RU2134453C1 - Method for neuronetwork coordinated control of traffic flows in area of control - Google Patents

Abstract

FIELD: coordinated control of traffic flows in areas of control with employment neuron networks. SUBSTANCE: method is based on measurement of intensity of traffic flows at the entrances of crossings of the area of control, selection of coordination plan closest to the control totality as a result of measurements, and supply of control signals to the traffic light at the "entrances exits" of the area of control at the control traffic light objects are preliminarily assigned, then the layout of traffic light objects inside the area of control is selected, all traffic light objects are furnished with a microcomputer with a neuroprogram and means of problem fetching, all traffic light objects connected through parameters of flows are compared with those assigned at the area boundaries, and if essential deviations are available, feedback is effected by supplying signals from the control traffic light objects to the internal traffic light objects; as a result, self-adaptation of the traffic flow control system is accomplished. EFFECT: enhanced efficiency of automatic traffic flow control with the aid of trained and self-trained neuron network conditioned by an increase within the preset time interval of traffic capacity of transport facilities in the control area, decrease of traffic delays, prevention of jammings, etc. 3 dwg

Description

 The invention relates to coordinated traffic control, methods of controlling traffic flows in regulatory areas using neural networks.

 There is a known method of controlling traffic flows at an intersection, in which the intensity of traffic flows in each direction of the intersection is measured by counting the number of vehicles in each direction of the intersection for a period of polling and sequential switching of traffic signals, in which the average intensity of traffic flows at the intersection is determined for each of possible for this intersection of regulatory phases by counting for the period of the survey, the value of which is 10-30 times more than the value of the traffic light control cycle, the number of vehicles in all directions of movement of each phase and its division by the number of directions of movement involved in this phase. For each direction of movement of the intersection, the absolute value of the deviation of the intensity of traffic flows on it from the value of the average intensity of traffic flows of the regulation phase in which this direction is involved is determined, the absolute deviations of the traffic intensities for the phases included in each of the possible control programs are summarized, the obtained amounts are compared and choose the control program with the smallest sum of deviations of traffic intensities for the next survey period stv in the directions included in this control program, from the values of the average intensities of the flows of the corresponding control phases (a. from the USSR N 1399800, MKI G 08 G 1/01).

Benchmarking analysis
A.S. N 139980. A method for controlling traffic flows at an intersection is proposed. A decentralized control method with an adaptive choice of a regulatory program is proposed. As a reference, choose the average parameters (intensities) and make attempts to improve it. The method does not take into account preference for district-wide routes. The method is based on less efficient electronic means with traditional programs.

 The present invention. A method of coordinated neural network traffic management in the regulatory area is proposed. The district management method is based on a practical selection of tasks. The proposed method allows you to control traffic, taking into account the flexible change of time modes, reservations, emergencies, etc. The method is developed taking into account preferences for the regulatory area in case the intersection is related to the main directions of traffic in the regulatory area. The method is carried out on the basis of modern electronic means using a microcomputer with a neuroprogram that is customizable and adjustable, the system trained and self-taught during operation can combine the functions of measuring, receiving signals, making a trained decision and transmitting signals.

 A known method of controlling the movement of vehicles at an intersection based on recording the time of arrival of vehicles in the direction of the intersection, measuring during the time interval from the minimum to the maximum time of the included phase of traffic lights for the included direction of movement of time intervals between every two consecutive moments of arrival of vehicles in this direction, comparing the values of the measured time intervals with a given inte the time shaft and the inclusion of the phase of the traffic light control, if the next measured time interval exceeds a predetermined interval at which for each of the turned off phases of the traffic light at equal time intervals, ten or more times less than the minimum phase duration, measure the total waiting time of the enable signal by vehicles, received on the direction of the phase of traffic light regulation, compare them with each other, include the phase of traffic light regulation with the highest at the time of comparison The total waiting time for vehicles (see A.S. N 1462395, MKM G 08 G 1/01).

Benchmarking analysis
A.S. 1462395. A method of controlling the movement of vehicles at an intersection. Designed to reduce the total delay time of vehicles at the intersection, i.e. considered a single traffic light, which makes decisions without communication with other traffic lights. The control method is regulated by strict control logic, although it involves some search to reduce the transit time of the transport. The current situation (experience) is not remembered anywhere and is not studied for the future. The method is based on less efficient electronic means with traditional programs.

 The proposed invention. A method of coordinated neural network traffic management in the regulatory area is proposed. Designed to manage a controlled area, all traffic lights are interconnected and serve to control the regulatory area. The district management method is based on a practical selection of tasks. The proposed method allows you to control traffic, taking into account the flexible change of modes in time, reservations, emergencies, etc. The method is developed taking into account preferences for the routes of the regulatory area in case the intersection is related to the main directions of movement of the regulatory area. The method is carried out on the basis of modern electronic means using a microcomputer with a neuroprogram that is customizable and adjustable, the system trained and self-taught during operation can combine the functions of measuring, receiving signals, making a trained decision and transmitting signals.

 A known method of coordinated control of traffic flows, taken as a prototype, based on measuring the intensity of traffic flow at the entrances of intersections of streets of the control area, selecting as a result of the measurement of the coordination plan and signaling to traffic lights in accordance with the selected plan, which measures at the outputs of intersections of the control area , together with the intensity of traffic flows at the inputs of the intersection, they transform into traffic intensity in each direction At the intersection of the control area, in accordance with the highest traffic intensity in one of the directions, the next intersection with the maximum traffic intensity is sequentially selected and interviewed, the information obtained determines the set of directions with the maximum traffic intensity, compares it with the control sets, selects the closest traffic direction control population and select a coordination plan calculated in advance for the selected control population ( m. A.S. USSR N 510957, G 08 G 1/095).

Benchmarking analysis
A. s. N 510957. A method for controlling traffic flows is proposed. A strict coordination scheme is foreseen, a change in time is not stipulated and there are no cases of failure of individual traffic lights, emergencies, emergency route flows, etc. Automatic control is not provided. The proposed coordinated vehicle management is based on a passive approach, as measure and transform what is, compare with the control sets and select the closest one. The method is based on measuring the intensity of the traffic flow, choosing a plan and signaling in accordance with this, is carried out on the basis of imperfect electronic means. The system is not automatic.

 The proposed method. A method of coordinated neural network traffic management in the regulatory area is proposed. The proposed method is based on a flexible scheme, including a change in time. The method is based on training, self-training of the entire network of traffic lights by an object. The neuroprogram control of traffic light objects provides various control options, including cases of failure, the choice of the necessary direction of movement, memorization and accounting of emergency situations, etc. provides automatic traffic management. The method involves an active approach: establish this, the possible time of passage of vehicles at the "exits-entrances" of the area with the allocation of preferred routes of vehicles, etc. and in accordance with this task, the desired time of passage of vehicles through a network of traffic lights, which would ensure it compliance with all local conditions and restrictions. The proposed method has a specific implementation based on traffic signals objects with neuroprograms of management and devices for recognition, adoption of management decisions, the method works according to a flexible system, trained and self-adjusting, all traffic lights are interconnected. The method is carried out automatically.

 The proposed method of coordinated neural network traffic flow control in the regulatory area is designed to control traffic in the entire area, trains the system on the basis of a practical selection of tasks that take into account flexible change of control modes over time, redundancy, emergency situations, etc., configure the system and adjust it, etc. e. provide feedback and system self-tuning. The presented system is trained and self-taught during operation, uses modern electronic devices that combine the functions of measuring, receiving signals, making a trained decision, transmitting information and actuating actuators.

 The aim of the invention according to a method of coordinated neural network traffic control in the regulatory area is the effective automatic control of traffic using a learning and self-learning neural network, associated with an increase in the vehicle traffic capacity in a controlled period of time, reducing traffic delays, preventing traffic jams, highlighting preferred driving directions.

 The goal is achieved by the fact that according to the method of coordinated neural network traffic flow in the regulatory area, based on measuring the intensity of traffic flows at the intersections of the streets of the regulatory area, selecting the coordination plan closest to the control set as a result of measurements, and supplying control signals to the traffic light, pre-assign the desired indicators for preference coefficients from the "entrance" to the "exit" of the controlled area at the control traffic lights objects, after which they select the layout of internal traffic lights within the regulatory area, which allows for a strategy taking into account the conditions and limitations, provide all the traffic lights of the microcomputer with the neuroprogramme means for the mutual exchange of information about the "inputs and outputs" of vehicles, train on the basis of a practical selection of tasks traffic lights connected by an information network equipped with a microcomputer with a neuroprogram that takes into account the flexible change of modes over time, redundancy, possible failures and random situations, various preferences for special vehicles, compare the received flow parameters with those assigned at the borders of the district and, if there are deviations, correct them, i.e. provide feedback by sending signals about the change of preference coefficients from control traffic signals objects equipped with a microcomputer with neuroprograms to internal traffic signals objects, as a result, they carry out self-adjustment of the traffic flow control system.

 The proposed method of coordinated neural network traffic flow control in the regulatory area due to the appointment of the desired traffic flow control indicators, the choice of the location of traffic lights equipped with a microcomputer with a trained neuro program and information exchange tools that takes into account all situations in the regulatory area, comparing the received flow parameters with the assigned indicators and adjusting they allow you to effectively carry out the process of automatic regulation, and the benefit arya learning opportunities, and self douchivaniya traffic light devices equipped with a microcomputer control neyroprogrammnym, with the desired performance managed area allows a more efficient way to automatically implement traffic management in the regulatory area, it uses modern neyroprogrammy. The proposed method, together with the neuro-control program, can be local and coordinated in the directions of movement with the allocation of priorities and special assignments.

 The following terms and concepts are accepted below.

 Regulatory area is a certain area of the city that has certain trunk directions.

 Under the internal traffic light objects refers to a set of devices that provide traffic control at the intersection. The device includes directly a traffic light, motion sensors, a digital neurocontroller with an analog-to-digital converter and a microcomputer with a neuroprogram. The digital neurocontroller receives input from neighboring traffic lights and provides information to neighboring traffic lights. A microcomputer with a neuroprogram is trained on the basis of field examinations and self-learns during operation.

 Control traffic light objects are the same in composition as internal traffic light objects in composition, with the difference that their neuroprogram additionally compares the elapsed time with the established control criteria for the time of passage of vehicles and provides controlled signals to internal traffic lights with an indication of preference coefficients. A preference coefficient is calculated (see formula 4 below).

 A neutroprogram means a program trained on the basis of field surveys taking into account a given set of preference factors and which self-learns during operation. Based on the neural program, as a result of the training of each traffic light object (internal and control), tuning is performed for optimal operation, like training with the preference coefficients indicated to it. During operation, each traffic light object receives additional information and performs its control work by completing what is required by the neuroprogram algorithm (see Fig. 3), i.e. traffic light objects work with direct and feedback, interact with each other.

In FIG. 1 a scheme of coordinated neural network traffic management in the regulatory area,
in FIG. 2 - control device of a traffic light object,
in FIG. 3 is a block diagram of a neural network training and self-study program.

 The scheme of the coordinated neural network traffic control within the boundaries of regulation region 1 (Fig. 1), which has a real network of streets 2, 3, 4, 5, 6, 7 with intersections, at the entrances and exits of the regulation region are located 8, 9, 10, 11 - control traffic light objects with sensors 12 traffic intensity on the streets to measure traffic flows at the "inputs-outputs" of the regulatory area and internal traffic light objects 13-16 with their sensors 12 traffic intensity.

 The traffic light object diagram (Fig. 2) contains motion sensors 12 associated with a digital neurocontroller 17, consisting of an analog-to-digital converter 18 and a microcomputer with a neuroprogram 19, outputs connected to traffic light objects 13-15. The input of the digital neurocontroller 17 receives information from neighboring traffic lights, and the output goes to neighboring traffic lights. The outputs of the traffic light objects 13, 14, 15 are connected to the inputs of the information sensors 12.

 All traffic lights equipped with a microcomputer with neuroprograms are connected by an information network.

 The method of coordinated neural network traffic management in the regulatory area is as follows.

 Pre-measure and calculate the intensity of traffic flows at the "inputs and outputs" of the regulation area at the control traffic lights 8, 9, 10, 11 using sensors 12 traffic intensity installed on each lane, set the necessary traffic parameters taking into account the coefficients of preference for directions in area of regulation, choose as a result of measurements a coordination plan that is closest to the control population.

 The desired strategic indicators of traffic flows at the "entrances and exits" of the regulatory area are assigned, including the time that vehicles pass through the boundaries of the district (from maximum to minimum), especially highlighting the directions and flows of public transport, special transport with an indication of their preference coefficient, taking into account all local conditions and restrictions.

 After that, the layout of the placement of internal traffic lights 13-16 equipped with a microcomputer with a neuro-program inside the regulation area is selected, which allows providing a strategy taking into account the conditions and limitations of internal traffic lights 13-16 equipped with a microcomputer with a neural network program, which serves to process information and control traffic lights objects within this region, depending on this information necessary for the effective operation of the method, and thereby choose a coordination plan as a result of measurements, the most close to the control population.

 They supply all traffic light objects of a microcomputer with a neural network program of additional information from neighboring internal traffic light objects 13-16 through established information communication channels and means for the mutual exchange of information about the "inputs and outputs" of vehicles. The task of each internal traffic light object 13-16 with a microcomputer with a neural network program is to minimize transport delays at the boundaries of the district by increasing its throughput, taking into account the priority indicated for it, for passing through the main directions of traffic and for special vehicles.

 They train all traffic lights connected with each other by an information network with a microcomputer with a neuro program based on data taken from control traffic lights of objects 8-11, and send control signals to the traffic lights. The situation on the road at any given moment is reflected in the readings of the traffic intensity sensors 12 and includes the features of a particular moment: time of day, weather conditions, time of year, flexible change of modes over time, redundancy, possible failures, emergency situations, various preferences for special vehicles etc. A Mirkocomputer with a neuro-control program learns at the intersections of the regulatory area during this regulatory cycle to predict the state at the intersection of the regulatory area in the next regulatory cycle and respond accordingly by changing the operation order of internal traffic lights 13-16.

 The method at the traffic light object, for example, 13 (see Fig. 2) is as follows: measure the intensity of traffic flows at the inputs of the intersections of the regulatory area, provide information from the sensors 12 of the traffic intensity from adjacent traffic lights to a digital neurocontroller 17, containing analog-digital the transducer 18 and the microcomputer with the neuroprogram 19, are converted into digital form in the analog-to-digital transducer 18 and fed to the input of the microcomputer 19 with the neuroprogram, in which a set of coefficients of dpochteniya. With the help of a neuroprogram, they make a control decision and submit the results to traffic lights at the intersection of the regulatory area.

 From the outputs of the microcomputer with the neuroprogram 19, the information about the state at the given intersection of the regulatory area is supplied to the inputs of neighboring traffic light objects of the devices. The final information is fed to control traffic lights 8, 9, 10, 11, where the obtained flow parameters are compared with those assigned at the boundaries of the region, calculated by formulas 3, 4, and the supply of control signals to internal traffic lights 13-16.

 The scheme of training and retraining of each traffic light object connected by an information network with neighboring ones is shown in FIG. 3 and includes the input of data from the intensity sensors of neighboring internal and control traffic lights, the implementation of direct functioning based on preliminary training and retraining, calculating the learning step, reverse functioning, modification of tuning parameters. The received tuning parameters are used for the next cycle of the program.

Then proceed to set up, for this
calculate the theoretically possible and assign the desired (control) time of passage of traffic flows in the main (priority) directions and through this managed area
ΔT _i = (T _output - T _input ) i, (1)
where ΔT _i is the control time of the vehicle through the controlled area of regulation,
T _exit - the time the vehicle exited the area,
T _entry - the time the vehicle entered the area,
i is the number of the main direction of movement.

где

суммарное время задержки на n внутренних светофорных объектах,
с - суммарное время прохождения транспортных средств между внутренними светофорными объектами и подхода к контрольным светофорным объектам.For each internal traffic light object 13-16, they conduct training according to a flexible scheme at their intersection with a step-by-step set of preference factors η in the main directions. At the control traffic lights, the program calculates the elapsed time of the passage of traffic flows by the formula (2):

Where

total delay time at n internal traffic light objects,
c - the total time of passage of vehicles between internal traffic lights and access to the control traffic lights.

т. е. если контрольное время больше или равно затраченному времени, то процесс происходит без изменения. Если условие (3) не выполняется, то производят корректировку процесса за счет увеличения коэффициентов предпочтения η на внутренних управляемых светофорных объектах 13-16 по формуле (4):

где k - номер внутреннего светофорного объекта.The control time is compared with the elapsed time, i.e. (1) and (2), if

that is, if the control time is greater than or equal to the time spent, then the process occurs without change. If condition (3) is not satisfied, then the process is adjusted by increasing the preference coefficients η at the internal controlled traffic lights 13-16 according to the formula (4):

where k is the number of the internal traffic light object.

 Thus, in the neuroprogram, feedback and adjustment of the method of controlling traffic flows and self-tuning are automatically carried out. With new changes in the preference coefficient from the control traffic lights 8-11, the internal traffic lights 13-16 are rebuilt, give new information to all traffic lights, and the whole network of the controlled area is self-tuning. When training a neuroprogram on the basis of objective real data at different periods of the day, the conditions for changing the operation of individual traffic lights of the entire area are included, tuning, retraining occurs automatically and constantly.

 Thus, the use of the method of coordinated neural network traffic flow control in the regulatory area allows you to effectively automatically control traffic flow by choosing the layout of traffic lights equipped with a microcomputer using a learning and self-learning neuro program that takes into account all situations in a regulated area, training based on a practical selection of tasks all traffic lights connected by the information network, comparing the obtained parameters, adjusting and feedback by sending signals to control objects, self-tuning the control system.

 The method provides an active approach in which this possible time of passage of vehicles is established and, in accordance with this, the desired time of passage of vehicles through a network of traffic lights is set, which ensures its implementation taking into account all conditions.

 The proposed method has a specific implementation based on traffic signals with control neuroprograms, works on a flexible system, trained and self-tuning.

 Feasibility study.

 Traffic light regulation in Krasnoyarsk is carried out according to a strict scheme without taking into account the traffic intensity of vehicles and pedestrian flows and has the following disadvantages: increased transit time of vehicles and pedestrians, congestion and associated significant fuel consumption, increased gas pollution, etc.

The use of ASUD "Gorod" on the basis of third-generation technical equipment according to the results of implementation in a number of cities (Tashkent, Alma-Ata, Novosibirsk) gives a change in traffic indicators: reduced vehicle delay by 10%, increased communication speed by 10-12%, decreased the number of road traffic accidents, improved operational monitoring of traffic, etc. This system is the most modern, but it has difficulty in creating it, consisting in sorting out options with subsequent general and local correction th program, there is no possibility douchivaniya, training and self-control devices, bulky information and control devices (see. Kremenets YA Technical means of traffic management. M .: Transport, 1990)
The proposed method for coordinated neural network traffic management in a controlled regulatory area is the next step in improving the district's traffic management system, is able to eliminate the above disadvantages due to the possibility of training, further training and self-training of neural program control devices with phased implementation from full-scale information collection with a set of several temporary control programs to automatically information management system, simplifying info mation and control devices due to their combination. The proposed method, together with the neuro-control program, can be local and coordinated (along highways) with the allocation of priorities and special assignments.

Claims (1)

 The method of coordinated neural network traffic flow control in the regulatory area, based on measuring the intensity of traffic flows at the inputs of intersections of the regulatory area, selecting the coordination plan closest to the control population as a result of measurements, and supplying control signals to the traffic light, characterized in that the desired indicators are assigned according to preference coefficients at the "inputs and outputs" of the regulatory area at the control traffic lights on the basis of After measuring and calculating the intensities of traffic flows at the "inputs and outputs" of the regulatory area, after which they select the layout of internal traffic lights within the regulatory area, which allows for a control strategy taking into account the conditions and limitations, supply all the traffic lights of the microcomputer with a neuroprogram and means for the mutual exchange of information about the intensity of traffic flows at the "inputs and outputs" of the regulatory area, all traffic lights connected by the information network are taught, data from a microcomputer with a neuroprogram based on a practical selection of tasks that takes into account the flexible change of modes over time, redundancy, possible failures, emergencies, various preferences for special vehicles, etc., compare the obtained flow parameters with the assigned desired indicators according to preference coefficients at the "inputs-outputs "the regulatory area and in the presence of significant deviations provide feedback by sending signals according to preference coefficients from control traffic lights at the" inputs and outputs "p Jon regulation, a microcomputer equipped with neyroprogrammoy on internal traffic lights inside the control area.

Images