CN-121982885-A - Intelligent decision method and system for port mobile signal lamp

CN121982885ACN 121982885 ACN121982885 ACN 121982885ACN-121982885-A

Abstract

The invention relates to an intelligent decision method and system for a port mobile signal lamp, wherein the method comprises the steps of collecting traffic element information in the action range of each mobile signal lamp in the port in real time, constructing a local traffic state vector, carrying out feature fusion to obtain an overall traffic state vector after carrying out synchronization and time sequence correction on the local traffic state vector based on a synchronization correction function, taking the overall traffic state vector as a state, utilizing a reinforcement learning model to output the action of each mobile signal lamp at the current moment, wherein the action meets phase constraint and position movement constraint, carrying out online training on the reinforcement learning model based on an instant rewarding function, calculating average rewarding lifting amount based on the reinforcement learning model after training, and updating model parameters if the average rewarding lifting amount is larger than a preset value, wherein the system is used for realizing the method. Compared with the prior art, the invention provides a signal lamp tube control method capable of dynamically shifting along with the operation requirement of a port.

Inventors

WANG LEI
WU XUDONG
WENG JINXIAN
Sun Zhouwei

Assignees

上海海事大学

Dates

Publication Date: 20260505
Application Date: 20260115

Claims (10)

1. The intelligent decision method for the port mobile signal lamp is characterized by comprising the following steps of: the method comprises the steps of collecting traffic element information in the action range of each mobile signal lamp in a port in real time, constructing a local traffic state vector, carrying out synchronization and time sequence correction on the local traffic state vector based on a synchronization correction function, and carrying out feature fusion to obtain an overall traffic state vector; The method comprises the steps of taking the whole traffic state vector as a state, utilizing a reinforcement learning model to output the action of each mobile signal lamp at the current moment, enabling the action to meet phase constraint and position movement constraint and comprise phase change quantity and mobile target position correction quantity, carrying out online training on the reinforcement learning model based on an instant rewarding function, calculating average rewarding lifting quantity based on the trained reinforcement learning model, updating model parameters if the average rewarding lifting quantity is larger than a preset value, and otherwise, not updating the model parameters.
2. The intelligent decision-making method for port mobile signal lamp according to claim 1, wherein said traffic element information includes lane number, number of vehicles in each lane, vehicle identification in each lane and average speed of each lane, said vehicle identification is 01-type variable, and when When representing a manually driven vehicle, when The time of which means that the vehicle is not driven by a person, Represent the first A vehicle; Then for any of the mobile signal lights, in The local traffic state vector at the moment is: , represent the first Lane numbering of lanes; Is shown in Time of day (time) The number of vehicles in the lane; Is shown in Time of day (time) Lane in lane no Vehicle identification of the vehicle; Is shown in Time of day (time) Average speed of lane; represent the first And a mobile signal lamp.
3. The intelligent decision method for port mobile signal lamp according to claim 1, wherein the synchronization correction function is: , Expressed in the correction time sequence Is a function of the local traffic state vector of (a), Represent the first The data transmission delay of the individual signal lamps, Is shown in Local traffic state vector at.
4. The intelligent decision-making method for port mobile signal lamp according to claim 1, wherein said phase constraint is , Indicating the duration of the green light in phase switching, And Respectively representing the minimum value and the maximum value of the phase green light duration; the position movement constraint comprises a position movement trigger constraint and a position movement distance constraint, wherein the position movement trigger constraint is that and only that The position movement distance constraint is that each position movement of any moving signal lamp does not exceed a preset distance; indicating the update period in which the kth mobile signal is located, Representing the update trigger period of a kth mobile signal lamp; And has 。
5. The intelligent decision-making method for port mobile signal lamp according to claim 1, wherein said instant prize function is: , Wherein, the And All represent weight coefficients and have ; The average passing speed of the unmanned collecting card is represented; The maximum passing speed of the unmanned integrated card theory is represented; representing the average waiting time of the vehicle; Indicating a vehicle reference waiting time.
6. An intelligent decision system for a port mobile signal lamp, which is characterized by comprising: The local sensing unit comprises a short-range radar, a camera and a positioning module, and is used for collecting traffic element information in the action range of each mobile signal lamp in a port in real time, collecting environment change parameters after the mobile signal lamp unit executes actions output by the port area edge calculation scheduling unit, and transmitting the traffic element information and the environment change parameters to the mobile signal lamp unit; the mobile signal lamp unit comprises a signal lamp device, a local controller and a wireless communication subunit, wherein the local controller is used for receiving the traffic element information, constructing a local traffic state vector, carrying out synchronization and time sequence correction on the local traffic state vector based on a synchronization correction function, and then carrying out feature fusion to obtain an overall traffic state vector; The port area edge calculation scheduling unit comprises a strategy management subunit, a training subunit and a database, wherein the strategy management subunit performs the following steps of taking the whole traffic state vector as a state, using a reinforcement learning model to output the action of each mobile signal lamp at the current moment, an instant rewarding value and a state vector at the next moment, wherein the actions meet phase constraint and position movement constraint and comprise phase change quantity and a movement target position correction quantity, the training subunit performs online training on the reinforcement learning model by using an instant rewarding function, calculates an average rewarding lifting quantity based on the trained reinforcement learning model, and updates model parameters if the average rewarding lifting quantity is larger than a preset value, otherwise does not update the model parameters.
7. The intelligent decision-making system for port mobile signal lamp according to claim 6, wherein said traffic element information includes lane number, number of vehicles in each lane, vehicle identification in each lane and average speed of each lane, said vehicle identification is type 01 variant, and when When representing a manually driven vehicle, when The time of which means that the vehicle is not driven by a person, Represent the first A vehicle; Then for any of the mobile signal lights, in The local traffic state vector at the moment is: , represent the first Lane numbering of lanes; Is shown in Time of day (time) The number of vehicles in the lane; Is shown in Time of day (time) Lane in lane no Vehicle identification of the vehicle; Is shown in Time of day (time) Average speed of lane; represent the first And a mobile signal lamp.
8. The intelligent decision system for port mobile signal lamp as recited in claim 6, wherein said synchronization correction function is: , Expressed in the correction time sequence Is a function of the local traffic state vector of (a), Represent the first The data transmission delay of the individual signal lamps, Is shown in Local traffic state vector at.
9. The intelligent decision system of a port mobile signal lamp according to claim 6, wherein the phase constraint is: , indicating the duration of the green light in phase switching, And Respectively representing the minimum value and the maximum value of the phase green light duration; the position movement constraint comprises a position movement trigger constraint and a position movement distance constraint, wherein the position movement trigger constraint is that and only that The position movement distance constraint is that each position movement of any moving signal lamp does not exceed a preset distance; indicating the location update period in which the kth mobile signal is located, A position update trigger period representing a kth mobile signal lamp; And has 。
10. The intelligent decision system for port mobile signal lights according to claim 6, wherein said instant prize function is: , Wherein, the And All represent weight coefficients and have ; The average passing speed of the unmanned collecting card is represented; The maximum passing speed of the unmanned integrated card theory is represented; representing the average waiting time of the vehicle; Indicating a vehicle reference waiting time.

Description

Intelligent decision method and system for port mobile signal lamp Technical Field The invention relates to the technical field of intelligent traffic and traffic control, in particular to an intelligent decision method and system for a port mobile signal lamp. Background Along with the promotion of automatic harbor construction, unmanned container transport vehicles (unmanned inner collector cards) have been born in harbor areas for horizontal transport tasks from a yard to the front edge of a wharf, thereby remarkably saving manpower and improving efficiency. However, at present, most ports only realize unmanned operation in the interior, and container trucks (external collector cards) transported to the outside still rely on manual driving, so that a typical scene of mixing unmanned vehicles and manned vehicles in the ports is formed. In the mixed running environment, the unmanned inner vehicle generally runs along a preset track, and a perception system is relied on to identify obstacles and slow down or stop the vehicle to ensure safety. And the manual driving integrated card is influenced by subjective judgment and operation time pressure, and the actions such as robbing, temporary lane changing and the like often occur, so that traffic collision frequently occurs. Once the unmanned vehicle meets the manual driving vehicle, the unmanned system always completely avoids or waits for a long time due to a safety strategy, traffic stagnation is easy to cause, and the overall transportation efficiency of a storage yard and a wharf area is reduced. Although unmanned vehicles operate efficiently in a regularized environment, in a dynamic hybrid scenario containing human intervention, their decisions are still limited by the conservation of perceived accuracy and driving strategy. Therefore, it is difficult to realize efficient and orderly mixed traffic by means of an automatic driving system alone. It is necessary to introduce a traffic signal control and behavior guiding mechanism to reasonably schedule and guide the manually driven vehicles, and coordinate the passing rights of different vehicles from the system level, so as to reduce conflict, improve the overall working efficiency and ensure the stable operation of the unmanned transportation system. Most of the conventional port traffic control systems depend on fixed signal lamps and structural rules, the special characteristics of port mixed traffic environments are not considered, such as China patent application CN113538937A, the port mixed traffic control system is provided, adjacent unmanned collecting card lanes and manual collecting card lanes are arranged in ports, the orderly traffic of the manual collecting cards and the unmanned collecting cards is realized through the cooperation of hardware such as traffic control centers, traffic signal lamps, ground lifting upright posts, ground indication lamps and the like, when the unmanned collecting cards need to use the manual collecting card lanes, the application is submitted to the control center, after the control center judges that the manual collecting cards pass, the signal lamps are controlled to display forbidden traffic, the lifting upright posts are lifted, the ground indication lamps are warned, the manual collecting card traffic is limited, and the safe use of the application area of the unmanned collecting cards is ensured. Although the intelligent safe passing of the artificial collecting card and the unmanned collecting card is guaranteed to a certain extent, the intelligent safe passing of the artificial collecting card and the unmanned collecting card is dependent on fixed lane division, fixed signal facility positions and preset passing rules, a harbor road presents non-standardization and high dynamic performance, an operation route is frequently adjusted, a plurality of types of main bodies of an artificial driving vehicle, the unmanned collecting card and special equipment are mixed, the speed difference is obvious, the behavior randomness is high, traffic flow is driven by operation tasks, the characteristics of burstiness, short-time concentration and unbalanced direction are presented, the regulation and the control are difficult to be carried out by depending on fixed phase logic, meanwhile, the harbor road is difficult to use fixed signal facilities, and the signal arrangement positions need to be continuously migrated along with the operation. Therefore, it is a technical problem to provide a method for controlling a signal lamp capable of dynamically moving along with the operation requirement of a port. Disclosure of Invention The invention aims to overcome the defects of the prior art and provide an intelligent decision method and system for a port mobile signal lamp, which are used for constructing an instant rewarding function aiming at traffic efficiency, safety and operation scheduling cooperativity by taking port traffic information as state representation and rea