To address the ant colony algorithm's problems of slow convergence speed and tendency to easily fall into the local optimum, we propose an ant colony algorithm based on dynamic evolution and interactive learning mechanism. Based on the concept of niche, we construct a dynamic evolutionary model within the two populations and adopt an adaptive evolutionary mechanism. According to the optimization status of the subpopulations within the population, the algorithm eliminates the subpopulations by stages, and the evolutionary direction is adjusted reasonably to prevent the subpopulations from falling into local optimum. At the same time, the two populations adopt an interactive learning mechanism to measure the population. The standard deviation of the fitness of each subgroup can adjust the interaction period adaptively, reduce the communication overhead among the populations, and adopt the strategy of learning object difference pairing to improve the communication efficiency and solving accuracy. Finally, we use a number of TSP examples with different scales for experimental analysis and compared with other multi-population algorithms. Results show that the algorithm performs better in improving the solution accuracy and optimization speed.