Multi-objective optimization ai traffic signal control method and system

Abstract

The application discloses a multi-target optimization AI traffic signal control method and system, and relates to the field of intelligent traffic systems.The method comprises the following steps: establishing a traffic signal control model, wherein the model comprises multiple optimization targets, and each optimization target is configured with an initial weight; setting a bone conduction sensor network in a traffic signal control area to obtain real-time motion information of blind users; detecting whether there is a visually impaired person; if there is a visually impaired person, adjusting the initial weight of each optimization target to obtain a dynamically adjusted weight; running the traffic signal control model according to the dynamically adjusted weight to generate a traffic signal control instruction; and controlling the display state of a traffic signal lamp according to the traffic signal control instruction, and simultaneously starting a variable sound and light guiding system to guide the visually impaired person to pass through an intersection.The application solves the technical problem that the existing multi-target optimization AI traffic signal control cannot meet the travel demand of visually impaired persons, and achieves the technical effect of meeting the travel demand of visually impaired persons.

Description

Technical Field

[0001] This application relates to the field of intelligent transportation systems, and in particular to a multi-objective optimization AI traffic signal control method and system. Background Technology

[0002] With the acceleration of urbanization, traffic congestion, pedestrian safety, and the convenience of travel for special groups (especially the visually impaired) have become critical issues that urgently need to be addressed in urban traffic management. These issues not only affect urban traffic efficiency but also directly relate to the quality of life and public safety of citizens. Currently, the main approach to solving these problems is to use traditional traffic signal control systems. These systems are typically designed with signal control strategies based on fixed optimization objectives (such as maximizing traffic flow) and are supplemented with basic pedestrian crossing facilities. However, current methods lack detailed consideration and dynamic adaptability to the needs of different travel groups (especially the visually impaired), resulting in significant deficiencies in ensuring pedestrian safety, particularly the safe travel of the visually impaired.

[0003] At present, multi-objective optimization AI traffic signal control has technical problems that make it difficult to meet the travel needs of visually impaired people. Summary of the Invention

[0004] This application provides a multi-objective optimization AI traffic signal control method and system. It employs a traffic signal control model that incorporates traffic flow, pedestrian safety, and optimization for visually impaired individuals' travel. Initial weights are assigned to each objective. A bone conduction sensor network is installed within the control area to acquire real-time motion information of users on tactile paving. Based on this real-time motion information, the system detects whether visually impaired individuals are using the tactile paving. If a visually impaired individual is detected, the weights of each optimization objective are dynamically adjusted. Based on the adjusted weights, the model generates traffic signal control commands, controls traffic lights according to the commands, and activates a variable sound and light system to guide visually impaired individuals safely across the street. These technical means effectively meet the travel needs of visually impaired individuals.

[0005] This application provides a multi-objective optimization AI traffic signal control method, comprising: establishing a traffic signal control model, the model including multiple optimization objectives, specifically traffic flow optimization, pedestrian safety optimization, and travel optimization for visually impaired persons, each optimization objective being configured with initial weights; setting up a bone conduction sensor network within the traffic signal control area, and acquiring real-time motion information of tactile paving users based on the bone conduction sensor network; detecting the presence of visually impaired persons based on the real-time motion information; if visually impaired persons are present, adjusting the initial weights of each optimization objective to obtain dynamically adjusted weights; running the traffic signal control model based on the dynamically adjusted weights to generate traffic signal control commands; and controlling the display state of traffic lights based on the traffic signal control commands, while simultaneously activating a variable audio-visual guidance system to guide visually impaired persons across intersections.

[0006] In a possible implementation, the establishment of the traffic signal control model involves the following processes: establishing a data acquisition layer to collect traffic flow data, pedestrian flow data, and motion data of visually impaired individuals based on sensors; establishing a data processing layer to perform filtering, noise reduction, and feature extraction preprocessing on the collected data; establishing an optimization decision layer to dynamically adjust the phase and duration of traffic lights based on the preprocessed data using a chaotic particle swarm optimization algorithm and a dynamic weight adjustment strategy, thereby generating traffic signal control commands; establishing an execution layer to control the display state of the traffic lights according to the traffic signal control commands; and integrating the data acquisition layer, the data processing layer, the optimization decision layer, and the execution layer to establish the traffic signal control model.

[0007] In a possible implementation, a bone conduction sensor network is set up within the traffic signal control area. Based on the real-time motion information of the tactile paving user obtained from the bone conduction sensor network, the following processing is performed: tactile paving layout information within the traffic signal control area is obtained, and the tactile paving boundary contour is extracted from the layout information; the tactile paving complexity and area are calculated based on the boundary contour, and the number and location of sensors are matched and obtained based on the complexity and area; multiple distributed bone conduction sensors are deployed on the tactile paving according to the number and location of the sensors, resulting in a bone conduction sensor network; bone conduction signals are obtained from the bone conduction sensor network, and the bone conduction signals are filtered, amplified, and have motion features extracted to obtain the real-time motion information of the tactile paving user.

[0008] In a possible implementation, if a visually impaired person is present, the following processing is also performed: based on the real-time motion information and the nearest distance, a first distributed bone conduction sensor is determined; based on the bone conduction sensor network, a first neighboring bone conduction sensor of the first distributed bone conduction sensor is determined; and the first distributed bone conduction sensor and the first neighboring bone conduction sensor are activated to a high-sensitivity mode.

[0009] In a possible implementation, the step of detecting whether a visually impaired person exists based on the real-time motion information involves performing the following processing: filtering the real-time motion information for environmental noise to obtain noise-reduced motion information; extracting visual impairment features from the noise-reduced motion information to obtain a visually impaired motion pattern signal; and comparing the visually impaired motion pattern signal with a confidence threshold to detect whether a visually impaired person exists.

[0010] In a possible implementation, if visually impaired individuals exist, the initial weights of each optimization objective are adjusted to obtain dynamically adjusted weights, and the following processing is performed: if visually impaired individuals exist, the number and location information of visually impaired individuals are obtained based on the real-time motion information; a motion dot matrix map of visually impaired individuals is generated based on the number and location information of visually impaired individuals; the relative positional relationship between the motion dot matrix map of visually impaired individuals and the zebra crossing at the intersection is calculated to obtain a travel weight adjustment coefficient for visually impaired individuals; and the initial weights of each optimization objective are adjusted based on the travel weight adjustment coefficient for visually impaired individuals to obtain dynamically adjusted weights.

[0011] In a possible implementation, the following processing is performed: the variable sound and light guidance system includes a traffic light brightness adjustment module and an audio prompt module. The traffic light brightness adjustment module is used to dynamically adjust the brightness of the traffic lights according to the location of the visually impaired person and traffic flow. The audio prompt module is used to dynamically adjust the intensity and content of the audio prompts according to the location of the visually impaired person and traffic signal status.

[0012] This application also provides a multi-objective optimization AI traffic signal control system, comprising: a traffic signal control model establishment module for establishing a traffic signal control model, the model including multiple optimization objectives, specifically traffic flow optimization, pedestrian safety optimization, and travel optimization for visually impaired persons, each optimization objective being configured with initial weights; a real-time motion information acquisition module for setting up a bone conduction sensor network within the traffic signal control area and acquiring real-time motion information of tactile paving users based on the bone conduction sensor network; a visually impaired person detection module for detecting the presence of visually impaired persons based on the real-time motion information; a weight adjustment module for adjusting the initial weights of each optimization objective if visually impaired persons are present, obtaining dynamically adjusted weights; a traffic signal control command generation module for running the traffic signal control model based on the dynamically adjusted weights and generating traffic signal control commands; and a traffic control module for controlling the display state of traffic lights according to the traffic signal control commands, and simultaneously activating a variable audio-visual guidance system to guide visually impaired persons across intersections.

[0013] The proposed multi-objective optimization AI traffic signal control method and system first establishes a traffic signal control model, which includes multiple optimization objectives: traffic flow optimization, pedestrian safety optimization, and travel optimization for visually impaired individuals. Each objective is assigned an initial weight. Then, a bone conduction sensor network is set up within the traffic signal control area to acquire real-time motion information of users on the tactile paving. Next, based on this real-time motion information, the presence of visually impaired individuals is detected. If a visually impaired individual is found, the initial weights of each optimization objective are adjusted to obtain dynamically adjusted weights. Based on these dynamically adjusted weights, the traffic signal control model is run to generate traffic signal control commands. Finally, based on these commands, the display status of traffic lights is controlled, and a variable audio-visual guidance system is activated to guide visually impaired individuals across intersections. This achieves the technical effect of meeting the travel needs of visually impaired individuals. Attached Figure Description

[0014] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings of the embodiments of the present invention will be briefly described below. Flowcharts are used in this application to illustrate the operations performed by the system according to the embodiments of the present application. It should be understood that the preceding or following operations are not necessarily performed precisely in sequence. Instead, various steps can be processed in reverse order or simultaneously as needed. Furthermore, other operations can be added to these processes, or one or more steps can be removed from these processes.

[0015] Figure 1 This is a flowchart illustrating the multi-objective optimization AI traffic signal control method provided in an embodiment of this application.

[0016] Figure 2 This is a schematic diagram of the structure of a multi-objective optimization AI traffic signal control system provided in an embodiment of this application.

[0017] Figure labeling: Traffic signal control model establishment module 10, real-time motion information acquisition module 20, visually impaired person detection module 30, weight adjustment module 40, traffic signal control instruction generation module 50, traffic control module 60. Detailed Implementation

[0018] The above description is merely an overview of the technical solution of this application. In order to better understand the technical means of this application and to implement it in accordance with the contents of the specification, and to make the above and other objects, features and advantages of this application more obvious and understandable, specific embodiments of this application are given below.

[0019] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description of this application will be provided in conjunction with the accompanying drawings. The described embodiments should not be considered as limitations on this application. All other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0020] In the following description, references to "some embodiments" describe a subset of all possible embodiments. However, it is understood that "some embodiments" can be the same or different subsets of all possible embodiments and can be combined with each other without conflict. The terms "first" and "second" are used merely to distinguish similar objects and do not represent a specific ordering of objects. The terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or server that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or modules not explicitly listed or inherent to these processes, methods, products, or devices. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used herein is for the purpose of describing embodiments of this application only.

[0021] This application provides a multi-objective optimization AI traffic signal control method, such as... Figure 1 As shown, the method includes:

[0022] Step S100: Establish a traffic signal control model. The model includes multiple optimization objectives, specifically traffic flow optimization, pedestrian safety optimization, and travel optimization for visually impaired persons. Each optimization objective is configured with an initial weight.

[0023] Specifically, three optimization objectives are defined for the traffic signal control model: traffic flow optimization, pedestrian safety optimization, and travel optimization for visually impaired individuals. The traffic signal control model is a mathematical model or simulation system for optimizing traffic signal control, capable of considering multiple optimization objectives and generating corresponding traffic signal control commands. Traffic flow optimization is achieved by reducing vehicle waiting time and improving road capacity; pedestrian safety optimization ensures pedestrians have sufficient time to cross the road and reduces conflicts with vehicles; and travel optimization provides safe crossing conditions for visually impaired individuals, including appropriate waiting time and clear audio-visual guidance. An initial weight is assigned to each optimization objective, reflecting the relative importance of each objective in the early stages of model operation. The weight assignment can be based on historical data, expert opinions, or simulation results. The traffic signal control model is constructed using mathematical or computer simulation software (such as MATLAB, Python, etc.), with the optimization objectives and initial weights as input parameters.

[0024] In one possible implementation, the establishment of the traffic signal control model, step S100 further includes step S110, setting up a data acquisition layer to collect traffic flow data, pedestrian flow data, and motion data of visually impaired individuals based on sensors. Specifically, the data acquisition layer is responsible for collecting and transmitting various data required for traffic signal control. Within the traffic signal control area, different types of sensors are deployed according to actual needs, such as vehicle flow sensors (e.g., geomagnetic sensors, infrared sensors), pedestrian flow sensors (e.g., cameras, infrared sensors), and bone conduction sensors specifically for detecting visually impaired individuals (or combined with other technologies such as RFID tags, smartphone apps, etc., if the visually impaired individual carries specific devices). The sensors collect traffic flow data (including vehicle quantity, speed, type, etc.), pedestrian flow data (including pedestrian quantity, walking speed, direction, etc.), and motion data of visually impaired individuals (e.g., location, speed, direction, etc.) in real time. The collected data is transmitted to the data processing center via wired or wireless means (e.g., Wi-Fi, Bluetooth, LoRa, NB-IoT, etc.).

[0025] Step S120: A data processing layer is built to preprocess the collected data through filtering, noise reduction, and feature extraction. Specifically, the data processing layer is responsible for data preprocessing, storage, and quality control. It performs filtering (removing noise components from the signal and improving the signal-to-noise ratio), noise reduction (similar to filtering but focusing on reducing random errors and noise in the data), and feature extraction (extracting key information such as peak traffic flow and pedestrian flow patterns). The preprocessed data is stored in a database for subsequent analysis and use. Data accuracy and integrity are ensured through verification and outlier detection.

[0026] Step S130: Establish an optimization decision layer. Based on the preprocessed data, utilize the chaotic particle swarm optimization algorithm and dynamic weight adjustment strategy to dynamically adjust the phase and duration of traffic lights, generating traffic signal control commands. Specifically, the optimization decision layer is responsible for generating traffic signal control commands based on real-time data, employing the chaotic particle swarm optimization algorithm (CPSO). This algorithm combines the advantages of chaos theory and particle swarm optimization, enabling it to find better solutions to complex multi-objective optimization problems. The weights of each optimization objective are dynamically adjusted based on real-time traffic conditions (such as vehicle flow, pedestrian flow, and the number of visually impaired individuals) to balance the relationship between traffic flow optimization, pedestrian safety optimization, and visually impaired individual travel optimization. Based on the preprocessed data and the dynamic weight adjustment strategy, the phase (the switching order of red, green, and yellow lights) and duration (the duration of each phase) of the traffic lights are calculated and generated.

[0027] Step S140: Establish an execution layer to control the display state of traffic lights according to the traffic signal control instructions. Specifically, the execution layer is responsible for controlling the display state of traffic lights according to the traffic signal control instructions, transmitting the generated traffic signal control instructions to the traffic light control system through a communication network, and the traffic light control system adjusts the display state of the traffic lights (red light, green light, yellow light) according to the received instructions.

[0028] Step S150: Integrate the data acquisition layer, data processing layer, optimization decision layer, and execution layer to establish a traffic signal control model. Specifically, the data acquisition layer, data processing layer, optimization decision layer, and execution layer are integrated through a software architecture (such as microservice architecture, distributed system, etc.) to form a complete traffic signal control model. The model's performance and effectiveness are verified through simulation testing and field testing to ensure it meets actual needs. Based on the test results, the model is optimized and adjusted to improve its accuracy and reliability. This implementation method, by building a data acquisition layer, data processing layer, optimization decision layer, and execution layer and integrating them into a complete traffic signal control model, can achieve a balance between multiple objectives while ensuring traffic flow optimization and considering pedestrian safety and travel optimization for visually impaired individuals.

[0029] Step S200: Set up a bone conduction sensor network in the traffic signal control area, and obtain real-time motion information of tactile paving users based on the bone conduction sensor network.

[0030] Specifically, suitable sensors for bone conduction are selected, capable of detecting real-time motion information of tactile paving users, such as position, speed, and direction. A sensor network is deployed along the tactile paving within traffic signal-controlled areas, ensuring coverage of all important intersections and pedestrian crossings. Sensors are connected via wired or wireless means to form a complete monitoring system. The sensor network collects motion information from tactile paving users in real time and transmits the data to a central control system for processing. Bone conduction sensors, which detect human movement by transmitting sound or vibration signals through bone, are used to detect the real-time motion information of tactile paving users.

[0031] In one possible implementation, the step of setting up a bone conduction sensor network within the traffic signal control area and acquiring real-time movement information of tactile paving users based on the bone conduction sensor network further includes step S210, acquiring tactile paving layout information within the traffic signal control area and extracting the tactile paving boundary contours from the tactile paving layout information. Specifically, the tactile paving layout information within the traffic signal control area is acquired using a Geographic Information System (GIS) or an electronic map provided by the urban planning department. This information exists in vector data form, including the starting point, ending point, and path of the tactile paving. Using GIS software or image processing algorithms, the acquired tactile paving layout information is parsed to extract the boundary contours of the tactile paving, that is, the vector data is converted into polygon or polyline representations for analysis.

[0032] Step S220: Calculate the complexity and area of ​​the tactile paving based on its boundary contour. Then, match and obtain the number and location of sensors based on these parameters. Specifically, calculate the complexity of the tactile paving based on the extracted boundary contour. Complexity can be measured by various indicators, such as the length, tortuosity, and number of intersections of the tactile paving, which can be automatically calculated by an algorithm. Similarly, based on the extracted boundary contour, calculate the area covered by the tactile paving by calculating the area of ​​the polygon or the region enclosed by the polyline. Based on the complexity and area of ​​the tactile paving, as well as parameters such as the coverage and sensitivity of the bone conduction sensors, calculate the required number and location of sensors using an algorithm to minimize the number of sensors while meeting coverage requirements.

[0033] Step S230: Based on the number and location of the sensors, multiple distributed bone conduction sensors are deployed on the tactile paving to obtain a bone conduction sensor network. Specifically, based on the calculation results of step S220, appropriate bone conduction sensor models and quantities are selected. Sensors are deployed on the tactile paving according to the deployment locations and quantities determined in step S220, either by burying the sensors under the tactile paving surface or fixing them to the surface. During deployment, ensure the sensors are accurately positioned, securely fixed, and avoid interfering with tactile paving users.

[0034] Step S240 involves acquiring bone conduction signals from the bone conduction sensor network, filtering, amplifying, and extracting motion features from these signals to obtain real-time motion information of the user on the tactile paving. Specifically, when a user walks on the tactile paving, the vibration of their bones generates bone conduction signals. These signals are captured by sensors deployed on the tactile paving and converted into electrical signals. Digital or analog filters are used to filter the acquired bone conduction signals to remove noise and interference. To improve the signal-to-noise ratio and facilitate subsequent processing, the filtered signals are amplified. Useful motion features, such as walking speed, direction, and cadence, are extracted from the amplified signals. These features can be automatically calculated by algorithms and used for subsequent analysis and decision-making. This implementation calculates the complexity and area of ​​the tactile paving and matches the number and location of sensors based on these parameters, ensuring that the coverage and detection accuracy of the sensor network meet actual needs, thus establishing an efficient bone conduction sensor network.

[0035] Step S300: Based on the real-time motion information, detect whether there is a visually impaired person.

[0036] Specifically, the central control system receives and analyzes data from the sensor network. By comparing real-time motion information with preset thresholds or patterns, it determines whether a visually impaired person is present.

[0037] In one possible implementation, the step of detecting the presence of a visually impaired person based on the real-time motion information further includes step S310, which involves filtering the real-time motion information for environmental noise to obtain noise-reduced motion information. Specifically, environmental noise in the real-time motion information is identified by analyzing outliers, high-frequency noise, or interference signals that do not conform to other normal motion patterns. Once environmental noise is identified, filtering algorithms (such as Kalman filtering, median filtering, or low-pass filtering) are applied to remove or reduce this noise. After filtering, noise-reduced motion information is output, which more accurately reflects the actual motion state of the user on the tactile paving.

[0038] Step S320: Visual impairment feature extraction is performed on the noise-reduced motion information to obtain visual impairment motion pattern signals. Specifically, based on known characteristics of visually impaired individuals' walking (such as unsteady gait, slow pace, frequent pauses, etc.), relevant features are selected and extracted from the noise-reduced motion information. Machine learning algorithms (such as support vector machines, neural networks, or decision trees) are used to classify and identify the extracted features. These algorithms are trained to identify features related to the walking patterns of visually impaired individuals. The system outputs the identified visual impairment motion pattern signals, which represent specific characteristics of the tactile paving user's walking.

[0039] Step S330 involves comparing the visually impaired movement pattern signal with a confidence threshold to detect the presence of a visually impaired person. Specifically, the confidence level of the identified visually impaired movement pattern signal is calculated by evaluating the signal's strength, stability, and its matching degree with other known visually impaired walking patterns. Based on system performance and accuracy requirements, a confidence threshold is set to determine whether the identified signal is reliable enough to confirm the presence of a visually impaired person. The calculated confidence level is compared with the set threshold; if the confidence level is higher than the threshold, the system considers the presence of a visually impaired person; otherwise, it considers the absence of a visually impaired person. This implementation, through environmental noise filtering and visual impairment feature extraction, can more accurately identify signals related to the walking of visually impaired individuals, thereby reducing false alarms and false negatives and ensuring accurate detection of visually impaired individuals under different circumstances.

[0040] In step S400, if there are visually impaired individuals, adjust the initial weights of each optimization objective to obtain dynamically adjusted weights.

[0041] Specifically, when a visually impaired person is detected, the initial weights of each optimization objective are dynamically adjusted according to a pre-defined weight adjustment strategy. This strategy can be based on factors such as the number, location, and speed of visually impaired individuals. New dynamically adjusted weights are then calculated based on the weight adjustment strategy and input into the traffic signal control model.

[0042] In one possible implementation, if visually impaired individuals are present, the initial weights of each optimization objective are adjusted to obtain dynamically adjusted weights. Step S400 further includes step S410, whereby, if visually impaired individuals are present, the number and location information of visually impaired individuals are obtained based on the real-time motion information. Specifically, based on the identified walking characteristics of visually impaired individuals, the number and relative positions of visually impaired individuals are estimated using the layout and signal strength differences of the sensor network.

[0043] Step S420: Generate a motion dot plot of visually impaired individuals based on their number and location information. Specifically, visualization software or libraries (such as Matplotlib, ECharts, etc.) are used to convert the number and location information of visually impaired individuals into a dot plot. In the dot plot, each dot represents a visually impaired individual, and the position and color of the dot can encode its location information and quantity (e.g., color intensity indicates quantity). By combining timestamp information, spatiotemporal data can be further integrated to generate a dynamic motion distribution map to show the changing trend of visually impaired individuals over time.

[0044] Step S430: Calculate the travel weight adjustment coefficient for visually impaired individuals based on the relative positional relationship between the movement map of visually impaired individuals and the zebra crossing at the intersection. Specifically, using a Geographic Information System (GIS) or map service API, the relative positional relationship between each point on the movement distribution map of visually impaired individuals and the zebra crossing at the intersection is analyzed through geometric operations such as distance calculation and direction determination. Based on the positional relationship analysis results, a travel priority adjustment strategy is formulated. For example, if visually impaired individuals are close to the zebra crossing and there are many of them, their travel priority needs to be increased; conversely, their priority should be decreased. According to the priority adjustment strategy, a travel priority adjustment factor for visually impaired individuals is calculated. This factor is used to adjust the weights of each optimization objective in subsequent steps.

[0045] Step S440: Adjust the initial weights of each optimization objective according to the travel weight adjustment coefficient for visually impaired persons to obtain dynamically adjusted weights. Specifically, based on the travel priority adjustment factor for visually impaired persons, apply a weight adjustment algorithm (such as linear interpolation, nonlinear mapping, etc.) to adjust the initial weights of each optimization objective. The adjusted weights are then used as new dynamic weights and updated in the traffic signal control model. In this way, the model can optimize in real time according to the current traffic conditions and the needs of visually impaired persons. This implementation dynamically adjusts the weights of each optimization objective in the traffic signal control model based on the real-time detected number and location information of visually impaired persons. By increasing the weight of optimizing travel for visually impaired persons, it ensures that traffic signal control pays more attention to the needs of visually impaired persons, thereby improving their safety when crossing intersections.

[0046] Step S500: Based on the dynamically adjusted weights, run the traffic signal control model and generate traffic signal control instructions.

[0047] Specifically, the traffic signal control model is run using dynamically adjusted weights as input parameters. The model employs an optimization algorithm to find the optimal traffic signal control scheme that simultaneously satisfies multiple optimization objectives. Based on the algorithm's results, corresponding traffic signal control commands are generated, including the display times and order of red, green, and yellow lights.

[0048] Step S600: According to the traffic signal control command, control the display status of the traffic lights and simultaneously activate the variable sound and light guidance system to guide visually impaired people through the intersection.

[0049] Specifically, the generated traffic signal control commands are sent to the traffic light control system to control the display status of the traffic lights. When a visually impaired person is detected, a variable audio-visual guidance system is automatically activated. This system provides safe guidance for visually impaired people to cross the road through sound and light signals. The display status of the traffic lights is ensured to be synchronized and coordinated with the signals of the variable audio-visual guidance system to provide a consistent and safe road crossing experience. This embodiment of the application employs a traffic signal control model that includes traffic flow, pedestrian safety, and optimization for the travel of visually impaired people. Initial weights are assigned to each objective. A bone conduction sensor network is set up within the control area to acquire the movement information of users on the tactile paving in real time. Based on the real-time movement information, it detects whether a visually impaired person is using the tactile paving. If a visually impaired person is detected, the weights of each optimization objective are dynamically adjusted. Based on the adjusted weights, the model is run to generate traffic signal control commands. The commands control the traffic lights, and the variable audio-visual system is activated to guide visually impaired people to cross the street safely. These technical means achieve the technical effect of meeting the travel needs of visually impaired people.

[0050] In one possible implementation, step S600 further includes step S610, wherein the variable sound and light guidance system includes a traffic light brightness adjustment module and an audio prompt module. The traffic light brightness adjustment module is used to dynamically adjust the brightness of the traffic lights according to the location of the visually impaired person and traffic flow. The audio prompt module is used to dynamically adjust the intensity and content of the audio prompts according to the location of the visually impaired person and traffic signal status.

[0051] Specifically, a variable sound and light guidance system is a guidance system that can dynamically adjust its sound and light characteristics according to specific conditions (such as the location of visually impaired persons and traffic signal status). The system includes a traffic light brightness adjustment module and an audio prompt module.

[0052] The traffic light brightness adjustment module obtains the location information of visually impaired individuals and traffic flow data from the traffic signal control model. It applies a preset brightness adjustment algorithm that considers factors such as the distance between the visually impaired individual and the traffic light, the density of traffic flow, and ambient lighting conditions. Based on the algorithm's calculations, the brightness of the traffic light is adjusted. For example, if the visually impaired individual is close to the traffic light and the traffic flow is low, the brightness can be appropriately reduced to decrease light pollution; conversely, the brightness needs to be increased to ensure that other pedestrians or drivers around the visually impaired individual can clearly see the traffic light status. Dynamic brightness adjustment is achieved by controlling the traffic light's circuitry or LED driver module.

[0053] The audio prompt module also obtains the location information of visually impaired individuals and the real-time traffic signal status from the traffic signal control model. Based on the traffic signal status (e.g., red, green, yellow light), it generates corresponding audio prompts. For example, it plays a prompt saying "It's a red light now, please wait" when the light is red. Taking into account factors such as the distance between the visually impaired individual and the audio playback device, and the ambient noise level, a preset volume adjustment algorithm is applied to determine the appropriate audio prompt intensity. The audio prompt is then played at the set intensity and content through an audio playback device (e.g., a speaker). This implementation dynamically adjusts the traffic light brightness and audio prompts, enabling visually impaired individuals to perceive traffic signal status more accurately, thereby reducing the risk of traffic accidents.

[0054] In one possible implementation, if a visually impaired person is present, the method further includes step S700, determining a first distributed bone conduction sensor based on the real-time motion information and the nearest distance. Specifically, motion information of the tactile paving user is acquired in real-time from the bone conduction sensor network; this information includes, but is not limited to, position, speed, and direction. For each distributed bone conduction sensor, the straight-line distance between its position coordinates and the tactile paving user's current position is calculated by performing geometric operations between the sensor's position coordinates and the tactile paving user's position coordinates. The distance calculation results of all sensors are compared, and the sensor closest to the tactile paving user is identified as the first distributed bone conduction sensor. A unique identifier (such as an ID number) for the first distributed bone conduction sensor is recorded.

[0055] Step S800: Determine the first nearest neighbor bone conduction sensors of the first distributed bone conduction sensor based on the bone conduction sensor network. Specifically, analyze the topology of the bone conduction sensor network, i.e., the connection relationships between sensors. Based on the network topology, search for directly connected neighbor sensors starting from the first distributed bone conduction sensor using an adjacency matrix or adjacency list in graph theory. Among the neighbor sensors, select the sensor closest to the first distributed bone conduction sensor or with the best signal quality as the first nearest neighbor bone conduction sensor according to preset priority rules (such as distance, signal strength, etc.). Record the unique identifier of the first nearest neighbor bone conduction sensor and its relative positional relationship with the first distributed bone conduction sensor.

[0056] Step S900: Activate the first distributed bone conduction sensor and the first neighboring bone conduction sensor to high-sensitivity mode. Specifically, based on the sensor's hardware characteristics and control protocol, send a mode switching command to the first distributed bone conduction sensor and the first neighboring bone conduction sensor. The command specifies the configuration parameters for the high-sensitivity mode, such as sampling rate, gain, and filter, to ensure that the sensors can more accurately capture and identify the motion information of the user on the tactile paving. Monitor the sensor's state changes to ensure they have successfully switched to high-sensitivity mode and are operating stably. If a sensor fails to switch successfully or malfunctions, trigger a feedback mechanism, such as sending an alarm or restarting the sensor. This implementation selectively activates some sensors to high-sensitivity mode, rather than setting all sensors to high-sensitivity mode, thus reducing system energy consumption and cost while maintaining detection accuracy and response speed.

[0057] The following is a specific example. Three types of sensors are installed on the road: traffic flow sensors (geomagnetic, used to count the number, speed, and type of vehicles), pedestrian sensors (cameras, used to monitor the number of pedestrians, walking direction, and cadence), and visually impaired person sensors (bone conduction, used to identify the position and movement of tactile paving users through bone vibration). All sensors transmit raw data to the cloud via a low-power network, LoRaWAN, at least once per second. For traffic flow data, a low-pass filter is used to remove high-frequency noise (such as lightning interference); for pedestrian data, a median filter is used to eliminate transient outliers (such as those caused by wind); for bone conduction signals, a band-pass filter is used to retain the 0.5-5Hz human motion frequency signal. Outliers (e.g., outliers with vehicle speeds exceeding 100 km / h) are removed using the Z-score method. For tactile paving sensor data, if there is no movement for 30 consecutive seconds, it is marked as "idle". The key features generated are: for traffic flow, the average traffic density (vehicles / minute) over 15 minutes is calculated; for pedestrian safety, the conflict risk index when pedestrians cross the road is calculated (…). Among them, R ped This is a conflict risk index, where v is the average speed of pedestrians crossing the road. safe It is the safe walking speed threshold, and h is the intersection width. safe It is a safe passage width threshold); for visually impaired people, extract cadence and movement direction stability features (step length less than 0.5 meters is judged as slow movement).

[0058] Bone conduction signals were analyzed using a random forest model. Normal walking exhibits a stable gait frequency (1-1.5 steps / second) and smooth acceleration; visually impaired individuals show a low gait frequency (<0.8 steps / second), frequent pauses, and large lateral swaying. The model outputs a "probability of visual impairment" (0-100%). A probability >80% indicates a visually impaired person, triggering priority adjustment; a probability ≤80% is considered normal pedestrians, with no weight adjustment.

[0059] The default base weights are traffic flow (40%), pedestrian safety (40%), and visual impairment optimization (20%). When a visually impaired person is detected, the weights are adjusted based on their distance from the zebra crossing. The adjustment factor is e^(-distance / 20 meters), meaning the closer the person, the higher the weight. The adjustment rules are: Traffic flow weight = original weight × (1 - adjustment factor), Pedestrian safety weight = original weight × (1 - 0.5 × adjustment factor), Visual impairment optimization weight = original weight × (1 + adjustment factor). For example, if a visually impaired person is 10 meters from the zebra crossing, the adjustment factor is e^(-0.5)≈0.6065. In this case, the traffic flow weight is 40%×(1-0.6065)≈40%×0.3935≈15.74%, the pedestrian safety weight is 40%×(1-0.5×0.6065)≈40%×0.7017≈28.07%, and the visual impairment optimization weight is 20%×(1+0.6065)≈20%×1.6065≈32.13%. Since 15.74%+28.07%+32.13%=75.94%, the total weight is not equal to 1 and needs to be normalized. The weights are scaled proportionally to a sum of 1. The normalized results are 15.74%→20.73%, 28.07%→36.98%, and 32.13%→42.31%.

[0060] The CPSO algorithm needs to minimize three metrics simultaneously: average vehicle waiting time, pedestrian crossing risk index, and waiting uncertainty (standard deviation) for visually impaired individuals. Its scoring formula is a weighted sum: Score = ω1 × T veh +ω2×R ped +ω3×σ blind Where ω1, ω2, and ω3 are dynamic weights, and T veh R is the average waiting time for vehicles. ped It is a conflict risk index, σ blind This represents the standard deviation of the waiting time for visually impaired individuals. The algorithm randomly generates a large number of "candidate solutions," each corresponding to a set of traffic light duration combinations. Based on dynamic weights, the solution with the highest comprehensive score is selected, ultimately generating the traffic light phase and duration (e.g., red light 60 seconds → green light 40 seconds → yellow light 5 seconds).

[0061] When a visually impaired person is detected, the brightness of the signal light near the visually impaired person automatically increases by 20%, while the brightness of the other lights decreases to avoid glare; a guiding sound (such as "Green light is on, turn right") is played through a directional speaker, and the volume is dynamically adjusted according to the distance (80 decibels within 1 meter, 50 decibels beyond 5 meters).

[0062] In the above text, refer to Figure 1 A multi-objective optimization AI traffic signal control method according to embodiments of the present invention is described in detail. Next, reference will be made to... Figure 2 A multi-objective optimization AI traffic signal control system according to an embodiment of the present invention is described.

[0063] The multi-objective optimization AI traffic signal control system according to embodiments of the present invention addresses the technical problem of existing technologies failing to meet the travel needs of visually impaired individuals, thereby achieving the technical effect of satisfying their travel needs. The multi-objective optimization AI traffic signal control system includes: a traffic signal control model establishment module 10, a real-time motion information acquisition module 20, a visually impaired person detection module 30, a weight adjustment module 40, a traffic signal control command generation module 50, and a traffic control module 60.

[0064] The traffic signal control model establishment module 10 is used to establish a traffic signal control model, which includes multiple optimization objectives, specifically traffic flow optimization, pedestrian safety optimization, and travel optimization for visually impaired persons. Each optimization objective is configured with an initial weight. The real-time motion information acquisition module 20 is used to set up a bone conduction sensor network within the traffic signal control area and acquire the real-time motion information of users on the tactile paving based on the bone conduction sensor network. The visually impaired person detection module 30 is used to detect the presence of visually impaired persons based on the real-time motion information. The weight adjustment module 40 is used to adjust the initial weights of each optimization objective if visually impaired persons are present, obtaining dynamically adjusted weights. The traffic signal control command generation module 50 is used to run the traffic signal control model based on the dynamically adjusted weights and generate traffic signal control commands. The traffic control module 60 is used to control the display status of traffic lights according to the traffic signal control commands and simultaneously activate a variable sound and light guidance system to guide visually impaired persons across the intersection.

[0065] The specific configuration of the traffic signal control model establishment module 10 will be described in detail below. As mentioned above, the traffic signal control model establishment module 10 may further include: a data acquisition layer building unit for building a data acquisition layer, which collects traffic flow data, pedestrian flow data, and visually impaired person movement data based on sensors; a data processing layer building unit for building a data processing layer, which performs filtering, noise reduction, and feature extraction preprocessing on the collected data; an optimization decision layer building unit for building an optimization decision layer, which dynamically adjusts the phase and duration of traffic lights based on the preprocessed data using a chaotic particle swarm optimization algorithm and a dynamic weight adjustment strategy, and generates traffic signal control commands; an execution layer building unit for building an execution layer, which controls the display state of traffic lights according to the traffic signal control commands; and an integration unit for integrating the data acquisition layer, the data processing layer, the optimization decision layer, and the execution layer to establish the traffic signal control model.

[0066] The specific configuration of the real-time motion information acquisition module 20 will be described in detail below. As mentioned above, a bone conduction sensor network is set up within the traffic signal control area. The real-time motion information acquisition module 20 acquires the real-time motion information of the tactile paving user based on the bone conduction sensor network. The real-time motion information acquisition module 20 may further include: a tactile paving layout information acquisition unit for acquiring the tactile paving layout information within the traffic signal control area and extracting the tactile paving boundary contour from the tactile paving layout information; a sensor deployment information acquisition unit for calculating the tactile paving complexity and tactile paving area based on the tactile paving boundary contour, and matching and acquiring the number and location of sensor deployments based on the tactile paving complexity and tactile paving area; a bone conduction sensor deployment unit for deploying multiple distributed bone conduction sensors on the tactile paving according to the number and location of the sensors, thereby obtaining a bone conduction sensor network; and a real-time motion information acquisition unit for acquiring bone conduction signals from the bone conduction sensor network, filtering, amplifying, and extracting motion features from the bone conduction signals to acquire the real-time motion information of the tactile paving user.

[0067] If a visually impaired person is present, the system may further include: a first distributed bone conduction sensor determination module for determining a first distributed bone conduction sensor based on the real-time motion information and the nearest distance; a first neighboring bone conduction sensor determination module for determining a first neighboring bone conduction sensor of the first distributed bone conduction sensor based on the bone conduction sensor network; and a high sensitivity mode activation module for activating the first distributed bone conduction sensor and the first neighboring bone conduction sensor to a high sensitivity mode.

[0068] The specific configuration of the visually impaired person detection module 30 will be described in detail below. As mentioned above, based on the real-time motion information, the module detects whether a visually impaired person exists. The visually impaired person detection module 30 may further include: an environmental noise filtering unit for filtering environmental noise from the real-time motion information to obtain noise-reduced motion information; a visually impaired feature extraction unit for extracting visually impaired features from the noise-reduced motion information to obtain a visually impaired motion pattern signal; and a comparison unit for comparing the visually impaired motion pattern signal with a confidence threshold to detect whether a visually impaired person exists.

[0069] The specific configuration of the weight adjustment module 40 will be described in detail below. As mentioned above, if visually impaired individuals exist, the initial weights of each optimization objective are adjusted to obtain dynamically adjusted weights. The weight adjustment module 40 may further include: a visually impaired individual information acquisition unit, used to acquire the number and location information of visually impaired individuals based on the real-time motion information if visually impaired individuals exist; a visually impaired individual motion dot matrix generation unit, used to generate a visually impaired individual motion dot matrix based on the number and location information of visually impaired individuals; a visually impaired individual travel weight adjustment coefficient calculation unit, used to calculate the visually impaired individual travel weight adjustment coefficient based on the relative positional relationship between the visually impaired individual motion dot matrix and the zebra crossing at the intersection; and an initial weight adjustment unit, used to adjust the initial weights of each optimization objective based on the visually impaired individual travel weight adjustment coefficient to obtain dynamically adjusted weights.

[0070] The specific configuration of the traffic control module 60 will be described in detail below. As mentioned above, the traffic control module 60 may further include: a variable sound and light guidance system building unit for constructing a variable sound and light guidance system, wherein the variable sound and light guidance system includes a traffic light brightness adjustment module and an audio prompt module. The traffic light brightness adjustment module is used to dynamically adjust the brightness of the traffic lights according to the location of the visually impaired person and traffic flow, and the audio prompt module is used to dynamically adjust the intensity and content of the audio prompts according to the location of the visually impaired person and traffic signal status.

[0071] The multi-objective optimization AI traffic signal control system provided in the embodiments of the present invention can execute the multi-objective optimization AI traffic signal control method provided in any embodiment of the present invention, and has the corresponding functional modules and beneficial effects of the execution method.

[0072] Although this application makes various references to certain modules in the system according to the embodiments of this application, any number of different modules can be used and run on user terminals and / or servers. The various units and modules included are only divided according to functional logic, but are not limited to the above division, as long as the corresponding functions can be achieved; in addition, the specific names of each functional unit are only for easy distinction between each other and are not used to limit the scope of protection of this invention.

[0073] The specific embodiments described above do not constitute a limitation on the scope of protection of this application. Those skilled in the art should understand that various modifications, combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this application should be included within the scope of protection of this application. In some cases, the actions or steps described in this application can be performed in a different order than that shown in the embodiments and still achieve the desired results. Furthermore, the processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

Claims

1. A multi-objective optimization AI traffic signal control method, characterized by, The method comprises: establishing a traffic signal control model, the model comprising multiple optimization objectives, specifically traffic flow optimization, pedestrian safety optimization, and visually impaired person travel optimization, each optimization objective being configured with an initial weight; the establishment of the traffic signal control model comprising: building a data acquisition layer to collect traffic flow data, pedestrian flow data, and visually impaired person movement data based on sensors; building a data processing layer to filter, denoise, and feature extract the collected data for preprocessing; building an optimization decision layer to dynamically adjust the phase and duration of the signal light based on the preprocessed data using a chaotic particle swarm optimization algorithm and a dynamic weight adjustment strategy, and to generate traffic signal control instructions; building an execution layer to control the display state of the traffic signal light according to the traffic signal control instructions; integrating the data acquisition layer, the data processing layer, the optimization decision layer, and the execution layer to establish a traffic signal control model; setting up a bone conduction sensor network in the traffic signal control area and obtaining real-time movement information of blind lane users based on the bone conduction sensor network; detecting whether there are visually impaired persons based on the real-time movement information; if there are visually impaired persons, adjusting the initial weights of the optimization objectives to obtain dynamic adjustment weights, comprising: if there are visually impaired persons, obtaining the number and position information of the visually impaired persons based on the real-time movement information; generating a visually impaired person movement dot matrix based on the number and position information of the visually impaired persons; calculating a visually impaired person travel weight adjustment coefficient based on the relative position relationship between the visually impaired person movement dot matrix and the zebra crossing; adjusting the initial weights of the optimization objectives based on the visually impaired person travel weight adjustment coefficient to obtain dynamic adjustment weights; running the traffic signal control model based on the dynamic adjustment weights to generate traffic signal control instructions; controlling the display state of the traffic signal light based on the traffic signal control instructions, and simultaneously starting a variable sound and light guidance system to guide visually impaired persons through the intersection; When a visually impaired person is detected, the weight is adjusted according to the distance from the zebra crossing, the travel weight adjustment coefficient = e (- the distance from the zebra crossing of the disabled person / 20 meters), according to the travel weight adjustment coefficient of the visually impaired person, the initial weight of each optimization target is adjusted by applying the weight adjustment algorithm, the adjustment rule is traffic flow weight = traffic flow initial weight × (1- travel weight adjustment coefficient), pedestrian safety weight = pedestrian safety initial weight × (1- 0.5 × travel weight adjustment coefficient), visually impaired optimization weight = visually impaired optimization initial weight × (1+ travel weight adjustment coefficient), and then the dynamic adjustment weight is obtained by normalization; the minimum of the three indexes in multi-objective optimization is respectively vehicle average waiting time , pedestrian crossing risk index and the standard deviation of the waiting time of the visually impaired , wherein: the conflict risk index is the conflict risk index, v is the average speed of the pedestrian crossing the road, is the safe walking speed threshold, h is the width of the intersection, is the safe passing width threshold.

2. The multi-objective optimization Al traffic signal control method of claim 1, wherein, the bone conduction sensor network in the traffic signal control area, and obtaining real-time movement information of blind lane users based on the bone conduction sensor network, comprising: obtaining blind lane layout information in the traffic signal control area and extracting blind lane boundary contours from the blind lane layout information; calculating blind lane complexity and blind lane area based on the blind lane boundary contours, and matching the sensor layout quantity and position based on the blind lane complexity and blind lane area; laying multiple distributed bone conduction sensors in the blind lane based on the sensor layout quantity and position to obtain a bone conduction sensor network; obtaining bone conduction signals from the bone conduction sensor network, filtering, amplifying, and extracting movement features of the bone conduction signals to obtain real-time movement information of blind lane users.

3. The multi-objective optimization Al traffic signal control method of claim 2, wherein, if there are visually impaired persons, further comprising: determining a first distributed bone conduction sensor based on the nearest distance based on the real-time movement information; determining a first neighbor bone conduction sensor of the first distributed bone conduction sensor based on the bone conduction sensor network; activating the first distributed bone conduction sensor and the first neighbor bone conduction sensor to a high sensitivity mode.

4. The multi-objective optimization Al traffic signal control method of claim 1, wherein, The method comprises the following steps: The real-time motion information is subjected to environmental noise filtering to obtain clean noise motion information; The clean noise motion information is subjected to visual impairment feature extraction to obtain a visual impairment motion mode signal; The visual impairment motion mode signal is compared with a confidence threshold to detect whether a visually impaired person exists.

5. The multi-objective optimization Al traffic signal control method of claim 1, wherein, The variable sound and light guiding system comprises a signal lamp brightness adjusting module and an audio prompting module, the signal lamp brightness adjusting module is used for dynamically adjusting the brightness of the signal lamp according to the position of the visually impaired person and the traffic flow, and the audio prompting module is used for dynamically adjusting the strength and content of the audio prompt according to the position of the visually impaired person and the traffic signal state.

6. A multi-objective optimization AI traffic signal control system characterized by, The system is used to implement the multi-objective optimization AI traffic signal control method according to any one of claims 1-5, and the system comprises: A traffic signal control model establishing module is used to establish a traffic signal control model, the model comprises a plurality of optimization objectives, specifically traffic flow optimization, pedestrian safety optimization and visually impaired person travel optimization, and each optimization objective is configured with an initial weight, The traffic signal control model establishing module is also used to build a data acquisition layer, collect traffic flow data, pedestrian flow data and visually impaired person motion data based on sensors; A data processing layer is built to filter, denoise and feature extract the collected data for preprocessing; An optimization decision layer is built to dynamically adjust the phase and duration of the signal lamp based on the preprocessed data, generate traffic signal control instructions by using a chaotic particle swarm optimization algorithm and a dynamic weight adjusting strategy; An execution layer is built to control the display state of the traffic signal lamp according to the traffic signal control instructions; The data acquisition layer, the data processing layer, the optimization decision layer and the execution layer are integrated to establish a traffic signal control model; A real-time motion information acquisition module is used to set a bone conduction sensor network in a traffic signal control area and acquire real-time motion information of blind lane users based on the bone conduction sensor network; A visually impaired person detection module is used to detect whether a visually impaired person exists based on the real-time motion information; A weight adjusting module is used to adjust the initial weight of each optimization objective to obtain a dynamic adjustment weight if a visually impaired person exists; The weight adjusting module is also used to acquire the number and position information of the visually impaired person based on the real-time motion information if a visually impaired person exists; A visually impaired person motion dot matrix is generated based on the number and position information of the visually impaired person; A visually impaired person travel weight adjustment coefficient is calculated based on the relative position relationship between the visually impaired person motion dot matrix and the zebra crossing of the intersection; The initial weight of each optimization objective is adjusted to obtain a dynamic adjustment weight based on the visually impaired person travel weight adjustment coefficient; A traffic signal control instruction generation module is used to run the traffic signal control model based on the dynamic adjustment weight to generate traffic signal control instructions; A traffic control module is used to control the display state of the traffic signal lamp based on the traffic signal control instructions, and simultaneously start a variable sound and light guiding system to guide the visually impaired person to pass through the intersection. The traffic control module is also used to adjust the weight according to the distance of the visually impaired person from the zebra crossing when the visually impaired person is detected, the travel weight adjustment coefficient is e(-distance of the disabled person to the zebra crossing / 20 meters), the initial weight of each optimization target is adjusted by applying a weight adjustment algorithm according to the travel weight adjustment coefficient of the visually impaired person, the adjustment rule is traffic flow weight=traffic flow initial weight×(1-travel weight adjustment coefficient), pedestrian safety weight=pedestrian safety initial weight×(1-0.5×travel weight adjustment coefficient), visually impaired optimization weight=visually impaired optimization initial weight×(1+travel weight adjustment coefficient), and then the dynamic adjustment weight is obtained by normalization; the minimum of the three indexes in the multi-objective optimization is respectively vehicle average waiting time , pedestrian crossing risk index and standard deviation of the waiting time of the visually impaired , wherein: the conflict risk index is the conflict risk index, v is the average speed of the pedestrian crossing the road, is the safe walking speed threshold, h is the width of the intersection, is the safe passing width threshold.

Images