Abstract: The fusion of global navigation satellite system, strapdown inertial navigation system, odometer and zero velocity updates can provide accurate position and attitude references for ground mobile surveying systems. For the existences of mathematics model bias and measurement outliers, the traditional federated filter cannot isolate the interference from fault subsystems efficiently when fusing multisources information. Based on the equivalence of adaptive filter and federated filter and the principle of robust filter, an algorithm is proposed to calculate the information sharing coefficient by using the pre-innovation and then adjust it automatically. A field vehicle test shows that compared with the traditional federated filter the proposed adaptive federated filter can resist the measurement outliers and degrade the impact of the mathematics model errors effectively, and the accuracy and robustness of the integrated navigation system can be improved.