Variable Speed Limit Control In The Work Zone Area Of Two-way Eight-lane Highway Under The Existence Of Connected Automated Vehicles

Two-way eight-lane highway is the main artery of the regional transportation,playing an important role in supporting and serving the local social and economic development.As China turns to high-quality development,the traditional incremental expansion has been transformed into stock renewal and the early-built highways have gradually entered the stage of maintenance.As the traffic demand continuously grows,the highway work zone area has become the main bottleneck that restricts the traffic operation safety and efficiency.The Internet of Vehicles and its intermediate technologies are being gradually applied.For a long time in the incoming future,the traffic flow on the road will be the heterogeneous traffic flow composed of traditional human-driven vehicles and connected automated vehicles.Variable speed limit(VSL)control is a frequently adopted method in highway work zone area.Investigating its application under the existence of connected automated vehicles will help to improve the traffic operation safety and efficiency in the highway work zone area and further meet the requirements to build China’s strength in transportation.The author clarified the organizational form of the work zone area of two-way eight-lane highway and chose the work zone with the inner two lanes closed as the typical example for research.Taking the work zone area of the Shanghai-Nanjing highway with the same organizational form for instance,the author extracted the vehicle trajectories from the videos recorded by an UAV.The traffic flow characteristics such as vehicle type distribution,flow rate,average speed,and the headway as well as the lane-changing characteristics of each lane in different segments of the work zone area were investigated.The possible factors causing the congestion of work zone area were analyzed from both the macro and micro perspective and mechanism of congestion forming in this area was also revealed,which provided essential clues for further work like modeling the heterogeneous traffic flow and the VSL control.The calculation of the highway work zone capacity and common control methods were introduced and the practical and theoretical basis of the application of the intelligent transportation system and speed control in this area were clarified.The classifications of the development level of both the intelligent connected infrastructures and connected automated vehicles were integrated and the classification of the development level of intelligent connected transportation system was proposed.By predicting the developing trend of relevant technologies,the Internet of Vehicles development level studied in this paper was clarified.The principle as well as the effect of VSL control were elaboreted and the applicability of VSL control in the work zone area under the Internet of Vehicles was demonstrated.The performance and features of each car-following model were summarized,and the principles for choosing the suitable car-following model in work zone simulation were proposed.The intelligent driving model and the PATH model based on real vehicle test were adopted in this paper to model the traditional human-driven vehicles and connected automated vehicles respectively.The fundamental diagram of the heterogeneous traffic flow varying with the penetration rate of connected and automated vehicles was analyzed while considering the dynamic performance and geometric parameters of different vehicle types and the changing pattern of the theoretical maximum throughput of the road section varying with penetration rate of connected and automated vehicles was revealed.A reinforcement learning based VSL control model was established.The corresponding action space,state space,the reward and reward function considering traffic safety and efficiency were defined.The deep deterministic policy gradient algorithm was selected to yield the speed limit and the feasiblility of the control method was ensured by mapping the continuous action space to discrete action space.The proposed VSL control was validated through simulation conducted on SUMO.A typical work zone of Shanghai-Nanjing highway with inner two lanes closed was chosen as the example for parameter calibrations.The reinforcement learning based VSL control was trained under different connected automated vehicle penetration rates.The taffic operation states of both with and without VSL control were evaluated from efficiency and safety perspectives and the results showed that the proposed VSL control was able to improve the traffic efficiency and safety in the work zone area.