Application of Connected, Cooperative and Automated Vehicles in traffic light controller
Many of today’s vehicles and road users are connected through devices like mobile phones and GPS devices. In the near future, they will also interact directly with one another and with the infrastructure and management systems. The exchange of information between road users(V2V) and management systems(V2X) through such interaction, will give ways to the new possibilities to coordinate various traffic management systems and improve traffic light control performance.
The performance of traffic light controllers as part of intelligent traffic systems(ITS) is improving by incorporating more information about current and future traffic state. For example, the information that is exchanged between traffic light controller and (connected)vehicles will increase the predictivity of traffic state and as a result, improve the performance of adaptive and model predictive controllers (MPC).
Integration between different ITS applications and overall transportation system will also raise new opportunities for traffic light controllers to perform better. One can think of priority that is given to platoons of vehicles – i.e. formed for example by a traffic light in upstream section –to increase the traffic throughput.
Not only the traffic light controllers benefit from new sources of data, but also, communication between vehicles and traffic lights is crucial to increase the safety and performance of automated vehicles. Such communication helps the automated vehicle to anticipate traffic condition and cooperate with other traffic management systems (e.g. re-routing).
Connected Vehicles’ Data Fusion
New technologies motivate developing the new generations of sensors in the near future. These sensors will not only detect and report car trajectories but also provide other types of information such as vehicle class and intended route. Generally, the data that could be collected from different types of sensors is categorised as follow (1):
- Passive data: doesn’t require the driver to participate in data collection procedure, such as inductive-loops, radar, camera images and license-plate readers.
- Semi-passive data: is using drivers’ devices but doesn’t need consent and knowledge of the driver. Examples include roadside Bluetooth, Wi-Fi, tire-pressure sensor sniffers, radio-frequency identification, cellular hand-off signals.
- Active data: requires driver’s consent and motivation to provide the data. Examples include fleet-tracking networks and mobile applications that allow users to report scenarios such as accidents and construction work.
All these sort of data need to be processed in the future traffic light controller to provide better prediction and estimation of current traffic state. Particularly, active data sources that can track the vehicle along the road and provide information such as vehicle classes are important for adaptive controllers where a self-learning algorithm fine-tunes control’s variables and parameters such as Green split and extension time.
Travel time Reliability for Automated Vehicles
Automated vehicles are rapidly developing and in the near future will be introduced commercially. New traffic management systems including traffic light controllers should be prepared for the increased number of automated vehicles in their future policies and paradigms.
The introduction of automated vehicles in traffic will face a transition period where the conventional and different levels of automated vehicles will coexist and have to be managed so that the safety and efficiency of traffic flow is not negatively affected. This transition period requires that traffic light controllers effectively communicate and cooperate with automated vehicles while maintaining their traditional control policy and sources of data for nonconnected vehicles.
Automated vehicles need reliable information from traffic lights and other infrastructure data sources in addition to their in-car sensors in order to perform driving tasks. The information that is needed by automated vehicles at an intersection includes the queue length, time to green and green duration which should be obtained from traffic light controller.
In various cases the complexity of traffic situation necessitate intervention of the human driver in automated vehicle driving task. One reliable categorization of level of vehicle automation is provided by society of automotive engineers(SAE). This categorization(2) highlights the need to transfer different level of driving tasks- depending on level of automation- between vehicle system and humans.
Such transition often increases the mental load of human driver and requires to be planned ahead of time for safety reasons. Research shows that when the time of engagement of driver is predictable, drivers’ attention is more stable and concise(3). Therefore, it is very important that traffic light controller transmits reliable information(with high confidence factor) to automated vehicle.
Table 1 - Level of automation of vehicles according to SAE
Integrated Traffic Management System
Increasing number of road users, their heterogeneity and co-existence of conventional and automated vehicles introduces several problems in our future transportation systems. These problems may have negative effect on economic development, traffic accidents, emission, lost time and decreased level of user comfort. Traffic management systems aim at minimizing the traffic congestion and its negative effects and they are composed of two sorts of tools and applications(4):
- Infrastructure-free systems: cooperative congestion detection, congestion avoidance, accident detection and warnings, red light violation warning.
- Infrastructure-based systems: traffic light management, route suggestion, speed harmonization, platooning.
Large-scale traffic management requires an effective combination of ITS applications and data synchronization between these applications. In this paper the focus is on collaboration of traffic light controller and other ITS applications. In Table. 2 examples of information exchange between traffic light controller and infrastructure-free management systems could be seen. Table. 3 provides examples of infrastructure-based ITS applications and the information exchange with traffic light controller.
Table 2- Infrastructure-free ITS and traffic light controller data exchange
Table 3 - Infrastructure-based ITS and traffic light controller data exchange
An Example: Amsterdam Control with “Smart Traffic”
“Smart Traffic” is a cloud based, decentralised and model predictive traffic light controller and it is developed and launched by Sweco. It is based on real-time data fusion and it has a microscopic data-driven model as the predictive model.
“Smart Traffic” uses a broad set of data sources such as Extended Floating Car Data (xFCD) or WiFi-sniffers in addition to the induction-loop data and camera images. The information provided by sensors, vehicles and infrastructure(e.g. other TLCs) is used in “Smart Traffic” to maximise the traffic flow throughput by a multi-objective optimization method. Such methodology provides policy makers and road authorities with sufficient tools to change and practice new policies such as emission reduction and modal (e.g. bicycles) priority.
The possibility of using and incorporating new sources of data in “Smart Traffic” ensures a resilient product that can keep up with the rapid developing sensors technology and growth in data sources. Furthermore, the ease of configuration and flexibility of objective functions (travel time, stops, delay, etc.) in Smart Traffic provides road authorities with an efficient tool to manage an oversaturated urban area.
In addition, accurate prediction of traffic state in Smart Traffic (i.e. by microscopic simulation model) could be useful when it comes to intelligent traffic management, predicting travel time, and transit signal priority (TSP). For instance, priority feature will change/adjust signal timing in order to turn the signal green for specific class of vehicles(e.g. busses), resulting in less delay for these vehicle classes.
e characteristics of ”Smart Traffic” as a model predictive controller allows for handling the over-saturated intersections in urban area.
In a case study, “Smart Traffic” is applied on 4 intersections in the city of Amsterdam (The Netherlands). The simulation tool is PTV Vision’s micro simulation software Vissim and the simulated network(Figure. 1) contains vehicles, pedestrians, cyclists, busses and trams.
Besides induction- loops, “Smart Traffic” is using floating car data in this simulation study. The information that is provided by vehicles from Vissim is used in the optimization paradigm of “Smart Traffic” to provide priority to scheduled busses, trams and emergency vehicles.
Furthermore, floating car data are used to fine-tune the traffic state and prediction provided by the internal forecasting model of “Smart Traffic”.Figure. 2 provides an overview of all floating car data from different vehicle classes that were identified and incorporated in “Smart Traffic” in 1 hour simulation.
Results and discussion
The objective functions of “Smart Traffic” are configured to minimize emission while respecting priority policy. “Smart Traffic” achieves these goals by reacting to the demand in real-time and based on the stop and delay costs of each of the signal groups.
Figure. 3 demonstrates the speed profile of the simulated network together with red and green phases of one signal group. The optimisation method indicated cycle time, green split and offset based on the arriving pattern.
Figure 1 - Amsterdam network in Vissim simulator
Figure 2 - Vehicle classes in 1 hour simulation, class 0 is passenger cars, class 1 is buss, class 2 is tram, class 3 is bicycles and class 4 is pedestrians.
Figure 3 - Speed profile and signal phases provided by “Smart Traffic"