Abstract: In order to find out the key nodes of land use regulations, and provide scientific support for the formulation of the relevant policies and the measures, multi-agent system (MAS) offers a bottom-up approach to explore the internal mechanism of land use change through simulating the decision-making behavior of micro-level agents. With a case study of Xiaozhuang Village in Tanghe County, Henan Province, this paper took farmers' choice between cotton and peanut as an example, and attempted to explore the application possibility of MAS approach in the study of agricultural land use change by simulating land use decision-making behavior of farmer households. This study constructed a conceptual framework of land-use decisions for farmer households base-on MAS, in which household's decision-making behavior was affected by both internal factors (ability and willingness) and external factors (market, policy, natural conditions and interaction with other households). And the conceptual framework analyzed the cycle mechanism among these factors: Internal factors were the direct and core factors in households' decision-making process, external factors affected their decision-making behavior indirectly through influencing internal factors, i.e. ability or willingness, and the cumulative result of farmers' decisions could change land-use pattern of the region as a whole, which would further influence internal and external factors. Then a mathematical model was given based on the conceptual framework to simulate the decision-making process of the households: First, the internal-factor-based decision-making rule was formulated to get the internal decision-making value, which was regarded as the base value. Then external factors were taken into account to amend the base value and to get the final decision-making value. The results demonstrated that: 1)Internal decision-making value on peanut was generally higher than on cotton, although the value varied from one household to another. Market factors strengthened households' willingness to grow peanuts. The effects of interaction on households varied based on each household's characteristic. In general, external factors made households more likely to grow peanut rather than cotton, increased households' final decision-making value on peanut but decreased that on cotton. 2) Simulated value of area ratio (cotton/peanut) is 0.26:0.74, which was in good agreement with the actual value 0.26:0.74 (The simulated value was only 5.4% higher than the actual value). Thus, the simulation caught the key factors in households' decision-making process such as profit, labor, market, etc. It also well explained and predicted decision-making behavior of households, the land-use pattern and change of the region. Furthermore, the reason behind the results were discussed: Considering the labor-dependent characteristic, the suitability of natural conditions, and price stability, cotton was in a relatively inferior position than peanut in research field，so the households were very sensitive to the price drop of cotton in the year 2011, which brought about huge negative impact on local cotton cultivation, and peanuts became main autumn commercial crop in Xiaozhuang Village in the year 2012. 3) The relationship between the benefit of households and the market was analyzed: Although closely relevant with each other, they were essentially different. Market was a regional-scale indicator that represented the average level of the area as a whole. Benefit was a household-scale indicator whose value was influenced by some irregular or uncontrollable factors, which reflected the complexity of the households' decision-making behavior. The construction of the model was to simulate the complicated real world, meanwhile to abstract and simplify it, so how to balance the relationship between "simulation" and "simplification" was very important in the model design and expression. 4) In this case, labor was chosen as the ability factor, because contradiction between the labor-dependent characteristic of the crops (especially cotton) and the labor loss (caused by high opportunity cost) made labor the primary factor influencing, even determining households' ability. In practical application, the ability factors can vary under different situations. This model can provide theoretical foundation for the regulation of farmer households' land-use behavior, and help promote efficient agricultural land use fundamentally.