In the research, we present a robust cooperative positioning method in GNSS-challenging environments, in which observations from vision-aided technique are integrated with information provided by cooperative UAVs. In particular, the graph factor based on Student-t distribution is used to joint recursive estimation of the states and the measurement noise parameters of the nonlinear system. In order to decrease the computational complexity, the proposed cooperative positioning method uses a separable variational approximation to marginalize out the joint posterior distribution of states and noise parameter. The performance of the proposed method has been evaluated on the basis of numerical simulations. Results of a experimental test show that the proposed method provides a comparable performance of cooperative positioning to the traditional methods in GNSS-challenging environments.