Abstract: The existing average speed estimation models based on GPS (global positioning system) data of taxis are not applicable where high precision and low cost are of necessities. Two traditional models, namely a time integral model and a time interpolation model, based on the GPS data of taxis were modified, and the two modified models were combined to form a hybrid model. The three new models were verified using GPS data of taxis obtained from a part of road network of a large city. The results shown that the estimating precisions of the two modified models increased by 1.5% and 19.7%, respectively, compared with the original time integral model and time interpolation model, and those of the hybrid model by 2.8% and 8.8%, respectively, compared with the two modified models