Abstract: Abstract: The research on driving mechanism of the small-scale rural land use change has vast importance to make overall arrangements about the scale, the layout, and the time sequence of rural land. Taking Guanlin Town, Yixing City, Jiangsu Province as a case, and using remote sensing images, questionnaire survey and socioeconomic data, this study built a model based on multi-agent system and cellular automata to quantitatively analyze the influence of the natural and locational factors and agents' decision-making behaviors on area and space change of rural land and their driving mechanism. The results showed that: 1) Rural land use change was influenced by natural environment factors, social environment factors and distance variables. The concrete factors had varied influences on different types of rural land use changes. Arable land change was mainly influenced by soil pH value, per capita annual net income of farmers and distance to waters. The change probabilities would increase by 1.161 and 1.313 times respectively with soil pH value and per capita annual net income of farmers increasing by one unit, while the change probabilities would decrease by 0.780 time with one unit increasing in distance to waters. Aquiculture water surface change was mainly influenced by slope, per capita annual net income of farmers and distance to waters. The change probabilities would increase by 1.065 and 1.067 times respectively with one unit increasing in slope and distance to waters, while the change probabilities would decrease by 0.761 time with per capita annual net income of farmers increasing by one unit. Rural residential land change was mainly influenced by surface soil texture, counterpart funds of public welfare and distance to roads. The change probabilities would increase by 1.995, 1.380 and 1.179 times respectively with one unit increasing in surface soil texture, counterpart funds of public welfare and distance to roads. Enterprise land change was mainly influenced by slope, gross industrial output value and distance to roads. The change probabilities would increase by 1.064 and 1.130 times respectively with one unit increasing in slope and distance to roads, while the change probabilities would decrease by 0.300 time with one unit increasing in gross industrial output value. Ecological land change was mainly influenced by output value of agriculture, fishery and industry. The change probabilities would increase by 1.093, 1.017 and 1.371 times respectively with one unit increasing in output value of agriculture, fishery and industry. 2) The rural land use agents consisted of government, enterprises and farmers. Different types of rural land use changes were influenced by different agents' decision-making behaviors. Agents' decision-making behaviors of land use expansion were affected by their individual characteristics, economic characteristics and family characteristics, which could lead to area change of rural land. Relocation decision-making behaviors were affected by regional traffic factors such as distance to waters, roads, and markets, which resulted in spatial change of rural land. As for the planner and decision maker of rural land use, decision-making behaviors of government had substantial influence on change of each land use type. Area and space distribution changes of enterprise land were mainly influenced by decision-making behaviors of enterprises, with a weight of 0.41. Enterprise agents' decision-making behaviors of land use expansion were influenced by their individual characteristics and economic conditions such as total assets, age and academic composition of employees, idle land area. Relocation decision-making behaviors were influenced by zonal and traffic factors such as distance to markets and distance to roads. Area and space distribution changes of arable land, aquiculture water surface and rural residential land were mainly influenced by decision-making behaviors of farmers, with the weights of up to 0.30. Farmers' decision-making behaviors of land use expansion were influenced by their individual characteristics such as age and academic qualification, and domestic characteristics such as population and income. Relocation decision-making behaviors were influenced by arable land quality, water quality and distance to roads. Rural land use change was driven by the comprehensive effect of natural and locational factors and agents' decision-making behaviors. It is feasible to simulate rural land use change with the model combining cellular automata and multi-agent system. The Kappa coefficient of the model was 0.8087, which indicated that simulation results had high reliability. The research can provide scientific basis for the systematic study of rural land use change, and for the more accurate analysis of its driving mechanism.