This paper introduces a semi-distributed cooperative localization technique realized via multi-node time-of-arrival (TOA) and direction-of-arrival (DOA) optimal fusion: Each base-node estimates the position of target-nodes by joint TOA- DOA evaluation, and then, the target-node position estimation error is minimized by TOA-DOA optimal fusion across multiple base-nodes. The performance of the proposed technique is studied and compared to two GPS-based positioning techniques, i.e., GPS-aided TOA fusion and GPS-aided DOA fusion. The circular error probability (CEP) is derived theoretically and verified via simulations. The results confirm the superiority of the proposed localization technique in moderate scale mobile ad- hoc networks (MANETs) compared to the two GPS-based fusion schemes. Thus, while the proposed technique is applicable to MANETs in GPS-denied environments, it is also suitable for GPS available environments. Finally, compared to the centralized scheme, the positioning updating rate of the semi-distributed technique is higher and its power consumption in the reference base node is considerably lower.