A Scalable and Computationally Efficient Connected Vehicle-Based Signal Control Algorithm

This paper proposes a scalable traffic signal control algorithm that relies on information obtained from Connected Vehicles. The proposed framework utilizes a modified genetic algorithm to identify (near) optimal signal phase and timing plans that minimizes total vehicle delay at an isolated intersection. The phasing plans are completely flexible in terms of phase sequence and duration and thus the algorithm can respond well to changes in prevailing traffic conditions. The proposed algorithm is tested in a simulation environment and compared to an enumeration approach that has been previously used to solve a similar problem. The results show that significant efficiencies can be achieved in computational effort (over 95% shorter duration to obtain final signal timings) without sacrificing car delays (less than 1% increase in car delays) compared to the enumeration approach. Furthermore, the computation time required to run the algorithm can be set a priori by simply modifying the parameters of the genetic algorithm, which allows the proposed algorithm to be scalable with respect to the total input flow and the penetration ratio of connected vehicles while still being capable of running in realtime.