Routing Algorithms Based On Multi-Agent System In LEO Satellite Network

The next generation Internet will be a comprehensive communication network with integrated components of air-space-ground. Satllite network is very important among them. Low Earth Orbit satellite network has become the research hotspot due to its advantages of small transmission loss and low end-to-end delay characteristic. In order to realize efficient information transmission in LEO satellite network, routing problem must be faced above all. Design and implement efficient and reliable routing algorithm is an important research content for LEO satellite network with ISL.The traditional LEO satellite network routing algorithms trend to divide the dynamic topology of the network into a series of static topology and transplant the ground mature routing algorithms on it. The method takes into consideration the inheritance of existing technologies for satellite network, but its unadaptable characteristics are also exposed in the study. Because of this, scholars from home and abroad turn to study the adaptive routing. Due to the satellite handover influence and the diversity of optimization targets, most of the research achievements are no systematized and could not implement transverse comparison. On the basis of this, the paper takes detailed analysis on routing problem in LEO satellite network and study the problem through different views using Multi-Agent System. Some positive methods and meaning conclusions are given in this paper. Overall, the following aspects are studied:Firstly, the traditional routing algorithms often adopt greed mechanism, which is ambivalent to realize the maximum performance for the whole network and the routing algorithms based on intelligence need unified structure.Aimed at the defects, a new routing algorithm is proposed based on Multi-Agent System in this paper. Except that, the main characteristics and design process are introduced which gives referential significance for the last of the paper.Secondly, single path routing algorithm for LEO satellite network is unadaptable when satellite handover happened. It has to rerouting frequently due to the failure and restoration of ISL. Multipath routing is better than single path in terms of providing fault-tolerant pathes and polymerization bandwidth which is suitable for LEO satellite systems with global coverage and changing topology. A multipath routing algorithm based on Multi-Agent System is introduced in the paper. The routing algorithm uses forward agent to explore network and backward agent to update routing information.What's more, blackboard is introduced to realize direct communication between agents which accelerates the convergence speed. The innovation of route updating makes the algorithm track the topology changes in LEO satellite networkThirdly, load balance routing algorithms for LEO satellite network are incomplete and a novel load balance routing algorithm with traffic prediction is proposed in the paper. For LEO satellite systems with global coverage or changing traffic in sub-satellite point, the imbalanced characteristic of traffic distribution is reflected not only in time, but also in space. Radial Basis Function Neural Network is used to predict traffic of the grids on the earth and the results are used by predicted agent to revise the routing strategy on the current satellite. The proposed algorithm can dynamically adjust data between satellite nodes and effectively solve the problem mentioned above.Finally, LEO satellite network has to provide QoS routing to support multi-services and services with mixed demand. But traditional accurate routing algorithms are hard to achieve optimal solution timely. The complexity of QoS routing problem with multiple constraints is NP-Complete and the time to solve the problem will increase fast and become unbearable due to the scale of the problems. A novel QoS routing algorithm based on Multi-Agent System is proposed in this paper to solve the problem. In the algorithm, the QoS routing problem is divided into a series of subtasks which are completed by different intelligent agents. The algorithm not only acquires QoS satisfied paths, but also has low complexity and good expansibility which is suitable for different QoS indexs.