Dynamic Network Redundancy Configuration and Switching Method for Intelligent Driving Vehicles in Complex Information Source Scenarios

Abstract

The present invention discloses a method for dynamic network redundancy configuration and switching of intelligent driving vehicles in complex signal source scenarios, which relates to the fields of intelligent transportation and communication technologies. The method of the present invention detects the communication quality data of multiple communication links at different positions during the driving of driverless taxis, broadcasts the data to other surrounding driverless taxis through the vehicle-mounted communication system, and receives the communication quality data broadcast by other driverless taxis; without relying on a centralized server or a cloud platform, it decentralizes and fuses to generate a signal map only based on its own detected and received data; according to the signal map and the preset task priorities, it dynamically configures the network redundancy strategy; after executing the strategy, it records the score and broadcasts it, receives the scores of other vehicles, and optimizes and adjusts the strategy in advance. This method solves the problem of unstable communication in complex signal source scenarios and improves the communication reliability and efficiency of intelligent driving vehicles.

Description

Technical Field

[0001] The present invention relates to the fields of intelligent transportation and communication technologies, and more specifically, to a method for dynamically configuring and switching network redundancy for intelligent driving vehicles in complex signal source scenarios. Background Art

[0002] Intelligent driving vehicles are becoming increasingly important in the modern transportation field, especially playing a core role in urban transportation and autonomous taxi services. Such technologies rely on efficient communication systems to achieve vehicle perception, decision-making, and control. However, in complex signal source scenarios such as urban canyons, tunnels, or viaducts, the communication quality often becomes unstable due to signal interference or occlusion. This instability may cause delays or interruptions during the execution of critical tasks by the vehicle, affecting safety and operating efficiency. Therefore, how to ensure the stability and reliability of communication in complex environments has become a difficult problem that needs to be solved urgently in the current technological development. Summary of the Invention

[0003] The technical problem to be solved by the present invention is to provide a method for dynamically configuring and switching network redundancy for intelligent driving vehicles in complex signal source scenarios to solve the problems mentioned in the background art.

[0004] To achieve the above object, the present invention adopts the following technical solutions:

[0005] A method for dynamically configuring and switching network redundancy for intelligent driving vehicles in complex signal source scenarios, comprising the following steps:

[0006] Detect the communication quality data of multiple communication links at different positions during the driving process of the autonomous taxi;

[0007] Broadcast the communication quality data to other surrounding autonomous taxis through the vehicle-mounted communication system, and receive the communication quality data broadcast by other autonomous taxis;

[0008] Without relying on a centralized server or cloud platform, only based on the communication quality data detected by the autonomous taxi itself and the communication quality data received from other autonomous taxis, decentralized fusion is performed to generate a signal map of the autonomous taxi;

[0009] According to the signal map and the preset task priorities, dynamically configure the network redundancy strategy of the autonomous taxi;

[0010] Execute the network redundancy strategy and record the score of the strategy execution effect;

[0011] Broadcast the score to other surrounding autonomous taxis through the vehicle-mounted communication system, and receive the scores broadcast by other autonomous taxis;

[0012] Optimize and pre-adjust the network redundancy strategy of the driverless taxi based on the received scores.

[0013] In some embodiments, detecting the communication quality data of multiple communication links at different positions during the driving of the driverless taxi specifically includes:

[0014] Using the high-precision positioning system on the driverless taxi to record the position information of the driverless taxi during driving;

[0015] Through the communication module on the driverless taxi, real-time detect and record the signal strength, delay, and packet loss rate of the multiple communication links.

[0016] In some embodiments, based on the communication quality data detected by the driverless taxi itself and the communication quality data received from other driverless taxis, fusing to generate the signal map of the driverless taxi specifically includes:

[0017] Associate the communication quality data with the corresponding position information to form a position-communication quality data set;

[0018] Use a weighted average algorithm to process the position-communication quality data set to generate the signal map, where the weights are based on the freshness of the data and the distance between the driverless taxis.

[0019] In some embodiments, according to the signal map and the preset task priorities, dynamically configure the network redundancy strategy of the driverless taxi, specifically including:

[0020] According to the task priorities, divide the tasks of the driverless taxi into high-priority tasks, medium-priority tasks, and low-priority tasks;

[0021] For the high-priority tasks, configure a redundancy strategy of parallel transmission of multiple communication links;

[0022] For the medium-priority tasks and the low-priority tasks, select the single communication link with the best communication quality according to the signal map.

[0023] In some embodiments, the high-priority tasks include safety-related tasks, the medium-priority tasks include navigation and path planning tasks, and the low-priority tasks include passenger entertainment data transmission tasks.

[0024] In some embodiments, execute the network redundancy strategy and record the score of the strategy execution effect, specifically including: monitoring the task completion rate, average delay, and energy consumption during the execution of the network redundancy strategy;

[0025] Calculate the score using the following formula:

[0026] ;

[0027] Among them, S is the score, C is the task completion rate, D is the average latency, and E is the energy consumption. , , are preset weight coefficients.

[0028] In some embodiments, based on the received score, the network redundancy strategy of the driverless taxi is optimized and adjusted in advance, specifically including:

[0029] Collect the score data of the driverless taxi itself and other driverless taxis;

[0030] Use machine learning algorithms to analyze the score data and predict the execution effects of different network redundancy strategies at future locations;

[0031] According to the prediction results, adjust the network redundancy strategy of the driverless taxi in advance.

[0032] In some embodiments, the machine learning algorithm uses a reinforcement learning algorithm.

[0033] In some embodiments, the reinforcement learning algorithm is a deep Q-network.

[0034] In some embodiments, the training process of the deep Q-network includes:

[0035] Collect training data from the shared scores of multiple vehicles;

[0036] Construct a data set including location, policy configuration, task type, and score;

[0037] Use the Adam optimizer and the mean squared error loss function to train the deep Q-network;

[0038] Optimize the model parameters through a reward function based on the score.

[0039] The advantages of the present invention over the prior art are as follows. The decentralized signal map construction method adopted by the present invention enables each driverless taxi to broadcast multiple communication link quality data detected by itself to surrounding vehicles through the on-vehicle communication system during driving and receive communication quality data from other vehicles. Without relying on any centralized server or cloud platform, a real-time high-precision signal map is generated based on the data fusion of the vehicle itself and surrounding vehicles, completely eliminating the single-point failure risk brought by the centralized architecture. Even if the operator's network is interrupted or the background server is unavailable, vehicles can still maintain an accurate perception and dynamic update of the communication environment through mutual broadcasting and receiving, so as to ensure that the network redundancy strategy can take effect continuously in any environment. At the same time, the present invention also broadcasts the score of the execution effect of each network redundancy strategy (including task completion rate, average delay, and energy consumption) through the on-vehicle communication system, enabling surrounding vehicles to obtain the actual performance at different positions and with different strategy configurations in real time. The score information can be synchronized without an additional background database or dedicated communication link, which not only reduces the system deployment and maintenance costs but also avoids the delay or loss of score data caused by high server load or network quality fluctuations. Based on these real-time score data, a reinforcement learning algorithm is used to predict and optimize the future network redundancy configuration, which not only improves the timeliness and accuracy of decision-making but also enables the vehicle to adjust the network strategy in advance according to historical scores and real-time environmental changes, significantly enhancing the communication reliability and resource utilization efficiency. Thus, a dynamic network redundancy configuration and switching solution with high availability, low cost, easy expansion, and strong robustness is provided for driverless taxis in complex signal source scenarios. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Figure 1 is the overall flowchart of the present invention;

[0041] Figure 2 is the flowchart of communication quality detection and sharing of the present invention;

[0042] Figure 3 is the flowchart of signal map generation of the present invention;

[0043] Figure 4 is the flowchart of redundancy strategy configuration and optimization of the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0044] The following describes the specific embodiments of the present invention with reference to the accompanying drawings.

[0045] In complex source scenarios, the communication quality of intelligent driving vehicles (hereinafter referred to as "driverless taxis") is affected by various external factors, such as geographical location, building occlusion, weather changes, and network load. To ensure that driverless taxis can transmit data stably and efficiently during driving, especially when performing critical tasks, a flexible network redundancy configuration and switching method must be designed. As shown in Figure 1, the present invention realizes the intelligent selection and management of communication methods in different scenarios by real-time detecting communication quality, sharing data, generating signal maps, dynamically configuring redundancy strategies, and combining a scoring mechanism and prediction optimization. The following is a detailed description and implementation plan of this method.

[0046] More specifically, during the driving process of a driverless taxi, it is necessary to monitor the communication quality of multiple communication links in real time to provide data support for subsequent signal map generation and strategy configuration. This process mainly relies on the in-vehicle high-precision positioning system and communication module, as shown in Figure 2.

[0047] The high-precision positioning system, such as the Global Positioning System (GPS) or Global Navigation Satellite System (GNSS), can accurately record the position information of the driverless taxi. This information is usually stored in the form of latitude and longitude coordinates. For example, a vehicle is located at 39.9042°N, 116.4074°E at a certain moment.

[0048] At the same time, the communication module is responsible for detecting the quality data of multiple communication links, including signal strength, latency, and packet loss rate. A driverless taxi may support multiple communication methods simultaneously, such as 4G, 5G, and Wi-Fi, or even satellite communication. Each method corresponds to an independent communication link. The signal strength is measured in decibel-milliwatts (dBm), usually ranging from -120 dBm (extremely weak signal) to -50 dBm (strong signal); the latency is measured in milliseconds (ms), representing the time interval from packet transmission to reception; the packet loss rate is expressed as a percentage, reflecting the proportion of lost packets during data transmission. For example, at a certain location, the vehicle detects that the signal strength of the 5G link is -70 dBm, the latency is 50 ms, and the packet loss rate is 1%, while the signal strength of the 4G link is -85 dBm, the latency is 120 ms, and the packet loss rate is 3%. These data will be recorded and associated with the corresponding position information to form a preliminary data set.

[0049] To ensure the comprehensiveness of the data, the communication module needs to detect at a certain frequency (e.g., once per second) and continuously update the data during the vehicle's driving. This real-time nature can reflect the dynamic changes in communication quality. For example, when the vehicle enters a tunnel or a high-rise building dense area, the signal strength may drop significantly.

[0050] As shown in Figure 3, after detecting communication quality data, the driverless taxi needs to share it with other surrounding vehicles and at the same time receive the data shared by other vehicles to achieve collaborative communication quality perception. This process is completed through an in-vehicle communication system, and typical technologies include vehicle-to-everything (V2X) communication, where vehicle-to-vehicle (V2V) and vehicle-to-infrastructure (V2I) are the main modes.

[0051] Specifically, each driverless taxi will package the detected communication quality data (including location coordinates, link type, signal strength, latency, and packet loss rate) into data packets and broadcast them to surrounding vehicles through V2V communication. The broadcast range usually depends on the coverage ability of the communication system. For example, in an urban environment, the effective distance of V2V communication may be 300 to 500 meters. To avoid data conflicts, protocols such as time-division multiple access (TDMA) or carrier sense multiple access (CSMA) can be adopted to ensure the orderly transmission of data packets.

[0052] At the same time, the driverless taxi will also receive the communication quality data broadcast by other vehicles. These data will be stored in the local database and used for subsequent signal map generation. For example, assume that a vehicle detects the signal strength of a 5G link at location A as -65 dBm and broadcasts it. Another surrounding vehicle can combine this data with the data it detects at location B (such as the 5G signal strength of -75 dBm) after receiving the data. This collaborative mechanism significantly expands the data collection range of a single vehicle, enabling the signal map to cover a wider area.

[0053] Without relying on a centralized server or cloud platform, a signal map reflecting the communication quality at different locations is generated in a decentralized manner based only on the communication quality data detected by the driverless taxi itself and the communication quality data received from other driverless taxis. This process is divided into two steps: data association and weighted averaging.

[0054] First, associate the communication quality data with the corresponding location information to form a location-communication quality data set. Each data point contains location coordinates (latitude and longitude) and the communication quality metrics (signal strength, latency, packet loss rate) at that location. For example, a data point may be "Location: 39.9042°N, 116.4074°E, 5G link: signal strength -70 dBm, latency 50 ms, packet loss rate 1%".

[0055] Then, use a weighted averaging algorithm to process the data set to generate the signal map. The design of the weights is based on two key factors: the freshness of the data and the distance between vehicles.

[0056] Data freshness reflects the timeliness of data. Newly collected data usually better represents the current communication status, so it has a higher weight. The freshness weight can be calculated using an exponential decay function. For example, , where t is the current time, is the data recording time, and λ is the decay coefficient (e.g., 0.1). If a piece of data was recorded 5 seconds ago and λ = 0.1, then .

[0057] The distance weight reflects the correlation between the data point and the target location. Data points closer to the target location have a greater impact on the communication quality at the target location. The distance weight can be calculated using a Gaussian function. For example, , where d is the distance between the two vehicles, and σ is the standard deviation (e.g., set to 50 meters). If the two vehicles are 30 meters apart, then .

[0058] The combined weight is .

[0059] When generating the signal map, for each location point on the map (e.g., grid points divided at 10 - meter intervals), collect the data points near that location (e.g., within a radius of 50 meters), and perform a weighted average of the communication quality metrics according to the combined weight. For example, assume there is data from three vehicles near a certain location:

[0060] Vehicle 1: Signal strength - 70 dBm, = 0.8;

[0061] Vehicle 2: Signal strength - 75 dBm, = 0.6;

[0062] Vehicle 3: Signal strength - 65 dBm, = 0.05;

[0063] Then the combined signal strength at this location is:

[0064] ;

[0065] The delay and packet loss rate are also calculated in a similar manner, and finally a signal map containing the communication quality at each location is formed.

[0066] As shown in Figure 4, according to the generated signal map and the priority of the task, the driverless taxi needs to dynamically configure the network redundancy strategy to balance communication reliability and resource utilization efficiency. The division of task priorities and strategy configuration are the core of this process.

[0067] Tasks are divided into three categories according to their importance and real - time requirements:

[0068] High-priority tasks mainly involve safety-related functions such as automatic emergency braking, collision warning, and obstacle detection. These tasks require extremely high reliability and extremely low latency of the communication link.

[0069] Medium-priority tasks include navigation and path planning, such as real-time traffic information update and optimal path calculation. These tasks have certain requirements for communication quality but allow a certain degree of fault tolerance.

[0070] Low-priority tasks, such as passenger entertainment data transmission (such as video streaming or music downloading), have lower requirements for communication quality and the lowest priority.

[0071] For tasks with different priorities, different redundancy strategies are configured:

[0072] For high-priority tasks, a redundancy strategy of parallel transmission using multiple communication links is adopted. For example, both 5G and 4G links are used to transmit the same data packet simultaneously. Even if one link is interrupted due to signal attenuation or interference, the other link can still ensure data transmission. This method significantly improves the reliability of communication but increases energy consumption and bandwidth occupancy.

[0073] For medium-priority and low-priority tasks, a single link with the best communication quality is selected according to the signal map. For example, if the signal map shows that the signal strength of the 5G link at the current location is -65dBm and the latency is 40ms, while the 4G link is -85dBm and the latency is 100ms, then the 5G link is selected to transmit data. This method reduces resource waste while meeting the task requirements.

[0074] During actual driving, the vehicle may need to switch strategies frequently. For example, when the vehicle enters an area with weak signals (such as a tunnel), high-priority tasks may switch from a dual-link to a triple-link (such as adding Wi-Fi) to further improve reliability.

[0075] After configuring the redundancy strategy, the driverless taxi will execute the strategy and evaluate its effect by monitoring key metrics, and finally calculate a score. These metrics include task completion rate, average latency, and energy consumption.

[0076] The task completion rate (C) represents the proportion of tasks successfully completed, expressed as a percentage. For example, if 100 data packets are sent and 95 of them are successfully transmitted, the completion rate is 95%.

[0077] The average latency (D) is the average time of data transmission, measured in milliseconds. For example, if the latencies of 10 data packets are 50ms, 60ms, etc., the average latency is their average value.

[0078] The energy consumption (E) reflects the communication power consumption during the execution of the strategy, which is measured in watts (W) and can be measured by the power consumption sensor of the communication module.

[0079] The formula for calculating the score is:

[0080] ;

[0081] where S is the score, , , are the weight coefficients. The design of this formula comprehensively considers the impacts of task completion rate, latency, and energy consumption:

[0082] • C encourages a high task completion rate. The higher the completion rate, the higher the score;

[0083] - • D penalizes high latency. The lower the latency, the higher the score;

[0084] - • E penalizes high energy consumption. The lower the energy consumption, the higher the score.

[0085] The value range of the weight coefficients is usually from 0 to 1, and + + = 1 to keep the score standardized. For example, for high-priority tasks, = 0.6, = 0.2, <� = 0.2 can be set to highlight the importance of the completion rate; for low-priority tasks, it can be adjusted to = 0.3, = 0.3, = 0.4 to pay more attention to energy consumption. Assume that in a certain execution, the completion rate is 95%, the latency is 100 ms, the energy consumption is 10 W, and the weights are 0.6, 0.2, 0.2, then the score <� .

[0086] After executing the strategy and calculating the score, the driverless taxi will broadcast the score to surrounding vehicles through the vehicle-mounted communication system and receive the scores of other vehicles. This process is similar to the broadcast of communication quality data, aiming to share the information of the strategy effect.

[0087] The score data packet contains the strategy configuration information (such as the link type used), task type, and score value. For example, "Location A, using 5G + 4G redundant transmission, high-priority task, score 35". These data are broadcast through V2V communication, and the received scores are stored locally for subsequent optimization.

[0088] By collecting the scoring data of itself and other vehicles, the driverless taxi can further utilize machine learning algorithms to analyze this data, predict the effects of different strategies at future locations, and adjust the strategies in advance.

[0089] First, construct a dataset that includes location, strategy configuration, task type, and score. For example, "Location A, 5G single link, medium-priority task, score 40; Location B, 5G + 4G redundancy, high-priority task, score 35". Then, use a machine learning model for analysis.

[0090] For model selection, a Deep Q-Network (DQN) can be adopted. This is a reinforcement learning algorithm suitable for dynamic strategy selection. The inputs of the model include location information, communication quality data, and task type, and the output is the expected scores of different strategies.

[0091] The training process is based on historical score data, and the model parameters are optimized through a reward function (using the score S as the reward). For example, DQN contains a neural network composed of multiple fully connected layers (such as 64 neurons in the input layer, 128 neurons in the hidden layer, and the output layer corresponding to the number of strategies), and is trained using the Adam optimizer and the mean squared error loss function. The training data can be collected from the shared scores of multiple vehicles to ensure the generalization ability of the model.

[0092] During driving, the vehicle uses the trained model to predict the strategy scores at the next location based on the current location and future path. For example, if the prediction shows that the score of using 5G + 4G redundancy at Location C (38) is higher than that of 5G single link (32), the vehicle switches to the redundant strategy in advance when approaching Location C. This prediction and adjustment ability significantly improves the adaptability of communication.

[0093] Through the above steps, the driverless taxi can achieve dynamic network redundancy configuration and switching in complex signal source scenarios, ensuring the efficiency and reliability of communication. Whether it is the high real-time requirement of safety tasks or the resource-saving demand of entertainment tasks, this method can provide comprehensive communication support through data sharing, intelligent configuration, and optimized adjustment.

[0094] The above is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention, according to the technical solution and inventive concept of the present invention, makes equivalent substitutions or changes, and should be covered by the protection scope of the present invention.

Claims

1. A method for dynamic network redundancy configuration and switching of intelligent driving vehicles in a complex signal source scenario, characterized in that The steps include: Detect the communication quality data of multiple communication links at different positions during the driving of the driverless taxi; Broadcast the communication quality data to other surrounding driverless taxis through the vehicle-mounted communication system, and receive the communication quality data broadcast by other driverless taxis; Without relying on a centralized server or cloud platform, generate the signal map of the driverless taxi in a decentralized manner only based on the communication quality data detected by the driverless taxi itself and the communication quality data received from other driverless taxis; Dynamically configure the network redundancy strategy of the driverless taxi according to the signal map and the preset task priorities; Execute the network redundancy strategy and record the score of the strategy execution effect; Broadcast the score to other surrounding driverless taxis through the vehicle-mounted communication system, and receive the scores broadcast by other driverless taxis; Optimize and pre-adjust the network redundancy strategy of the driverless taxi based on the received scores.

2. The method for dynamic network redundancy configuration and switching of an intelligent driving vehicle in a complex signal source scenario according to claim 1, wherein Detecting the communication quality data of multiple communication links at different positions during the driving of the driverless taxi specifically includes: Using the high-precision positioning system on the driverless taxi to record the position information of the driverless taxi during driving; Through the communication module on the driverless taxi, detect and record the signal strength, delay, and packet loss rate of the multiple communication links in real time.

3. The dynamic network redundancy configuration and switching method for intelligent driving vehicles in complex signal source scenarios according to claim 1, wherein Fusing to generate the signal map of the driverless taxi based on the communication quality data detected by the driverless taxi itself and the communication quality data received from other driverless taxis specifically includes: Associate the communication quality data with the corresponding position information to form a position-communication quality data set; Use a weighted average algorithm to process the position-communication quality data set to generate the signal map, where the weights are based on the freshness of the data and the distance between the driverless taxis.

4. The dynamic network redundancy configuration and switching method for intelligent driving vehicles in complex signal source scenarios according to claim 1, characterized in that Dynamically configure the network redundancy strategy of the driverless taxi according to the signal map and the preset task priorities specifically includes: According to the task priorities, divide the tasks of the driverless taxi into high-priority tasks, medium-priority tasks, and low-priority tasks; For the high-priority tasks, configure a redundancy strategy for parallel transmission of multiple communication links; For the medium-priority tasks and the low-priority tasks, select the single communication link with the best communication quality according to the signal map.

5. The method for dynamic network redundancy configuration and switching of an intelligent driving vehicle in a complex signal source scenario according to claim 4, wherein, The high-priority tasks include safety-related tasks, the medium-priority tasks include navigation and path planning tasks, and the low-priority tasks include passenger entertainment data transmission tasks.

6. The dynamic network redundancy configuration and switching method for intelligent driving vehicles in complex information source scenarios according to claim 1, characterized in that Execute the network redundancy strategy and record the score of the strategy execution effect specifically includes: monitoring the task completion rate, average delay, and energy consumption during the execution of the network redundancy strategy; Calculate the score using the following formula: ； Among them, S is the score, C is the task completion rate, D is the average latency, E is the energy consumption, , , are preset weight coefficients.

7. The method for dynamic network redundancy configuration and switching of an intelligent driving vehicle in a complex signal source scenario according to claim 1 or 6, characterized in that, Optimize and pre-adjust the network redundancy strategy of the driverless taxi based on the received scores specifically includes: Collect the score data of the driverless taxi itself and other driverless taxis; Use machine learning algorithms to analyze the score data and predict the execution effects of different network redundancy strategies at future positions; According to the prediction results, pre-adjust the network redundancy strategy of the driverless taxi.

8. The method for dynamic network redundancy configuration and switching of an intelligent driving vehicle in a complex signal source scenario according to claim 7, wherein, The machine learning algorithm uses a reinforcement learning algorithm.

9. The dynamic network redundancy configuration and switching method for intelligent driving vehicles in complex signal source scenarios according to claim 8, characterized in that, The reinforcement learning algorithm is a deep Q-network.

10. The method for dynamic network redundancy configuration and switching of an intelligent driving vehicle in a complex signal source scenario according to claim 9, wherein, The training process of the deep Q-network includes: Collect training data from the shared ratings of multiple vehicles; Construct a data set including location, policy configuration, task type, and rating; Use the Adam optimizer and the mean squared error loss function to train the deep Q-network; Optimize the model parameters through a reward function based on ratings.

Images