Abstract: The semantic information descript by various agents was hard to comprehend, and the precision of attribute classification for semantic concept was low in agriculture drought mitigation process. In order to solve the problem, the multi-agent collaborative reasoning fitness model was established, based on the agriculture drought mitigation disposal process. At the same time, the multi-agency collaborative rule reasoning multi-objective robust optimization model was established to solve the multi-agent synergy rigid and widespread uncertainty problem in reasoning process. NSGA-Ⅱalgorithm was used to solve calculation example, and the different risk levels of decision solutions were obtained. Simulation analysis showed that the method exhibited higher system stability and validity.