When conventional algorithms are used to solve the survey task planning problem of unmanned underwater vehicle (UUV) swarm under limited endurance and load constraints in a wide and sparsely distributed area, the poor convergence and low solution quality are common problems. In this article, an improved ant colony optimization algorithm is proposed. Firstly, by analyzing the constraints of individual UUV platform capability and swarm task, the constraint model and optimization model of UUV swarm task planning are formulated. Secondly, a method of unequal allocation of initial pheromone concentration is designed based on the difference between the average distance and the distance between task points, the optimal and worst thresholds of the optimization model are proposed to classify the ants and complete pheromone update, introduce an innovative “ gravity factor” that dynamically changes according to the iteration process to the state transition rule to increase the probability of the suboptimal node being selected in the early and middle of the algorithm. Thirdly, the validity of the algorithm optimization item is analyzed by statistical experiment and the algorithm parameters are optimized according to the number of optimal solutions and average convergence value. Finally, based on the cases from classic documents, by making comparison analysis with basic ant colony algorithm and elite ant colony algorithm, the proposed algorithm in this article improves the percentage of finding approximate optimal solution by 78% and 66% in 50 statistical experiments and average convergence in the 40th generation, which shows good global optimization capability and convergence performance. Experimental results of a designed typical UUV swarm mission planning case with a certain scale show the rapidity and effectiveness of this algorithm in solving the problem of the swarm survey task planning in a wide and sparsely distributed area.