Abstract:Surveying projects of near-earth asteroids continue to emerge, and obtain massive observation data. However, this pattern makes the obtained arc too short, and the traditional methods have great difficulty in orbit determination and identification with ill-posed problem in itself when the arc is short. Then how to effectively use these short arc is of great significance for discovering, monitoring and evaluating the threat of asteroids. Under the evolutionary algorithms, a calculation framework for too-short-arc is constructed with three-variable (a,e,M) optimization, which keeps the dimensionality low while makes the optimization results no longer rely on observational measurements. The differential evolution algorithm with fewer parameters and simple operation is used to conduct experiments using orbital simulation data of asteroids with different eccentricity, then the optimal solutions and their aggregation regions are analyzed. The large eccentricity orbits will have an impact on the sensitivity of the algorithm search due to its complexity, it is need to reduce the search space to improve the search ability. The results show that the algorithm performs well in small eccentricity problem, and can obtain valid results to provide information for subsequent work. And for large eccentricity problem, while the traditional method fails, the distribution of the algorithm still contains the real solution. For the phenomenon that the optimal solution is not obvious in the global distribution, it can be analyzed by combining the distribution density and fitness value. Further research on the issue of large eccentricity is needed in the future, the influence of different observation positions and observation time on the algorithm should be considered, and calculate by classification.