Electronics and information technologies / Електроніка та інформаційні технології

Adaptive algorithms, which current traffic systems are based on, exist for many decades. Information technologies have developed significantly over this period and it makes more relevant their application in the field of transport. This paper analyses modern trends in the development of adaptive traffic flow control methods. Reviewed the most perspective directions in the field of intelligent transport systems, such as high-speed wireless communication between vehicles and road infrastructure based on such technologies as DSRC and WAVE, traffic jams prediction having such features as traffic flow information, congestion, velocity of vehicles using machine learning, fuzzy logic rules and genetic algorithms, application of driver assistance systems to increase vehicle’s autonomy. Advantages of such technologies in safety, efficiency and usability of transport are shown. Described multi-agent approach, which uses V2I-communication between vehicles and intersection controller to improve efficiency of control due to more complete traffic flow information and possibility to give orders to separate vehicles. Presented number of algorithms which use such approach to create new generation of adaptive transport systems.

The change in the intensity of traffic flows observed during the day requires a corresponding change in traffic management parameters, such as cycle times and time of the enabling signals. Adaptive control, due to the presence of feedback from the traffic flow, allows you to take into account both daily changes in intensity and its fluctuations due to the random arrival of vehicles. Systems based on adaptive management have been in place for the last decades, and their application in both Metropolitan areas and smaller cities has proven to be effective. However, the modern development of information technologies and intelligent transport systems (its) allows to create qualitatively new methods of traffic management aimed at improving the convenience, efficiency and safety of transport.

Today, there are dozens of different implementations of adaptive transport management systems, and the most common are SCOOT and SCATS. Modern achievements in the field of its can significantly expand the current capabilities of adaptive transport management and create more advanced systems through the use of advanced sensors, electronics, computer and communication technologies, innovative management strategies.

One of the important directions of its development is the use of wireless telecommunications. Research conducted in this area in the 2000s showed that the existing Wi-Fi technology does not meet the objectives. To solve these problems, a new addition to the Wi-Fi standard - IEEE 802.11 p was created. The new Protocol is based on the technology of DSRC (Dedicated short range communication), which serves for short-range communication. The next generation technology is called WAVE (Wireless Access to Vehicular Environment) and provides high-speed data transmission. The shortest-range wireless networks are used for data exchange between devices inside the car, for example, for communication between the driver's smartphone and the car's systems. V2V communication includes the exchange of data with vehicles passing near or moving on the same route, as well as emergency broadcasting to vehicles located nearby. V2i connectivity uses the roadside infrastructure for data exchange and network connectivity with vehicles. Also, the car can have a direct Internet connection via a cellular network. The services developed at this point include a cooperative alert system, collision system for the detection of collisions, a cooperative security system intersections, warning of the approach of emergency vehicles, or areas with road work.

Another direction in its is the prediction of congestion. In the last few decades, the most common road forecasting techniques have been based on the Kalman filter and the integrated moving average autoregression (ARIMA) model. Currently, much attention is paid to methods that can perform forecasting based on several features, including traffic flow, degree of road occupancy, speed. Such algorithms include support vector machines (SVM), neural network (NN) system based on the rules of fuzzy logic (FRBS), genetic algorithms (GA) . The most effective methods to date can predict the occurrence of congestion for a period of 5 to 30 minutes with an accuracy of 95%.

The last decade is characterized by active development in the field of Autonomous and unmanned vehicles. This is demonstrated through the following projects: VIAC (2007-10),  HAVEit (2008-11), Cybercars-2 and CityMobil (2005-08 and 2008-11) [13], the GCDC competition (2009-11), e-Safety (2002-13), the DARPA competition and Google's Driverless Car. Work on advanced systems assist the driver (PSV) is conducted in such areas as an assistance system when changing lanes, security systems pedestrian warning system and collision mitigation, adaptive light control headlights, an assistance system when Parking, night vision system cruise control system of internal monitoring, allowing to detect the sleepy state of the driver and to warn about dangerous situations. The next step is the creation of cooperative adaptive cruise control systems based on V2V interaction, traffic sign and traffic light recognition systems, systems that use information from digital maps, for example, to select the appropriate speed before a steep turn. The development of the PRSP will affect the safety requirements of vehicles and over time the use of such systems will become mandatory, making the machines more Autonomous.

Multi-agent systems (MAS) are systems consisting of Autonomous intelligent agents interacting with each other and a passive environment in which agents exist and can be affected. The use of MACS to control the intersection is made possible by the development of V2i and V2V communications. The transition device is equipped with a controller implementing an algorithm for controlling the phases of the traffic light. In the case of Autonomous vehicles, the traffic light can simply perform a secondary function, since the main commands can transmit through V2i connectivity. The controller has a specific range, appearing in which the car transmits information about its position, speed and direction of movement. The controller collects this information from all vehicles and performs phase planning based on the data received. If necessary, commands are sent to the vehicles so that they can adjust their actions.

The proposed algorithms of the controller can differ significantly from system to system. Some are focused on better planning of traffic light phases by providing more information on traffic flows, others are engaged in monitoring the trajectories of vehicles for more efficient and safe passage of the intersection, others suggest to abandon the regulation of traffic light and actively use the possibilities of wireless communication through the transmission of messages. The quality of the proposed algorithms is evaluated by simulation, which shows a significant increase in the efficiency of the intersection in comparison with the classical traffic light control. In addition to the tasks of managing an isolated intersection, the MAC can be used in centralized traffic management systems throughout the city, which are engaged in the formation of routes, planning of traffic time and coordination of individual traffic light objects in order to avoid congestion .

Most of the existing methods of adaptive transport management are based on technologies available for several decades. The article presents a number of technologies, such as devices that provide wireless high-speed telecommunications with cars, and advanced driver assistance systems, the emergence and spread of which will lead to the creation of a new generation of adaptive transport management systems. Such systems will be able to collect detailed descriptions of traffic flows, including information on the route, speed and position of individual vehicles, and be able to transmit individual commands to these vehicles, allowing urban transport systems to cope with the constant growth in the number of vehicles and traffic volumes observed around the world.

Key words: adaptive control methods, intelligent transport systems, machine learning, fuzzy logic rules, genetic algorithms, road infrastructure, wireless vehicle interaction, multi-agent approach, traffic lights, intersection