Multi-robot task allocation (MRTA) and path planning are crucial for optimizing the performance of multi-robot systems. Effective task assignment can enhance system efficiency and minimize conflicts during path planning. To address MRTA and path planning in complex environments, this paper introduces a method that combines an improved genetic algorithm (IGA) with conflict-based search (CBS). First, the MRTA problem is modeled as a multiple traveling salesman problem (MTSP). K-means clustering is then applied to partition the task points into clusters, transforming the MTSP into multiple TSPs to reduce the problem’s dimensionality. Next, the IGA is used to solve each TSP, obtaining a high-quality traversal order without considering robot path conflicts. Finally, the CBS is used to solve the multi-agent path finding (MAPF) problem, obtaining conflict-free paths for multiple robots that satisfy the traversal sequence constraints. The simulation results demonstrate that the IGA-CBS effectively handles the MRTA and path planning problems in complex environments.