V2X Enabled Non-Signalized Intersections Management: A Function Approximation Approach

The significant enhancement of vehicular communications together with artificial intelligence (AI) have opened up new horizons for innovative data-driven traffic management solution within intelligent transportation system (ITS). In this paper, to alleviate progressively worse urban traffic, we propose an efficient vehicle-to-everything (V2X) communications enabled non-signalized intersection management framework for automated vehicles. First, a resource reservation model involved with different functional planes has been developed for vehicle collision avoidance in V2X enabled non-signalized intersections. Second, reinforcement learning (RL) solution is leveraged to enhance management efficiency of non-signalized scheduling. Furthermore, considering the dimensionality disaster problem of vehicle state caused by complicated traffic environment, we propose a function approximation based non-signalized intersection control (FA-NIC) algorithm to obtain optimal scheduling strategy. Simulation results are provided to demonstrate the effectiveness of our proposed non-signalized intersection management solution.