Motivated by potential applications such as the use of multiple robots to collaboratively localize multiple objects/events of interest through audio cues, we propose a novel solution for blind source separation (BSS) and localization using multiple microphone arrays. Unlike the existing state-of-the-art where direction of arrivals (DOAs) and locations are estimated sequentially, the proposed approach jointly estimates both by solving a constrained optimization problem, in which the dependence of the separated signals is minimized. By adopting the alternating direction method of multipliers (ADMM), the problem can be solved in a distributed way in which only the local DOA estimates are communicated among arrays in each iteration. Because the augmented Lagrangian used in ADMM naturally forces the estimates to approach true values by adding penalties on non-optimal ones, the critical scaling and permutation problems in BSS are solved simultaneously. Simulation results further demonstrate that the estimation accuracy can be significantly improved using the proposed method.