A method for estimating the positioning accuracy of vehicles in urban environments

GNSS-based positioning system could provide accurate and real-time positioning information. It is necessary to meet the requirements of GNSS positioning performance in terms of accuracy and integrity. For automatic driving, if a large positioning error is ignored, it might result in a bad decision making and endanger the safety of both drivers and passengers; the drivers or the intelligent systems inside the smart vehicles need to make decisions on when to brake so as to avoid the collisions. So it is necessary to estimate the positioning accuracy of vehicles in real-time.Many researches have been done to estimate the positioning accuracy of GNSS in urban environments. The accuracy indicator of GNSS receivers can roughly estimate the positioning accuracy. However, the performance of the indicator can be severely degraded by the limited satellite visibility, multipath effects, and interference. The protection level (PL) mentioned in the advanced Receiver Autonomus Integrity Monitoring (ARAIM) is used to estimate the boundary of the positioning errors, but it cannot be applied directly to vehicle applications since the complex environments will make the performance of the protection level worse. This paper presents a method for positioning accuracy estimation in urban environments. We use the decision tree model in machine learning, to train a model for estimating the positioning accuracy. The main advantage of the proposed method lies in the use of multiple features that indicate the positioning accuracy.To illustrate the effectiveness of the proposed method, we consider a scenario where a cluster of vehicles drive in the urban environments. The preliminary experimental results show that the method proposed can accurately estimate the positioning accuracy and inform users in time, the probability of which is more than 95%.