A Novel GNSS based V2V Cooperative Localization using Consistency Checks to Exclude Multipath Effect
Guohao Zhang, Weisong Wen and Li-Ta Hsu, The Hong Kong Polytechnic University, Hong Kong
The technology of autonomous driving is rapidly developing in recent years, aiming to contribute on the traffic safety including collision avoidance. One of the bottlenecks to realize the autonomous driving is the current localization accuracy cannot guarantee the operation safety in urban canyon. For vehicle localization, various sensors are employed to improve the positioning accuracy, such as the global navigation satellite system (GNSS), LiDAR, vision sensor and inertial navigation system (INS). With the development of vehicular communication technology, such as vehicle-to-vehicle (V2V), the sharing of the information of between vehicles is enabled during operation. The shared vehicular information can cooperatively aiding the localization for autonomous driving. Comparing to the transponder-based V2V localization approach (using RSSI or TDOA), the exchange of vehicular GNSS data can prevent the scenario that the direct communication between vehicles are obstructed . In the other words, the GNSS based V2V cooperative localization is outperforming the transponder-based approach in the current situation that both the vehicles with/without connectivity are driving on road.
GNSS localization is influenced by serval factors, including satellite orbit/clock bias, atmospheric delays, receiver clock bias and the multipath effect. By applying differential GPS (DGPS) technique, the systematic errors can be eliminated using the correction from a reference station . Since the development of V2V, it is possible to cooperate the GNSS raw measurements to improve positioning. By extending the idea of DGPS, Kai Liu et al.  developed a GNSS based V2V cooperative localization algorithm by implementing double difference (DD). Both the systematic errors and receiver clock bias error can be eliminated using DD approach, achieving better localization solutions for vehicles. This GNSS DD approach is able to obtain relative position between two vehicles in open-sky areas. However, the GNSS localization performance is greatly challenged due to multipath effect. In urban, the transmission of satellite signal is usually reflected by building surface, resulting in an extra traveling distance. This multipath effect further introduces positioning error for GNSS satellite positioning, which is very severe in dense urban, such as Hong Kong, Shanghai and Beijing with great density of high buildings. The multipath affected localization error even exceeds 50 meters during operation in urban and the double difference process may even increase the error. To mitigate the multipath effect, a robust process is required to firstly identify the multipath signal and exclude those signals before applying GNSS positioning estimation. Since we are in an era of multi-GNSS including, GPS, GLONASS, Galileo and BeiDou, the number of visible satellite is always high even in urban canyon. Hence, excluding the multipath affected signal will not cause the lack of satellite for positioning. By employing the idea of GNSS measurement consistency check , the signal with multipath effect can be recognized and excluded based on its pseudorange residual. After excluding the biased GNSS measurement, the double difference technique and cooperative localization algorithm can be applied to obtain accurate relative and absolute vehicles’ position, even in urban areas. The improved GNSS solution can be therefore bounded in a certain error level that effectively aiding autonomous driving system on navigation operation.
In this study, the objective is to develop a novel GNSS cooperative localization algorithm in urban area, mitigating the multipath effect and improving the localization performance. To achieve the objective, two layers of GNSS measurement greedy search based consistency-check are developed. The first one is the consistent check at single vehicle (single receiver) level. The inconsistent measurement will be identified as enormous measurement affected by multipath effects. The surviving GNSS measurements will be exchanged with the surrounding vehicles to conduct the DD approach. The single difference of the GNSS measurement between two vehicles from the same satellite eliminates the atmospheric delay and satellite clock/orbit error. Then, the second difference of the measurements from two satellites to the same receiver is able to eliminate the receiver clock bias. As a result, the DD GNSS measurements are solely affected by multipath effects that not successfully excluded by the first layer consistency-check. The second layer consistency-check is then implemented into the least square estimation for the relative position between two vehicles. The benefit of this double layer design is to increase the number of measurements to exclude the inconsistent measurements. By cooperating the 1) GNSS fixed absolute position that passed first layer check and 2) relative position between two vehicles that passed second layer check, the localization estimation of ego vehicle can be optimized with these measurement constrains.
The proposed algorithm is firstly tested by the simulated multipath-biased GNSS measurement based on 3D building model and ray-tracing algorithm. The positioning accuracy is evaluated by the comparison between the accuracies of conventional GNSS weighted least square estimation and double difference cooperative GNSS localization without excluding enormous error satellites. The improvement of proposed algorithm is further verified using real GNSS raw measurement received from multiple pedestrian as different nearby locations. The proposed multipath excluded cooperative GNSS localization algorithm is able to achieve higher accuracy on both absolute and relative position in urban operation. According to the experiment result, the proposed algorithm achieves about 10 meters accuracy in urban canyon.
1. de Ponte Müller, F., Survey on ranging sensors and cooperative techniques for relative positioning of vehicles. Sensors, 2017. 17(2): p. 271.
3. Parkinson, B.W. and P.K. Enge, Differential gps. Global Positioning System: Theory and applications., 1996. 2: p. 3-50.
4. Liu, K., et al., Improving positioning accuracy using GPS pseudorange measurements for cooperative vehicular localization. IEEE Transactions on Vehicular Technology, 2014. 63(6): p. 2544-2556.
5. Hsu, L.T., et al., Multiple Faulty GNSS Measurement Exclusion Based on Consistency Check in Urban Canyons. IEEE Sensors Journal, 2017. 17(6): p. 1909-1917.