The design of suitable estimator architectures for INS aided by relative measurements and information sharing among an unrestricted community of users is discussed. Both decentralized and centralized estimators are presented. The centralized estimator is based on the Extended Kalman Filter. The challenge of handling inter-vehicle correlation in the decentralized framework is highlighted and a Covariance Intersection (CI) based decentralized estimator is proposed. The centralized and decentralized estimators are evaluated in an automotive simulation with a community of 2,000 + low-cost INS users operating in a GNSS-denied zone. Both single-vehicle and community-wide performance is compared. The proposed CI-based decentralized estimator is demonstrated to be conservative and maintain consistency. The centralized system achieves an order of magnitude reduction in uncertainty by aggressively aiding the INS on board each vehicle. Lastly, a quantitative evaluation of bandwidth requirements for both centralized and our proposed CI-based decentralized cooperative navigation systems are presented. Copyright © 2014 Institute of Navigation.