This dissertation considers the geolocation of a point of interest (POI), i.e., determining the location of a POI in the world, using multiple cooperating uninhabited aerial vehicles (UAVs) with gimballing camera sensors. A square root sigma point information filter (SR-SPIF) is developed to provide a probabilistic estimate of the POI location. The SR-SPIF utilizes the UAV's onboard navigation system to save computation and also takes important properties for numerical accuracy (square root), tracking accuracy (sigma points), and fusion ability (information). The SR-SPIF is general and scales well to any tracking problem with multiple, moving sensors. In the development of the SR-SPIF, the errors in the navigation system output are assumed to be zero mean. However, in the practical application, there are non zero mean errors (biases), which degrade geolocation accuracy. Therefore, a decentralized approach to simultaneously estimate the biases on each UAV and the unknown POI location is developed. The new decentralized bias estimation approach provides accurate geolocation in spite of sensor biases and further scales well with the number of UAVs. Communication is an important part of a cooperative geolocation mission and in practice communication losses and delays are inevitable. Therefore, a new method for cooperative geolocation in the presence of communication loss, termed the predicted information method, is developed from a separable formulation of the extended information filter. The predicted information method is shown to give the exact solution for linear systems when the measurement dynamics are constant or known by all UAVs. In addition to theoretical developments, extensive experimental flight tests with ScanEagle UAVs have been performed. The experimental flight tests serve two purposes: 1) to develop practical guidelines for geolocation 2) to validate all of the new approaches presented in this dissertation. In addition to the flight tests, a high fidelity, distributed, hardware in the loop simulation test bed was developed and used as further validation of all new approaches.