Data-Driven Kalman Consensus Filtering for Connected Vehicle Speed Estimation in a Multi-Sensor Network

Abstract

The autonomous traffic system has imposed higher requirements on the speed estimation of connected vehicles, where the speed of connected vehicles, as one of the control conditions for refined traffic management, plays a crucial role in the evaluation and optimization of network performance. In this paper, we propose a multi-source speed measurement sensor network consensus filtering (MSCF) algorithm based on information weight for the problem of optimal speed consistency estimation for connected vehicles. Specifically, we first utilize dynamic linearization techniques and data-driven parameter identification algorithms to handle the derived state equations of connected vehicles. We then establish observation models for four different types of sensors and construct distributed direct and indirect measurement models by dynamically adjusting the information weights of sensor nodes. Following this, we combine the Kalman consistency filtering algorithm to derive the speed state estimation update rate and design a new state estimator to achieve the optimal consistent convergence estimation for connected vehicles’ speed. The MSCF algorithm can solve the problem of consistency filtering for noisy sensor data under observation- and communication-constrained conditions, enabling each sensor node to obtain a consistent convergence estimation value for the speed of the connected vehicle. The convergence of the algorithm is proved using the Lyapunov function. Through numerical simulation, the results are verified, indicating that compared to existing methods, this method can achieve a higher precision speed estimation effect.