Research On The Problems Of The Description And The Simulation Of Mobile Agents

Developments in the research of network technology and computer science motivate the revolution of the distributed computing model. To find an efficient way to share the large amount of network resources and make up for the drawbacks of the conventional distributed computing model such as C/S model, a new model known as mobile computation is proposed and developed. Naturally, the representation of the mobile computation, the mobile agents, became the focus of the research and application in the distributed computing recently. Because of its distinct features such as mobility and reactivity which fit the requirements of distributed computing in the WAN well, mobile agent is adopted by people to develop the algorithms to solve various kinds of distributed problems.However, it remains a big problem to find an appropriate way of evaluating the performance of the mobile agent based algorithms. The difficulty mainly lies in two aspects: the one is the lack of a common standard of the execution model of the agents, which makes it hard to make a consistent and sound evaluation for the agent algorithms; the other is that the conventional ways in evaluating the performance of algorithms, mathematical analysis, live code deployment, and conventional simulation, are no longer fit for the ones based on mobile agents, mainly due to its features such as mobility and reactivity.To evaluate the performance of the mobile agent algorithms, this dissertation focused on looking for the solutions of some important problems in the description and simulation of the mobile agents:1. The theoretical research on mobile computation has made a lot of achievements, however there is a gap between the application and theoretical research, making the mobile agent based algorithms hard to be connected with the mobile computation theory. Meanwhile, for the scarcity of the common standard for describing the mobile agent's execution model, it is prone to be ambiguous in describing, understanding and implementing the mobile agent algorithms. In fact, this problem is a fundamental one to evaluate the performance of mobile agents. In the dissertation, a model called SMA is proposed to describe mobile agent's structure and execution flow. And with the model, a description language also called SMA language is proposed to formally describe the essence of the SMA model. To make a formal semantic explanation of the SMA language and connect it with the other formal approaches available in theoretical research of mobile computation, the encoding approaches for translating the SMA agent program to the equivalent description in Mobile UNITY andπ-Calculus are also provided.2. The research on the simulation of mobile agents is stagnant because of the lack of a concrete standard of the agent's execution model. Some works only present a simple qualitative analysis and is not enough to be realized and applied to the practical work at all. Based on the SMA model and the theory of modeling and simulation, the thesis builds a parallel discrete event system model for the mobile agents as the theoretical basis of establishing the simulation model for the mobile agents. Further, the thesis also explores the semantics of discrete event system model, which is a topic often ignored by the researchers.3. As mentioned above, due to the features of mobile agents, the conventional simulation approaches are difficult to be applied in the mobile agents. An efficient simulation approach called direct execution simulation, which has worked efficiently in the simulating the conventional distributed problems, is adopted by the thesis to build a simulation model called MADES (Mobile Agent Direct Execution Simulation). It lay the basis for solve the problem of simulating of mobile agent's simulation and performance evaluation eventually.4. SMA model and MADES provide a feasible way for describing and simulating the execution of mobile agents. The final solution of the problem is to build a simulation environment that meets the conditions. In the thesis, the popular mobile agent system, IBM Aglets is analyzed and a new simulation environment called SimulAgent is introduced. SimulAgent is a mobile agent direct execution simulation environment built on the SimulAgent Aglets, a mobile agent platform fitting for the SMA model. It has been used in concrete mobile agent's simulation and performance evaluation.The major contributions of this dissertation include: 1) A standard model called SMA for describing the execution of mobile agents is proposed to avoid the ambiguity of mobile agent's description and implementation.2) A formal description language, SMA language is proposed and the encoding approaches for translating the SMA programs to Mobile UNITY andπ-Calculus descriptions is presented.3) Based on the theory of modeling and simulation, a parallel discrete event system model called SMA-DEVS is proposed for simulating the mobile agents. The semantics of the discrete event system is also explored.4) The direct execution simulation approach is adopted in the thesis to build a simulation model for mobile agents. The architecture and implementation of the model, called MADES, are introduced.5) A direct execution simulation environment, SimulAgent, based on the modification of IBM Aglets, is built to implementing the SMA model and the MADES model and is applied to the simulation and performance evaluation of mobile agents.