Radio based Cooperative Positioning for Vehicle-to-Vehicle Systems in Urban Scenarios

This master thesis aims at improving vehicle positioning in high-speed movement scenarios where global navigation satellite system (GNSS) does not work well. The standard 802.11p, as one of the WiFi family members, it can support to share vehicle information between vehicle and vehicle or between vehicle and infrastructure in a high-speed movement environment. Without the help of GNSS, the vehicles can estimate their position by sharing position information with other vehicles. In order to reach highly accurate positioning in urban scenarios, non-linear filters, such as extended Kalman filter (EKF), the square root of cubature Kalman filter (SCKF) and particle filters (PF) are investigated in this thesis. These filter algorithms are simulated in MATLAB to evaluate the positioning performance. Useful information from vehicles are used in the algorithms, such as velocity, acceleration of vehicles. To obtain realistic scenarios, vehicles are simulated in different road networks in SUMO and obtain the vehicle information from one another in GEMV2 and NS3. SUMO, GEMV2 and NS3 are the tools to help the simulation. In the simulations, the positioning accuracy is greatly improved when sharing vehicle information and utilizing the filter algorithm. This thesis compares the advantages and disadvantages of two filter algorithms. One is the square root of cubature Kalman filter, the other is particle filter. As a conclusion from the simulation result, the SCKF works better than the particle filter and it improves the accuracy of positioning when the GNSS is not well received.Global navigation satellite system makes vehicles to locate themselves with the help of satellites. Each vehicle needs at least three satellites sending time information from distance between it and the satellites. With geometry tool, the vehicles can estimate their positions on the earth approximately. However, the satellite signals travel a long distance to the vehicle so that the strength of the signal is weak to be detected. These signals are easily blocked by objects, which introduces an error in the positioning estimation. With the help of signal from base station, it improves the positioning. However, the signal from base station is also reflected and degenerated by the objects in the way of its propagation. Inspired by the idea that selecting a close signal access point to navigation, the nearest access points are picked up instead, what these points have the same function as satellite and base station is to measure the distance from themselves to vehicles in this thesis. These nearest access points are vehicles, since there are a lot of them running on the road. The thesis focuses on how the vehicles can obtain good accurate positions with the help of other vehicles. Vehicles supported by distance measurement among each other is the main idea in this thesis. The WiFi standard 802.11p supports vehicle-to-vehicle communication and vehicle-to-infrastructure communication, making vehicles share their information with others easily. Each vehicle collects all information from the surround vehicles to improve its own positions. 802.11p, as a member of IEEE 802.11 standard, supports sharing vehicle information in a high-speed dynamic environment. The vehicles share their own information in two ways. The first way is to send it directly to the destination vehicles. The second way is to send it to the infrastructure roadside unit or other wireless access points before sending to the target vehicles. The 802.11p standard provides broadcast protocol so that the distance measurement is estimated through the arrival time of the signal. The shared information includes position estimated by GNSS, velocity, acceleration and distance measurement of each vehicle. The thesis also compares the simulation results from the Kalman filter and the particle filter as different fusing tools. Both of these two filters are good at fusing shared information to obtain good positioning. The vehicular network simulation software simulates a real road network as not all the tests can be applied in real road networks with the hardware limitation. The software such as SUMO, GEMV2 and NS3 make vehicles run on the simulation road network and show all the vehicle information for further analysis. These software provide us the exact vehicle locations, velocities and corresponding distances among vehicles. The vehicle information is implemented into MATLAB to evaluate the filter algorithm performance on positioning. In order to make the simulation more general and see how accurate the positioning, several scenarios are created to examine the robustness of the algorithm. Several scenarios like distance, distance-velocity, GPS-distance, GPS-distance-velocity are discussed in each scenario