In the Grid computing environment, there are a lot of important issues, including information service, information security, resource management, routing, fault tolerance, and so on. Among the issues, the job scheduling is a major problem since it is a fundamental and crucial step in achieving the high performance. The job scheduling problem has been known as a combinatorial optimization problem. Scheduling in a Grid environment can be seen as an extension to the scheduling problem on local parallel systems.

In the research, we focus on applying the technologies of RFID and Grid computing to the architecture of EPC network. In the view of EPC network, there are a lot of applications needing computing power of Grid to solve huge quantity of coming EPC data. According to the architecture of EPC network and the environment of Grid computing, they have several places homologous characteristics with each other.

Therefore, we propose a new algorithm that modifies the traditional GA and integrates SA. Since the processes of GA and SA keep no memory, some problem?s status visited before may be visited again. In order to overcome the drawback, we design a learning scheme to remember visited statues to reduce the probability of the re-visiting improvement the performance in the search space. It can help to find the optimal or Near-optimal scheduling efficiently and avoid the resource deadlock. Furthermore, our proposed algorithm, HGASA- with learning scheme, also considers about several properties of Grid computing environment and EPC network, such as heterogeneity, dynamic adaptation and the independent relationship of jobs.