Towards a Multi Agents System Coupling Replication and Exception Handling

Multi agents systems are formed of different independent entities placed in several machines. When an entity or an agent fails, it is the whole system that may be in a failure case. Through this paper, we will propose an approach that may guarantee fault tolerance in multi agents systems using two different techniques which are replication and exception handling. Replication uses redundancy to handle the problem of failures, while, exception handling gives the possibility of recovery in case of fault signal. To use these two different mechanisms within the same system, we use the idea that a multi agents system has a global goal or a mission which it has to achieve; this global goal can be divided into sub goals that may be simple or complex. A simple goal is realized by one single agent that is considered in this case as an important agent and it is replicated. Exception handling is used in case of complex goals. Within this work, we define a complex goal as the one that must be realized by more than one agent. The major aim of our work is to create a strong multi agents system that may benefit of the two techniques proposed to guarantee fault tolerance.