| Mobile ad hoc network (MANET) is specific wireless network without fixed infrastructure. It differs fundamentally in both functionality and capability from wired network or wireless LAN. A significant amount of research has been dedicated to the development of MANET. However, most of the research efforts have been concentrated on fundamental networking issues such as routing and mobility management, rather than management. Now, the technologies for implementing and deploying MANET are becoming more mature and standards for ad hoc networking are emerging, it is time to pay attention to management issues. Due to the emerging need for rapid deployment of MANET in areas such as temporary conference and disaster recovery, fault management that guarantees robust network communication service becomes more important. However, most existing fault management methods are designed for wired networks, assuming static deterministic network elements dependency model and information. These assumptions can not be made in MANET due to its stochastic nature. Although several new fault management methods were suggested for MANET, they suffer from low efficiency, and large communication overhead. Consequently, it is imperative to design specific fault management for MANET.We applied artificial neural network technique to fault management in MANET. However, traditional Back-Propagation (BP) neural network suffers local minimum point trapping problem. To overcome it, we combines Generic Algorithm (GA) and BP algorithm to design a hybrid GA&BP training algorithm. The experiment results show that the hybrid solution is superior compared with the traditional BP algorithm. To effectively collect fault alarms while maintaining low overhead, a hierarchical fault information collection system based on SNMP is designed. Under Linux OS, we implement this fault management system and then test it in real MANET environment. The results show this system performs well and can achieve the design goal. |
PDF Full Text Request