algorithm, multiple antennas is required in the mobile relay, which increases the requirement on the size and cost of mobile relay. This thesis mainly researches on the position optimization of mobile relay system based on one bit feedback signal, in which multiple antennas is not required. Moreover, only one bit signal is required to feedback from the destination node to mobile relay, which requires little network resource to ensure its accuracy. The proposed system achieves position optimization of mobile relay with low cost and low complexity.This thesis further proposes two implementation algorithms in two scenarios, i.e., fixed orbit scenario and global optimization of mobile relay position scenairo. Obviously, the algorithms in the first scenario are suboptimal solutions, but it has faster convergence speed with low complexity and, more importantly, it can be extended to multiple mobile relay system. On the contrary, although the algorithms in the second scenario have better performance, they require more complexity and have slower convergence speed, which are difficult to be extended to multiple mobile relay system. For the first scenario, we can directly utilize some closed-loop phase synchronization algorithm in the traditional distributed antenna system to achieve the position optimization, and propose two new algorithms, where they are referred to as the perturbation algorithm with fixed step-size and the perturbation algorithm with dynamic step-size by utilized successive negative feedback(Dynamic step-size with SNF), respectively. However, both two algorithms mentioned above do not make full use of the positive feedback information, which causes slow convergence speed if the initial value of step-size is not be set reasonably. This thesis propose a new perturbation algorithm with dynamic step-size by utilized cumulative positive feedback(Dynamic step-size with CPF), which contains the step-size growth mechanism with cumulative positive feedback and the mechanism of threshold adjustment. The proposed Dynamic step-size with CPF algorithm has ideal convergence speed for various situations. For the second scenario, this thesis proposes two global optimization algorithms of mobile relay position, i.e., the random perturbation algorithm and the directional perturbation algorithm. In the first algorithm, the mobile relay tries gradually to close the position optimized globally by adjusting the movement direction of mobile relay continually. While there are two stages in the second algorithm, the mobile relay runs along a straight line orbit in each stage. The position optimized globally can be achieved when the position optimized in each stage is achieved. In comparison with the first algorithm, although the second one has to predefine two thresholds to control the convergence process in each phase, it often has faster total convergence speed. Through theoretical analysis and simulation experiments, the proposed algorithms in two scenarios have been analyzed carefully and their superiority also be verified.