In complex urban environments, vehicle positioning based on Global Navigation
Satellite Systems (GNSS) is prone to failure or accuracy degradation due to
signal blockage and multipath effects. To address this issue, this article
proposes a vehicle–road cooperative positioning method based on factor graph
optimization for GNSS-denied environments and evaluates its performance through
both simulation and real-vehicle experiments. In the proposed approach, road
codes deployed on the road surface serve as absolute position references on the
road surface, and high-precision vehicle position estimates are obtained in real
time by fusing roadside positioning information with onboard sensor measurements
using factor graph optimization. Furthermore, to reduce the experimental cost
and development cycle of the vehicle–road cooperative positioning method, a
performance simulation platform is developed based on the CARLA simulator,
RoadRunner, and the CARLA-ROS (Robot Operating System) bridge for real-time
communication between simulation and positioning modules. The platform supports
road code generation and deployment, customized scenario construction, and
positioning performance simulation, and a multi-objective optimization approach
is employed to obtain an optimal deployment scheme for road code spacing.
Finally, the effectiveness of the proposed vehicle–road cooperative positioning
method is validated through both simulation and real-vehicle experiments.
Real-vehicle experiments show that the proposed method reduces the average RMSE
(Root Mean Square Error) by 26.6% compared with the ESKF (Error State Kalman
Filter) baseline, achieving an RMSE of 0.25 m and a maximum error of 0.95 m at a
vehicle speed of 60 km/h and a 10 m code spacing, thereby confirming
decimeter-level continuous accuracy in GNSS-denied environments.
