RU2151424C1 - Device for controlling light signal object using computer with neural network software - Google Patents

Abstract

FIELD: automatic road traffic control equipment. SUBSTANCE: device has controller, which has computer, real-time clock, non- volatile memory unit, which stores crossing configuration, traffic detector, which has computer with neural network software, which is preliminary trained for recognition of type and amount of vehicles, direction of their movement, velocity, traffic lanes, by means of reading information from transport detectors. Controller has computer with neural network software, which is preliminary trained for practical task selection, which takes into account flexible change of regulation modes in time, redundancy, incidents, as well as special transport preferences and correction of information which is received from adjacent light signal objects. Device gathers information from traffic detector on the crossing and adjacent light signal objects and corrects preference rates from monitored light signal objects. Controller which has computer with neural network software is connected via information transmission network to adjacent and monitored light signal objects by means of receiver- transmitter and antenna. EFFECT: automatic flexible control using trained neural network, increased throughput capacity of road traffic through crossing. 6 dwg

Description

 A traffic light control device relates to automatic coordinated traffic control, traffic control at a traffic light object using neural networks.

 A known method of controlling the movement of vehicles at an intersection. A device that implements this method includes a switch 1, a control unit, a waiting time determination unit, a block (generating a switching signal (phases, a computing unit, OR elements 6.1-6.N, N - the number of phases of the traffic signal control, generator, triggers 8.1- 8., AND elements, counters, switches 16.1-16.N, AND elements 17.1-17 Nxn, counters 18.1-18. N, AND elements 19.1-19N, triggers 20.1-20.N, AND elements 21-21.N, switches 22.1-22. N, AND elements 23.1-23.Nxn, counters 24.1-24.N, switches 25.1-25.N, OR elements 28.1-28.N, OR element [1].

 The disadvantage of this device is a non-automatic rigid control circuit of a traffic light object at the intersection, the characteristics of this device are given in comparative analysis 1 (see table 1).

 The closest technical solution chosen for the prototype is the “Blik” traffic light remote control device, consisting of a controller containing a single-chip computer connected by two-channel communication with a transceiver and an antenna, and the controller is connected by a port card with traffic lights. The controller board contains a real-time clock, non-volatile memory with the intersection configuration recorded in it, a single-chip modem that provides encoding and decoding of radio signals [2].

 The disadvantage of this device is its non-automatic, inflexible control of a traffic light object, which is illustrated by a comparative analysis 2.

 This invention is intended to implement a method of coordinated neural network traffic management in the regulatory area according to the application N 97117398/09 (018114).

 The aim of the invention is the automatic flexible control of a traffic light object using a learning neural network, associated with an increase in the established period of time throughput of the traffic flow at the intersection.

The following requirements are imposed on a traffic light object for implementing this method:
- the traffic light object must be equipped with a computer with a trained neuroprogram,
- must have an information connection with neighboring and control traffic light objects,
- must have an information connection with a transport detector equipped with a computer with a neuroprogram previously trained to recognize the type and quantity of transport at the intersection, its direction of movement, speed, lane based on information from transport sensors.

 These requirements are used in the present invention, achieve the goal and serve to implement the method according to the application N 97117398.

 The goal is achieved in that in the control device for a traffic light object with neuroprograms, consisting of a controller containing a computer, a real-time clock, non-volatile memory with the intersection configuration recorded in it, the controller is connected by a bi-directional channel with a transceiver with an antenna and a bi-directional channel with a port card connected to lamps of a traffic light object, a transport detector is introduced into it, equipped with a computer with a neuroprogram, previously trained to recognize the type and quantity at the intersection transport, directions of its movement, speed, lane based on information from transport sensors, as well as the fact that the controller with a computer is equipped with a neuroprogram that has been preliminarily trained in a practical selection of tasks that takes into account the flexible change of control modes in time, redundancy, emergency situations, and also on preferences for special vehicles and the adjustment of incoming information from neighboring traffic lights, operating on the basis of information from a transport detector at this intersection and from neighboring traffic lights objects and correcting preference coefficients from control traffic lights at the "inputs / outputs" of the regulation area, while the controller with a computer with a neuroprogram is connected by a transceiver information network to neighboring and control traffic lights through a transceiver and an antenna.

 Due to the fact that a transport detector is introduced into the control device of a traffic light object with neuroprograms, it is equipped with a computer with a neuroprogram previously trained to recognize the type and quantity of transport at the intersection, its direction of movement, speed, lane based on information from transport sensors, and the controller The computer is equipped with a neuro program pre-trained in a practical selection of tasks that takes into account the flexible change of control modes in time, redundancy, emergency situations, and the same preferences for special vehicles and the adjustment of incoming information from neighboring traffic lights and functioning on the basis of information from a transport detector at this intersection and from neighboring traffic lights based on correcting preference coefficients from control traffic lights at the "inputs and outputs" of the regulation area, while the controller is equipped with a computer connected with a neuroprogram by a transceiver information network with neighboring and control traffic lights through a transceiver and an antenna, is achieved The purpose of the invention is an effective automatic flexible control of a traffic light object using a learning computer with a controller neuroprogram associated with an increase in the traffic flow capacity at the intersection in a specified period of time with a decrease in the delay time of the traffic flow and the prevention of congestion.

 The computer controller has a learning neuro program. For training, the dual functioning algorithm is used, which is a multidimensional optimization procedure for adjustable parameters with minimization of the operational error, which uses the Lagrange multiplier method [3, 4, 5]. A computer controller with a trained neuroprogram has already been used and tested in action to control the bearing capacity of a multi-span continuous beam and has shown good results, as well as described in the manual "Automatic control of structures using neural networks" [6, 7].

 The transceiver information network of a traffic light object with adjacent traffic light objects allows you to adjust the traffic flow at the intersection and the regulation area.

 The proposed control device for a traffic light object with a computer with neuroprograms has a specific implementation based on traffic light objects with neuroprograms trained in the preliminary selection of tasks, it works in a flexible way, and has the ability to retrain.

 The use of a control device for a traffic light object with a computer with neuroprograms makes it possible to effectively automatically control traffic flows due to a learning neuroprogram that takes into account all situations in a regulated area, training based on a practical selection of tasks of all traffic light objects connected by a transceiver information network, signaling to control traffic light objects, comparison received parameters, adjusting and implementing feedback, and implementing control system settings.

The control device of a traffic light object with a computer with neuroprograms is presented in the figures:
in FIG. 1 shows a General diagram of a control device for a traffic light object with a computer with neuroprograms,
in FIG. 2 is a general diagram of a traffic light control device,
in FIG. 3 - training scheme of the neural program of the computer of the controller, at the input the traffic intensity, at the output - the burning time of the green signal,
in FIG. 4 is a diagram of the functioning of the neural program of the controller with a computer,
in FIG. 5 is a training diagram of a detector neuroprogram for recognizing the type and amount of transport. The training scheme for the neural program of the computer of the controller is also similar (see Fig. 3), at the input - readings of transport sensors, at the output - traffic intensity,
in FIG. 6 is a diagram of the functioning of the computer of the detector neuroprogram.

 Below are used concepts and definitions as in the application N 97117398.

 The regulatory area is understood to mean a certain area of the city that has certain main directions. Under the "input-output" of the regulatory area refers to the measurement of the intensity of the traffic flow in the regulatory area.

 Under the internal traffic light objects refers to a set of devices that provide traffic control at the intersection. The device includes: a direct traffic light, a controller with a computer with a neuroprogram, a transceiver, a port card with lamps for a traffic light object, a transport detector with a computer with a neuroprogram.

 Control traffic light objects, the same composition as internal traffic light objects, with the difference that their neural program additionally compares the elapsed time with the control criteria for the time of passage of vehicles and provides controlled signals to internal traffic light objects with the indication of correction preference coefficients.

 The selection of tasks takes into account the flexible change of control modes in time, redundancy, emergency situations, as well as preferences for special vehicles and adjustment of incoming information from neighboring traffic lights.

 The control device for a traffic light object with a computer with neuroprograms at a traffic light object in regulation area 1 (Fig. 1) has a real network of streets 2, 3, 4, 5, 6, 7 with intersections. Street 6 is the main highway. At the "inputs and outputs" of the regulatory area are located 8, 9, 10, 11 - control traffic lights with transport sensors 12 installed on each lane of the regulatory area for measuring traffic flows at the "inputs and outputs" of the regulatory area and internal traffic lights 13, 14, 15, 16 with their transport sensors 12.

 The control device for a traffic light object with a computer with neuroprograms comprises a controller 17 with a computer equipped with a neuroprogram that has been preliminarily trained in a practical selection of tasks that takes into account the flexible change of control modes in time, redundancy, emergency situations, as well as preferences for special vehicles and correction of incoming information from neighboring traffic lights and operating on the basis of information from the transport detector 12 at this intersection 13 and from neighboring traffic lights 8, 9, 14, 15, 16 and through them with other using traffic lights based on correcting preference coefficients from control traffic lights 8, 9, 10, 11 at the "inputs / outputs" of regulation area 1, while a computer controller with a neuro program is connected by a transceiver information network to neighboring traffic lights 8, 9, 14, 15 , 16 and through them with other traffic light objects through the transceiver 18 and the antenna 19 (Fig. 2).

 A traffic detector 24 with a computer, equipped with a neuroprogram preliminarily trained to recognize the type and quantity of transport at the intersection, its direction, speed, lane based on information from transport sensors 12 installed on each, is introduced into the control device of a traffic light object with a computer with neuroprograms. lane at the intersection, connected by the output to the input of the controller 7 with a computer with a neuroprogram.

 The neuroprogram of the computer of the controller is trained on the basis of full-scale surveys and correcting coefficients from control traffic lights and which can be learned 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 based on correcting preference coefficients. During operation, each traffic light object receives additional information and performs its control work by completing what is required by the neuroprogram algorithm, i.e. traffic light objects work with direct and feedback, interact with each other.

 The device operates as follows.

The control device for a traffic light object with a computer with neuroprograms works, for example, at the traffic light object 13, as follows:
The controller with a computer with neuroprogram 17, previously trained on a practical selection of tasks that takes into account the flexible change of modes over time, redundancy, emergency situations, as well as preferences for special vehicles and correction of incoming information from neighboring and control traffic lights, receives information through the antenna 19 of the transceiver 18 from of all traffic light objects, as well as information from a transport detector 24 with a computer with a neuro program, previously trained to recognize the type and number of transpo that at the intersection, the direction of motion, speed, traffic lane based on information from the vehicle sensors 12.

 A computer controller with neuroprogram 17 can also receive information from other structures of the traffic flow control system.

 A controller 17 with a computer with a neuroprogram controls traffic flows through a port board 20 that gives on-off signals for certain groups of lamps 21, 22, 23 of a traffic light object 13, taking into account information on and through neighboring traffic light objects 8, 9, 14, 15, 16 with other traffic light objects, taking into account information based on correcting preference factors from control traffic light objects 8, 9, 10, 11.

 The controller 17 with a computer with a neuroprogram has a pre-trained neuroprogram based on field examinations based on correcting preference coefficients 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, i.e. traffic light objects work with direct and feedback, interact with each other.

 The training scheme of the neural program of the computer of the controller 17 is shown in FIG. 3 and is carried out by a known method [3, 7].

The functioning diagram of the computer neuroprogram of the controller 17 is shown in FIG. 4 and is carried out as follows:
The input of information from neighboring traffic light objects 8, 9, 14, 15, 16 and through them with other traffic light objects and information about the traffic flow from a transport detector with a computer with a neuroprogram 24 is received by a trained computer with a neuroprogram of the control controller 17, receiving time burning of the green signal depending on the traffic intensity with further calculation of the control cycle by the green signal with the output of the traffic light objects 21, 22, 23. Calculation of the control cycle by the green signal y is transmitted to the adjacent traffic lights 8, 9, 14, 15, 16 and through them to other traffic lights. If the transit time of the transport stream is longer than the set time, then corrective control signals are received from the control traffic light objects 8, 9, 10, 11.

 The training scheme of the neural program of the computer of the transport detector 24 is shown in FIG. 5.

The functioning of the neural program of the computer of the transport detector 24 is shown in FIG. 6 and includes:
input information input from transport sensors 12 installed in each lane,
the functioning of a trained computer with the neural program of the transport detector 24, obtaining traffic intensity,
transmitting the received traffic flow intensity to a computer with a controller neuroprogram 17.

 Execution example. The traffic flow approaching along the highway to a traffic light object with a computer with a neuroprogram 13 is recorded by transport sensors 12 and the information enters the transport detector 24 with a computer with a neuroprogram pre-trained to recognize the type and amount of transport at the intersection, its direction of movement, speed, lane, and information about the parameters of traffic flows in the directions enters the controller 17 with a computer with a neuroprogram, in which, taking into account information from neighboring traffic lights 8, 9, 14, 15, 16 and through them with other traffic light objects, where a control signal is generated for lamps 21, 22, 23 of the traffic light object 13 with a computer with neuroprograms, for example, green.

 If the skip time of the transverse transport does not satisfy the control preference, then the burning time (for example, the green signal) along the main highway increases.

Learning the neural program of the microcomputer 17 of the controller is as follows (see Fig. 3):
- Entering input information from sensors 12 vehicles installed in each lane.

 - The functioning of the controller pre-trained computer with the neural program of the transport detector 24. Getting traffic intensity.

 - Transfer of the received intensity to the computer with the neuroprogram of the controller 17.

 A traffic light object with a controller with a microcomputer with a neuroprogram is a local subsystem, which, according to its own neuroprogram, makes the optimal decision on the bandwidth based on information: from neighboring traffic light objects - the pass of cars that passed in its direction taking into account the travel time from the neighboring to the given traffic light object.

 Our own neuroprogram is trained on a selection of tasks that takes into account the flexible change of control modes in time, redundancy, emergency situations, etc., including the passage of the traffic flow through the intersection from maximum to minimum, especially highlighting the directions and flows of public transport, special vehicles indicating their preferences taking into account all local conditions and restrictions, work according to the incoming information according to the criterion of minimum delays (by maximum throughput) and rektiruyuschih control signals to control traffic lights 8, 9, 10, 11.

 The transport detector 24 with a computer with a neuroprogram can be performed on sound, video and magnetic basis.

 A controller 17 with a computer with a neuroprogram was used to control the bearing capacity of a multi-span continuous beam and showed good results [6].

Literature
1. A. p. N 1462395 A method for controlling the movement of vehicles at an intersection. Publ. BI N 8, 1989

 2. Remote control system for traffic lights and its components. Technical task. The product was developed and manufactured by the firm "Level", address 690105, Vladivostok, st. Borodinskaya 46/50, PO Box 105-59 (prototype).

 3. Bartsev S.I., Gilev S.E., Okhonin V.A. The duality principle in the organization of adaptive information processing networks / Collection "Dynamics of chemical and biological systems" // Novosibirsk, Nauka, 1989

 4. Introduction to artificial neural networks. Anil C. Jane / Journal of Open Systems. Neurocomputers. N 4 (24), 1997, p. 16.

 5. Kremenets Yu.A. Technical means of traffic management. M. Transport, 1990).

 6. A method for automatically controlling the bearing capacity of a multi-span continuous beam using a controller with a neural network program and a device for its implementation. Application for invention N 97107519, according to which a decision was obtained on the grant of a patent of the Russian Federation dated April 29, 1998).

 7. Automatic control of structures using neural networks. Abovsky N.P., Abrosimov P.S., Babanin V.B., Lankin Yu.P., Smolyaninova L.G., KrasGASA, Krasnoyarsk, 1996, p. 88.

 8. The method of coordinated neural network traffic management in the regulatory area. Application for invention N 97117398. Babanin V.B., Smolyaninova L.G., Abovsky N.P. and etc.

Claims (1)

 A traffic light control device with a computer with neuroprograms, consisting of a controller containing a computer, a real-time clock, a non-volatile memory with the configuration of the intersection recorded in it, connected by a bi-directional channel to a port card connected to the traffic light lamps, and the controller is connected to the transceiver information network with neighboring and control traffic light objects through a transceiver and an antenna, characterized in that the device has a transport detector equipped with a computer with no reprogram, pre-trained to recognize at the crossroads the type and quantity of transport, its direction of movement, speed, lane based on information from transport sensors, as well as the fact that the computer controller is equipped with a neuro program pre-trained on a practical selection of tasks that take into account flexible change control modes in time, redundancy, emergency situations, as well as preferences for special vehicles and adjustment of incoming information from neighboring traffic lights and functioning based on information from a transport detector at this intersection and from neighboring traffic lights and based on correcting preference coefficients from control traffic lights at the inputs and outputs of the regulation area.

Images