Intersection traffic control with individual virtual wall and deceleration avoidance in a dynamic traffic flow environment

Abstract

With the advancement of machine learning and sensor technology, along with information sharing via vehicle-to-vehicle communication, overall optimization of traffic control focused on autonomous vehicles is becoming feasible. As an intersection traffic control method in a fully autonomous driving society, intersection control using Virtual Walls (VW), which differs from conventional optical traffic signals, has been proposed. VWs function as virtual obstacles and dynamically control vehicle routes to achieve safe intersection passage for autonomous vehicles. However, existing VW methods have two limitations: (1) complete avoidance of inter-vehicle collisions is difficult, and (2) they have only been evaluated in static environments with a fixed number of vehicles and fixed initial positions, limiting their practicality. To address these issues, this research proposes two functional extensions. First, we propose an individual VW placement optimization algorithm that corresponds to combinations of vehicle generation positions and destinations in a dynamic traffic flow environment where vehicles are continuously generated. Second, we introduce a Deceleration Avoidance function based on vehicle trajectory prediction to achieve collision avoidance through minimal speed adjustments in situations that are difficult to address through route design using VW control alone. Simulation evaluation in both static and dynamic traffic flow environments confirmed that the proposed method significantly reduces the number of collisions compared to conventional methods and achieves effective intersection control even in dynamic traffic flow environments.