Design Of V2I-Based Autonomous Intersection Control Policy And Simulation Research

Autonomous vehicles with the vastly improved precision control and sensing are coming, which brings new hope for safe driving. The existing artificial intelligence techniques have been able to effectively ensure autonomous car driving safely on the open road. But for driving at the intersection with complex situations, heavy traffic and more accidents, both security and efficiency cannot be guaranteed. This paper introduces an autonomous intersection management solution, trying to leverage global positioning, wireless communications, in-vehicle sensing and computation technologies to enable autonomous vehicles to drive through the urban intersections safely and efficiently.Control policy, just like the brain of intelligent intersection management system, has a direct impact on the system performance. This paper proposes a reservation-based control policy called Batch-Light which can make full use of existing Intelligent Transportation System (ITS) technologies to make the management system adaptive to constantly changing traffic. A Time-Space Board model is proposed to model the vehicle’s route in discretization method. And based on this model, a greedy-based Conflict Matrix decision algorithm is proposed to get more vehicles reserve successfully on the premise of ensuring certain fairness. Furthermore, a k-Shift optimization algorithm is proposed to help some unlucky vehicles to pass through the intersection as much as possible by acceleration or deceleration, which maximizes the efficiency and reduces the time waste and space waste.This paper extends the AIM (Autonomous Intersection Management) simulator with feedback loop, by making use of the dual-directional coupled simulator interconnection technology, implementing a more realistic wireless communication simulation with NS3instead of the simple message transmission model in AIM. This work is used for supporting the research on the demand level of different control policies for communication, and in turn how communication performance affects the traffic control efficiency.We evaluate our Batch-Light with other control policies together on the simulator we build. Our experiment results show that Batch-Light outperforms FCFS (First Come First Served) and traditional traffic signal control policy both in unbalanced traffic and balanced traffic. And Batch-Light requires smaller amount of data transmission and as the noise increases, its efficiency goes down steadily. While FCFS requires larger amount of data transmission, and the effects of network performance on the control efficiency are not fixed. Sometimes with the increase of packet loss rate and the transmission delay, control efficiency will rise.