Intelligent driving vehicle driving control method, system, equipment and medium

Abstract

The invention discloses an intelligent driving vehicle driving control method, system and device and a medium. The intelligent driving vehicle driving control method comprises the steps that vehicle driving information and network quality information of a base station near a vehicle position are collected in real time; according to the vehicle driving information, the network quality information and a historical driving state of the vehicle in a historical similar driving path, utilizing a pre-trained time sequence prediction model to predict a vehicle network connection fault in a future preset time to obtain a vehicle network connection fault probability; determining a driving risk level of the vehicle according to the vehicle network connection fault and the probability thereof and a driving task of the vehicle; and according to the vehicle network connection fault, the driving task of the vehicle and the driving risk level, utilizing a pre-trained machine learning model to select and generate an optimal driving strategy including network resource scheduling and vehicle behavior guidance, and according to a priority sequence of driving services of the vehicle, controlling the corresponding vehicle to execute the optimal driving strategy.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of intelligent driving, and more particularly to an intelligent driving vehicle driving control method, system, device and medium. BACKGROUND

[0002] New energy and intelligent driving are two major trends in the current automobile industry. As one of the key directions of high-tech development, major vehicle manufacturers have begun to layout the intelligent driving field. At present, L4 level intelligent driving systems are gradually popularized in passenger vehicles on the market. The vehicle can complete all driving tasks within a specific design operation domain (ODD) without human driver intervention.

[0003] L4 level automatic driving relies on continuous and reliable data exchange between the vehicle and the cloud platform, roadside facilities and other vehicles. However, the current cellular network (such as 4G / 5G) management mainly adopts a passive or responsive fault handling mechanism. When the vehicle is interrupted, the data transmission rate is sharply reduced, or the connection is interrupted, the existing vehicle driving control method usually triggers an alarm according to the preset network signal static threshold (such as signal strength lower than -110dBm). After the alarm is generated, a predefined and fixed recovery process is usually performed, for example, trying to reconnect the vehicle to the nearest base station with the strongest signal, or involving the operation and maintenance personnel for analysis. However, since the above driving control method is only for recovery response after the fault occurs, the processing process has a delay. For L4 level automatic driving, a communication interruption of hundreds of milliseconds may cause decision errors when performing critical operations (such as lane changing on the highway, unprotected left turn at the intersection), which may cause serious safety accidents. Moreover, the existing vehicle driving control method adopts a one-size-fits-all recovery strategy for all vehicles that have triggered an alarm, without considering the current driving task and business priority of the vehicle. For example, for a vehicle that is performing high-precision positioning data difference calculation, and a vehicle that is only uploading logs, the existing control uses the same network recovery strategy, which leads to the problem of failing to prioritize critical businesses, and cannot meet the extreme reliability and continuity requirements of network connection for L4 level automatic driving, thereby posing a safety hazard. SUMMARY

[0004] The present application provides an intelligent driving vehicle driving control method, system, device and medium to solve the technical problem that the existing vehicle driving control has a response lag and recovery non-adaptability, which cannot meet the extreme reliability and continuity requirements of network connection for L4 level automatic driving, thereby causing a safety hazard.

[0005] In a first aspect, an intelligent driving vehicle driving control method is provided, comprising:

[0006] real-time collection of vehicle driving information and network quality information of base stations near the vehicle position;

[0007] According to the vehicle driving information, the network quality information, and the historical driving state of the vehicle on a historically similar driving path, a pre-trained time series prediction model is used to predict vehicle network connection failures within a future preset time, to obtain a vehicle network connection failure probability;

[0008] According to the vehicle network connection failure and its probability and the driving task of the vehicle, a driving risk level of the vehicle is determined;

[0009] According to the vehicle network connection failure, the driving task of the vehicle, and the driving risk level, a pre-trained machine learning model is used to select an optimal driving strategy including network resource scheduling and vehicle behavior guidance, and the corresponding vehicle is controlled to execute the optimal driving strategy according to the priority order of the driving task of the vehicle.

[0010] In a second aspect, an intelligent driving vehicle driving control system is provided, including units for executing the intelligent driving vehicle driving control method described above.

[0011] In a third aspect, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, and the processor executes the computer program to implement the steps of the intelligent driving vehicle driving control method described above.

[0012] In a fourth aspect, a computer readable storage medium is provided, which stores a computer program, and the computer program is executed by a processor to implement the steps of the intelligent driving vehicle driving control method described above.

[0013] Compared with the existing technology, the present invention integrates the real-time driving information from the vehicle and the network quality information from the network end, and uses the time series prediction model to make a probabilistic prediction of the network connection failure within the preset time in the future. Compared with the existing technology that only relies on the static threshold of the network side for post-judgment, the present invention can predict the vehicle network connection failure, so that the intelligent driving vehicle can change from passive response to active defense when driving, which is conducive to the subsequent avoidance measures. After predicting the failure, the vehicle's driving risk level can also be determined according to the vehicle network connection failure and its probability and the vehicle's driving task to evaluate the impact of the vehicle network connection failure on the vehicle's driving task, and then use pre-training to predict the vehicle network connection failure, the vehicle's driving task and the driving risk level. The machine learning model selects and generates an optimal driving strategy that includes network resource scheduling and vehicle behavior guidance, and controls the corresponding vehicles to execute the optimal driving strategy according to the priority order of the vehicle's driving business, so as to ensure that the vehicles performing the most critical driving business among all vehicles predicted to have faults obtain the highest priority network protection, realize adaptive, differentiated and refined fault recovery, and avoid the inefficiency and resource waste of the traditional "one-size-fits-all" strategy. It can be seen that the intelligent driving vehicle driving control method of the present invention can accurately predict and actively intervene before the network connection failure occurs, avoiding a potential network connection interruption, ensuring the continuity and reliability of the autonomous driving of L4 vehicles, and ensuring that the vehicles performing the most critical driving business obtain the highest priority network protection. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Figure 1 A schematic flow chart of a driving control method for an intelligent driving vehicle provided in the first embodiment of the present invention;

[0015] Figure 2 for Figure 1 A flow chart of a specific implementation of step S120;

[0016] Figure 3 for Figure 1 A schematic flow chart of a specific implementation of step S130;

[0017] Figure 4 for Figure 1 A schematic flow chart of a specific implementation of step S140;

[0018] Figure 5 A schematic flow chart of a driving control method for an intelligent driving vehicle provided in a second embodiment of the present invention;

[0019] Figure 6 1 is a schematic structural diagram of an intelligent driving vehicle control system provided by a first embodiment of the present invention;

[0020] Figure 7is a structural schematic diagram of an intelligent driving vehicle driving control system provided by a second embodiment of the present application.

[0021] Figure 8 is a structural schematic diagram of a computer device in an embodiment of the present application. DETAILED DESCRIPTION

[0022] The technical solutions in the embodiments of the present application will be clearly and completely described in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are some but not all of the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative work fall within the protection scope of the present application.

[0023] It should be understood that, when used in the specification and the appended claims, the terms "comprise" and "include" indicate the presence of the described features, integers, steps, operations, elements, and / or components, but do not exclude one or more other features, integers, steps, operations, elements, components, and / or groups thereof.

[0024] It should also be understood that the term "and / or" used in the specification and the appended claims of the present application means any combination of one or more of the associated listed items and all possible combinations, and includes these combinations.

[0025] Please refer to Figure 1 , Figure 1 is a schematic flowchart of an intelligent driving vehicle driving control method provided by a first embodiment of the present application. In the embodiment shown in the drawing, the intelligent driving vehicle driving control method comprises the following steps S110-S140:

[0026] S110, collecting vehicle driving information and network quality information of a base station near a vehicle position in real time.

[0027] In the present application, the vehicle driving information can include real-time position, speed, acceleration, planned driving path, and driving service type (such as high-precision positioning, environmental perception data uploading, etc.) of the vehicle; and the network quality information of the base station near the vehicle position can include network quality information of the base station and adjacent base stations at the real-time position of the vehicle. The network quality information can include wireless link quality indicators (RSRP (Reference Signal Receiving Power), RSRQ (Reference Signal Receiving Quality), SINR (Signal to Interference plus Noise Ratio)) of the base station, cell load, uplink / downlink rate, and other data.

[0028] S120. Based on the vehicle driving information, the network quality information, and the historical driving status of the vehicle on historically similar driving routes, a pre-trained time series prediction model is used to predict vehicle network connection failures within a preset future time period to obtain a vehicle network connection failure probability.

[0029] In this step, the time series prediction model uses network quality information and vehicle driving information as input, and combines the vehicle's historical driving status on similar routes to continuously predict the probability of vehicle network connection failure within a certain period of time (e.g., 5-30 seconds). Vehicle network connection failures can include network connection / handover failures, and historical driving status can include whether the vehicle's network connection has been disconnected.

[0030] like Figure 2 As shown, the step S120 may specifically include steps S121-S122:

[0031] S121. Utilize an LSTM-based time series prediction model to continuously predict the network quality of the base station network used by the vehicle within a preset time in the future based on the vehicle driving direction, speed, and driving path in the vehicle driving information and the network quality information of the base stations near the vehicle location.

[0032] In this embodiment, a pre-trained LSTM (Long Short-Term Memory)-based time series prediction model is used to perform a probabilistic prediction of vehicle network connection failures within a preset time in the future.

[0033] In this step, the area that the vehicle will reach within the future preset time can be continuously predicted based on the vehicle's driving direction, speed and driving path in the vehicle driving information, and the network quality of the base station network used by the vehicle within the future preset time can be predicted based on the network quality information of the base station and neighboring base stations at the vehicle's real-time location.

[0034] S122: Predicting a vehicle network connection failure based on the network quality and the vehicle's offline records in historical driving states along similar historical driving routes to obtain a vehicle network connection failure probability within a preset future time period.

[0035] In this step, the probability of a vehicle network connection failure within the future preset time period is predicted in combination with the vehicle's historical disconnection records on similar driving routes. If there have been multiple cases in the past where the intelligent driving vehicle has been disconnected in the area to which the vehicle is destined within the predicted future preset time period, the probability of a vehicle network connection failure is high.

[0036] Based on the above steps S121-S122, the LSTM-based time series prediction model can be pre-trained according to a fault prediction data set containing the occurred vehicle network connection fault, the corresponding vehicle driving information, the network quality information of the base station near the vehicle position, and the historical driving state of the vehicle on the historical similar driving path, that is, the LSTM-based time series prediction model can also make fault prediction in combination with the historical vehicle driving state of the historical similar driving path to further improve the accuracy of model prediction. For example, based on the LSTM-based time series prediction model, according to the speed, direction of a certain vehicle (vehicle A), and the current high user load of target cell B reached by the vehicle A and the drop record of the historically similar path, it is predicted that the probability (i.e., the vehicle network connection fault probability) of vehicle A failing to perform network switching due to entering the weak coverage edge of the adjacent cell (cell B) in 10 seconds is more than 95%.

[0037] S130, determining the driving risk level of the vehicle according to the vehicle network connection fault, the probability thereof, and the driving task of the vehicle.

[0038] Specifically, as shown in the step S130, the step S130 can include steps S131-S132: Figure 3

[0039] S131, if the probability of the vehicle network connection fault is greater than a preset threshold, determining the root cause of the vehicle network connection fault by using a root cause diagnosis method.

[0040] In this step, when the probability of the vehicle network connection fault is greater than a preset threshold (such as 60%), it indicates that the fault will occur in the future preset time period. In the present application, the root cause of the vehicle network connection fault is also determined by using a root cause diagnosis method. For example, in the above example of switching failure, the root cause of the fault is confirmed to be "switching failure caused by target cell congestion" by this step.

[0041] Preferably, this step can also determine the fault root cause by using a root cause diagnosis method through an LSTM-based time series prediction model.

[0042] S132, evaluating the influence of the vehicle network connection fault on the driving task of the vehicle according to the root cause of the vehicle network connection fault and the driving task of the vehicle, thereby determining the driving risk level of the vehicle.

[0043] In this step, the influence of the vehicle network connection fault on the driving task of the vehicle can be evaluated according to the root cause of the vehicle network connection fault and the dependence level of the network function of the current or to-be-executed driving task (such as lane changing, cruising, parking) of the vehicle based on a rule engine or a deep learning model for risk level determination, and the driving risk level (for example: low, medium, high, and extremely high) is determined.

[0044] ​It can be understood that the knowledge base or rule base can be predefined to define the level of dependence on network functions (e.g., the level of network delay requirement, etc.) for different driving tasks, for example, if the driving task is "following cruise", the network delay requirement is medium, which corresponds to the above-mentioned example of failure of entering the neighbor area switch after 10S, the rule engine or the deep learning model for risk level determination can evaluate the driving risk level of this potential failure as medium.

[0045] S140, according to the vehicle network connection failure, the driving task of the vehicle and the driving risk level, the pre-trained machine learning model is used to select and generate an optimal driving strategy containing network resource scheduling and vehicle behavior guidance, and the corresponding vehicle is controlled to execute the optimal driving strategy according to the priority order of the driving service of the vehicle.

[0046] In the present application, the machine learning model can be a model based on reinforcement learning, and after generating the driving strategy, the corresponding vehicle can be controlled to execute the optimal driving strategy according to the priority order determined by the driving service types (such as high-precision positioning, environmental perception data uploading, etc.) of all vehicles predicted to fail within a future preset time, to adaptively determine the control execution order, which can improve the utilization rate of network resources and ensure that the most critical driving service obtains the highest priority network guarantee and recovers the failure first, for example, if the priority of the driving service type of high-precision positioning is higher than that of the driving service type of environmental perception data uploading, the vehicle performing the driving service type of high-precision positioning is preferentially controlled to execute the optimal driving strategy for failure recovery.

[0047] In the present embodiment, as shown in Figure 4 The step S140 of generating an optimal driving strategy containing network resource scheduling and vehicle behavior guidance according to the vehicle network connection failure, the driving task of the vehicle and the driving risk level by using the pre-trained machine learning model can specifically include steps S141-S142:

[0048] S141, using a pre-trained machine learning model to generate a plurality of candidate driving strategies containing network resource scheduling and vehicle behavior guidance and corresponding reward functions according to the vehicle network connection failure, the driving task of the vehicle and the driving risk level.

[0049] In the present invention, the pre-trained machine learning model can implement network resource scheduling from aspects such as network rerouting, network resource reallocation and / or master-slave switching according to vehicle network connection failure, and generate an alternative driving strategy including network resource scheduling and vehicle behavior guidance in combination with the current or upcoming driving task and the driving risk level; the reward function includes a reward function determined by the degree of impact on the current driving task after executing the alternative driving strategy, and the lower the degree of impact on the current driving task, the higher the reward function value.

[0050] In this embodiment, the generated alternative driving strategies may include: not intervening in the current vehicle driving strategy; or forcing the vehicle to switch the signal connection path in advance to connect to a neighboring base station with a stable signal; or reducing the vehicle's driving speed and delaying entry into an area where the vehicle network connection failure has occurred. For example, in the example of a failed handover to a neighboring cell after 10 seconds, executing the strategy of not intervening would result in the vehicle being disconnected and driving degradation, and the reward function value generated by the model for this alternative driving strategy would be approximately -100. If the strategy of forcing the vehicle to switch the signal connection path in advance to a non-optimal but stable neighboring base station is executed, since only the channel connection is switched, the vehicle's speed and direction remain unchanged, and the network can be uninterrupted. Although the bandwidth is slightly lower, it does not affect the current driving task, and the reward function value is approximately +80. If the strategy of reducing the vehicle's driving speed and delaying entry into an area where the vehicle network connection failure has occurred is executed, the smoothness of completing the driving task may be affected, and the reward function value is approximately +20.

[0051] S142. Select the alternative driving strategy with the largest reward function value as the optimal driving strategy.

[0052] In this embodiment, since the lower the impact on the current driving task, the higher the reward function value, the alternative driving strategy with the largest reward function value is selected as the optimal driving strategy. While avoiding network connection failures, the smooth completion of the driving task is ensured as much as possible, thereby improving the overall driving safety and continuity of intelligent driving.

[0053] It can be understood that the intelligent driving vehicle driving control method of the present invention can be applied to the server, and multiple vehicles can communicate with the server through the 5G core network to exchange data. Corresponding to the above example of failure to enter the neighboring cell handover after 10S, if the strategy of forcing the vehicle (vehicle A) to switch the signal connection path in advance and switch to a non-optimal but stable signal neighboring cell base station (such as the base station of cell C next to cell B) is selected, the signal connection path is forced to be switched, and the specific implementation steps are as follows:

[0054] The server establishes a secure connection channel through the SSH protocol by using the Netconf (interface protocol, which ensures the atomicity of transaction operations) to connect to the AMF (Access and Mobility Management Function, access and mobility management function of the core network) specified port for policy instruction interaction (mandatory handover instruction), containing the SUPI of vehicle A and the physical cell identifier of target cell C;

[0055] The AMF calls the Nudm_SDM (user data management service) service to obtain the subscription data of the user (vehicle A) from the UDM (Unified Data Management Function), and verifies the slice access right (whether the S-NSSAI matches, etc.) of the target cell C;

[0056] The AMF sends a HandoverRequest (cell handover control signaling) message to the source base station (the base station at the current location of vehicle A) through the NG-AP (signaling protocol between the core network and the base station), carrying the physical cell identifier of the target cell C, QoS flow parameters and handover reason, and the source base station sends a HandoverRequest to the target base station (base station of target cell C) through the Xn interface (interface between base stations) to request resource allocation (beamforming configuration of target cell, etc.);

[0057] The source base station sends a handover command to vehicle A through an RRC reconfiguration message, which contains the SSB frequency point (synchronization signal block, used for vehicle and cell docking) of the target cell C, random access configuration (physical random access channel resource index) and other parameters. Vehicle A performs non-competitive random access and sends an RRC Reconfiguration Complete confirmation after completing access in the target cell C. The source base station sends a Path Switch Request (signaling message for path switching) to the AMF through the NG-AP to complete the signal connection path switching.

[0058] It can be seen that, in the above scheme, the intelligent driving vehicle driving control method of the application fuses real-time driving information from the vehicle and network quality information from the network side, and predicts the probability of vehicle network connection failure in the future preset time through a time series prediction model. Compared with the prior art which only relies on network side static threshold for post-judgment, the application can predict network connection failure, so that the intelligent driving vehicle driving changes from passive response to active defense, which is beneficial to subsequent evasion measures. After predicting the failure, the driving risk level of the vehicle can be determined according to the vehicle network connection failure, its probability and the driving task of the vehicle, so as to evaluate the influence of the vehicle network connection failure on the driving task of the vehicle. Then, the pre-trained machine learning model is used to select and generate the optimal driving strategy containing network resource scheduling and vehicle behavior guidance according to the vehicle network connection failure, the driving task of the vehicle and the driving risk level. The corresponding vehicle is controlled to execute the optimal driving strategy according to the priority order of the driving business of the vehicle, so as to ensure that the vehicle executing the most critical driving business among all the vehicles predicted to fail obtains the highest priority network guarantee, realize adaptive, differentiated and refined failure recovery. It can be seen that the intelligent driving vehicle driving control method of the application can accurately predict and actively intervene before the network connection failure occurs, avoid a potential network connection interruption, and resolve the potential network connection failure without affecting the driving safety, thereby ensuring the continuity and reliability of the automatic driving of the L4 vehicle.

[0059] It should be noted that the size of the serial number of each step in the above embodiment does not mean the order of execution, and the execution order of each process should be determined according to its function and inherent logic, and should not constitute any limitation on the implementation process of the embodiment of the application.

[0060] Referring to Figure 5 , Figure 5 The flowchart of the intelligent driving vehicle driving control method provided by the second embodiment of the application is shown in the figure. As shown in the figure, the method comprises the following steps S200-S290:

[0061] S200, real-time collection of vehicle driving information and network quality information of base stations near the vehicle position.

[0062] Initially, the vehicle driving information and the network quality information of the base stations near the vehicle position are acquired in real time, so as to predict the vehicle network connection failure in the future preset time period according to the real-time acquired information or detect the hardware failure of the key intersection in the driving path. Correspondingly, steps S210 and S250 are executed respectively. Since step S200 is the same as or similar to step S110, it will not be described here.

[0063] S210, predicting vehicle network connection failure in a future preset time according to the vehicle driving information, the network quality information, and historical driving states of the vehicle in a historically similar driving path, by using a pre-trained time series prediction model, to obtain a vehicle network connection failure probability.

[0064] This step is the same as or similar to step S120, and will not be described here.

[0065] S220, determining a driving risk level of the vehicle according to the vehicle network connection failure, the probability thereof, and the driving task of the vehicle.

[0066] This step is the same as or similar to step S130, and will not be described here.

[0067] S230, selecting an optimal driving strategy containing network resource scheduling and vehicle behavior guidance according to the vehicle network connection failure, the driving task of the vehicle, and the driving risk level by using a pre-trained machine learning model, and controlling the corresponding vehicle to execute the optimal driving strategy according to the priority order of the driving task of the vehicle.

[0068] This step is the same as or similar to step S140, and will not be described here.

[0069] S240, continuously monitoring the network connection state of the vehicle, so as to obtain the network connection stability of the vehicle.

[0070] In this step, after the vehicle executes the optimal driving strategy, the network connection state of the vehicle is continuously monitored to monitor whether the network connection of the vehicle after execution is stable, so as to guarantee the reliability and continuity of vehicle driving, and also to verify whether the executed optimal driving strategy is effective.

[0071] S250, detecting whether an RSU at a key intersection in a driving path is offline according to the driving path in the vehicle driving information.

[0072] In the present application, the vehicle obtains front intersection traffic light information and pedestrian warning through an RSU (Roadside Unit) used for V2I communication with the vehicle, and if the RSU fails, it may cause a driving accident.

[0073] In the present embodiment, whether an RSU at a key intersection in a vehicle driving path is offline can be detected by a heartbeat packet detection mechanism, and if it is offline, a failure alarm can be triggered immediately; wherein the key intersection can be a crossroads or a road intersection with heavy traffic.

[0074] S260, if the RSU is offline, identify the service range of the RSU, search for vehicles about to enter the service range according to the service range of the RSU and the vehicle travel information, and determine that the travel risk level of the vehicle about to enter the service range is high risk.

[0075] Due to the RSU offline, the intelligent driving vehicle is prone to cause red light running or collision risk due to the lack of RSU information of the key intersection, which may cause a major traffic accident, so the travel risk level is directly determined as high risk.

[0076] S270, generate a travel strategy containing source switching and vehicle speed reduction according to the RSU offline situation and the travel risk level of the vehicle, so as to download the traffic light information of the key intersection published by the urban traffic management center to the corresponding vehicle through the cellular network, and control the vehicle to reduce the maximum speed limit and improve the target detection sensitivity of the vehicle-mounted sensor.

[0077] In this step, the RSU offline situation includes whether the RSU is offline and the specific setting position of the RSU. Due to the RSU offline, the traffic light information and pedestrian information cannot be obtained from the RSU, so the source needs to be switched. In this embodiment, the urban traffic management center is used as a backup source. The urban traffic management center is the management center of the intelligent traffic system in the urban area (such as some intelligent networked vehicle demonstration areas and pilot areas), which can monitor, manage and optimize the traffic flow of the whole city or area in real time, and integrates data aggregation, processing, decision-making and command. It downloads the corresponding traffic light information to the vehicle about to enter the service range of the offline RSU through the 5G cellular network, realizes the seamless replacement of key information, and in order to make up for the lack of V2P (vehicle-to-pedestrian) information, the intelligent driving vehicle travel control method of this embodiment can also issue instructions for cautious driving, control the vehicle to reduce speed, and at the same time improve the target detection sensitivity of the vehicle-mounted sensor (such as camera, laser radar), so as to further improve the driving reliability of the intelligent driving vehicle.

[0078] It can be understood that the ordinary machine learning model can be used to generate a travel strategy containing source switching and vehicle speed reduction according to the RSU offline situation and the travel risk level of the vehicle. The RSU at different key intersections plays different roles in traffic management. The speed reduction amplitude of the vehicle and the accuracy improvement of the specific vehicle-mounted sensor in the travel strategy can be learned by the model according to the specific training set composed of the RSU offline situation and the travel risk level of the vehicle.

[0079] S280, uniformly process the vehicle travel information collected each time, the network quality information of the base station near the vehicle position, the predicted failure, the travel risk level, the generated travel strategy and the monitored network connection state after execution to obtain structured data.

[0080] In this embodiment, the vehicle driving information collected each time, the network quality information of the base station near the vehicle position, the predicted vehicle network connection failure and its probability, the hardware failure, the determined driving risk level, the generated multiple alternative driving strategies, and the selected optimal driving strategy and the monitored network connection state after execution can be uniformly formatted as a set of structured data, added to the training data set, and the above models are optimized.

[0081] S290, continuously iteratively updating the model according to the structured data.

[0082] In this step, the LSTM-based time series prediction model for failure probability prediction, the LSTM-based time series prediction model for root cause diagnosis, the rule engine or deep learning model for risk level determination, and the machine learning model for driving strategy generation are iteratively updated according to the input data (real-time collected data) and the data obtained by executing the above steps, respectively, to continuously optimize the model parameters, so that each model has the ability to evolve, and as the running time increases, the output result will be more and more accurate, and the decision will be more and more intelligent, forming a virtuous cycle. Understandably, the position of each data in the structured data can be fixed, each data has a unique identifier, and when the model is iterated, only the data required by each model itself is selected for iterative update.

[0083] As can be seen, in the above scheme, the intelligent driving vehicle driving control method of the present application can change from "after-response" to "prior prediction and active intervention", and can dynamically generate multiple alternative driving strategies according to the specific driving intention (driving task and driving business) of the vehicle and the vehicle network connection failure, and execute the optimal driving strategy according to the driving business priority order, so as to adaptively generate a driving strategy for failure recovery without affecting driving safety, and to resolve potential network failures without feeling, which can improve the safety, continuity and network resource utilization of L4 automatic driving, and also can iterate the model according to the input and generated data each time to continuously optimize the model.

[0084] Reference Figure 6 , Figure 6 The structure diagram of the intelligent driving vehicle driving control system provided by the first embodiment of the present application is shown in the figure. In the embodiment shown in the figure, the intelligent driving vehicle driving control system comprises a data acquisition unit 101, a failure prediction unit 102, a diagnosis and risk assessment unit 103, and a strategy generation unit 104, and the functions of each unit are described as follows:

[0085] The data acquisition unit 101 is used for real-time acquisition of vehicle driving information and network quality information of base stations near the vehicle position.

[0086] The fault prediction unit 102 is configured to predict a vehicle network connection fault in a future preset time period according to the vehicle driving information, the network quality information, and historical driving states of the vehicle in a historical similar driving path, and obtain a vehicle network connection fault probability by using a pre-trained time series prediction model.

[0087] The diagnosis and risk assessment unit 103 is configured to determine a driving risk level of the vehicle according to the vehicle network connection fault, the probability thereof, and a driving task of the vehicle.

[0088] The strategy generation unit 104 is configured to select an optimal driving strategy including network resource scheduling and vehicle behavior guidance by using a pre-trained machine learning model according to the vehicle network connection fault, the driving task of the vehicle, and the driving risk level, and control the corresponding vehicle to execute the optimal driving strategy according to a priority order of the driving task of the vehicle.

[0089] In some embodiments, the fault prediction unit 102 is specifically configured to:

[0090] The LSTM-based time series prediction model is used to continuously predict network quality of a base station network used by the vehicle in the future preset time period according to the vehicle driving direction, speed, and driving path in the vehicle driving information, and the network quality information of the base station near the vehicle position.

[0091] The vehicle network connection fault is predicted according to the network quality and a disconnection record in the historical driving state of the vehicle in the historical similar driving path, and a vehicle network connection fault probability in the future preset time period is obtained.

[0092] In some embodiments, the diagnosis and risk assessment unit 103 is specifically configured to:

[0093] If the vehicle network connection fault probability is greater than a preset threshold, a root cause diagnosis method is used to determine a root cause of the vehicle network connection fault.

[0094] According to the root cause of the vehicle network connection fault and the driving task of the vehicle, an influence of the vehicle network connection fault on the driving task of the vehicle is evaluated, and a driving risk level of the vehicle is determined.

[0095] In some embodiments, the strategy generation unit 104 is specifically configured to:

[0096] A plurality of candidate driving strategies including network resource scheduling and vehicle behavior guidance and corresponding reward functions are generated by using a pre-trained machine learning model according to the vehicle network connection fault, the driving task of the vehicle, and the driving risk level, wherein the reward function includes a reward function determined by an influence degree on the current driving task after the candidate driving strategy is executed.

[0097] select the candidate driving strategy with the maximum reward function value as the optimal driving strategy.

[0098] In some embodiments, the candidate driving strategies generated by the strategy generation unit 104 include:

[0099] not intervening in the current vehicle driving strategy; or

[0100] forcing the vehicle to switch the signal connection path to switch to the neighboring base station with stable signal; or

[0101] reducing the vehicle driving speed to delay entering the area where the vehicle network connection fault occurs.

[0102] It can be seen that the present application provides an intelligent driving vehicle driving control system, which can probabilistically predict the vehicle network connection fault within a future preset time through a time series prediction model, so that the intelligent driving vehicle driving changes from passive response to active defense, which is conducive to subsequent avoidance measures. After predicting the fault, the driving risk level of the vehicle can be determined according to the vehicle network connection fault, its probability and the driving task of the vehicle, so as to evaluate the influence of the vehicle network connection fault on the driving task of the vehicle. Then, the pre-trained machine learning model is used to select and generate the optimal driving strategy containing network resource scheduling and vehicle behavior guidance according to the vehicle network connection fault, the driving task of the vehicle and the driving risk level. The corresponding vehicle is controlled to execute the optimal driving strategy according to the priority order of the driving business of the vehicle, so as to ensure that the vehicle executing the most critical driving business among all the vehicles predicted to have faults obtains the highest priority network guarantee, realizes adaptive, differentiated and refined fault recovery, and it can be known that the present application can accurately predict and actively intervene before the network connection fault occurs, avoid a potential network connection interruption, guarantee the continuity and reliability of L4 vehicle automatic driving, and ensure that the vehicle executing the most critical driving business obtains the highest priority network guarantee.

[0103] Referring to Figure 7 , Figure 7 The present application provides a schematic block diagram of the intelligent driving vehicle driving control system. The intelligent driving vehicle driving control system provided by the present embodiment adds a monitoring verification unit 105, a hardware detection unit 106, a search determination unit 107, a data processing unit 108 and an updating unit 109 compared with the intelligent driving vehicle driving control system provided by the first embodiment. The functions of the added functional units are described in detail as follows:

[0104] The monitoring verification unit 105 is used to continuously monitor the network connection state of the vehicle after the vehicle executes the optimal driving strategy, so as to know the network connection stability of the vehicle.

[0105] The hardware detection unit 106 is configured to detect whether an RSU at a key intersection in a driving path is offline according to the driving path in the vehicle driving information.

[0106] The search determination unit 107 is configured to, if the RSU is offline, identify a service range of the RSU, search for a vehicle about to enter the service range according to the service range of the RSU and the vehicle driving information, and determine that a driving risk level of the vehicle about to enter the service range is high risk.

[0107] The strategy generation unit 104 is further configured to generate a driving strategy including source switching and vehicle speed reduction according to the RSU offline condition and the driving risk level of the vehicle, so as to download the traffic light information of the key intersection issued by the urban traffic management center to the corresponding vehicle through the cellular network, and control the vehicle to reduce the maximum speed limit and improve the target detection sensitivity of the vehicle-mounted sensor. Understandably, the strategy generation unit 104 can include a machine learning model based on reinforcement learning and a general machine learning model, which respectively correspond to the vehicle network connection fault and the hardware fault to generate the driving strategy.

[0108] The data processing unit 108 is configured to uniformly process the vehicle driving information collected each time, the network quality information of the base station near the vehicle position, the predicted fault, the driving risk level, the generated driving strategy, and the monitored network connection state after execution to obtain structured data.

[0109] The update unit 109 is configured to continuously and iteratively update the model according to the structured data.

[0110] It can be seen that the intelligent driving vehicle driving control system can also detect hardware faults to ensure the driving safety and continuity of the intelligent vehicle from multiple dimensions, and can also iteratively update the model according to each input and generated data to continuously optimize the model and further improve the accuracy of the driving strategy generation.

[0111] It should be noted that those skilled in the art can clearly understand the specific implementation process of the above-mentioned intelligent driving vehicle driving control system and each unit or component, which can refer to the corresponding description in the foregoing method embodiments. For the convenience and brevity of description, it will not be repeated here.

[0112] In one embodiment, a computer device is provided, and an internal structure diagram of the computer device can be as shown in FIG. 8. Figure 8The computer device includes a processor, a memory, a network interface and a database connected through a system bus. The processor of the computer device is configured to provide computing and control capabilities. The memory of the computer device includes a non-volatile and / or volatile storage medium, an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for running the operating system and the computer program in the non-volatile storage medium. The network interface of the computer device is configured to communicate with external devices through a network connection. Further, the computer device can also include a display screen and an input device (e.g., a mouse, a keyboard, etc.) for interaction.

[0113] In particular, the processor in the computer device implements the steps of the intelligent driving vehicle driving control method provided by the first embodiment and the second embodiment when executing the computer program.

[0114] In an embodiment, the present application can also provide a computer readable storage medium having a computer program stored thereon, the computer program being executed by a processor to implement the steps of the intelligent driving vehicle driving control method provided by the first embodiment and the second embodiment.

[0115] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiments can be completed by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer readable storage medium. When executed, the computer program can include the processes of the above-mentioned embodiments. Any reference to memory, storage, database or other medium used in the embodiments provided by the present application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. As an illustration but not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDR SDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM) and memory bus dynamic RAM (RDRAM), etc.

[0116] Those skilled in the art can clearly understand that, for the convenience and brevity of description, only the above-mentioned division of each functional unit and module is exemplified, and in actual application, the above-mentioned functions can be completed by different functional units and modules according to needs, that is, the internal structure of the system is divided into different functional units or modules to complete all or part of the functions described above.

[0117] The above examples are only used to illustrate the technical solutions of the present application, but not to limit it; although the present application has been described in detail with reference to the foregoing examples, those skilled in the art should understand that the technical solutions recorded in the foregoing examples can still be modified, or some technical features can be replaced by equivalents; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the scope of the technical solutions of the embodiments of the present application, and should be included in the protection scope of the present application.

Claims

1. A driving control method for an intelligent driving vehicle, characterized in that: The intelligent driving vehicle driving control method includes: Real-time collection of vehicle driving information and network quality information of base stations near the vehicle location; Based on the vehicle driving information, the network quality information, and the historical driving status of the vehicle on similar historical driving routes, a pre-trained time series prediction model is used to predict vehicle network connection failures within a preset time period in the future to obtain a vehicle network connection failure probability; determining a driving risk level of the vehicle based on the vehicle network connection failure and its probability and the vehicle's driving task; Based on the vehicle network connection failure, the vehicle's driving task and the driving risk level, a pre-trained machine learning model is used to select and generate an optimal driving strategy that includes network resource scheduling and vehicle behavior guidance, and the corresponding vehicle is controlled to execute the optimal driving strategy according to the priority order of the vehicle's driving business.

2. The intelligent driving vehicle driving control method according to claim 1, characterized in that: The method of predicting a vehicle network connection failure within a preset time period in the future using a pre-trained time series prediction model based on the vehicle driving information, the network quality information, and the historical driving status of the vehicle on similar historical driving routes, and obtaining a vehicle network connection failure probability specifically includes: Utilizing an LSTM-based time series prediction model to continuously predict the network quality of the base station network used by the vehicle within a preset future time period based on the vehicle's driving direction, speed, and driving path in the vehicle's driving information and the network quality information of base stations near the vehicle's location; The vehicle network connection failure is predicted based on the network quality and the offline record of the vehicle in the historical driving state of similar historical driving routes, so as to obtain the vehicle network connection failure probability within a preset time in the future.

3. The intelligent driving vehicle driving control method according to claim 1, characterized in that: Determining the driving risk level of the vehicle based on the vehicle network connection failure and its probability and the vehicle driving task specifically includes: If the vehicle network connection failure probability is greater than a preset threshold, determining the root cause of the vehicle network connection failure using a root cause diagnosis method; Based on the root cause of the vehicle network connection failure and the driving task of the vehicle, the impact of the vehicle network connection failure on the vehicle driving task is evaluated to determine the driving risk level of the vehicle.

4. The intelligent driving vehicle driving control method according to claim 1, characterized in that: The selecting and generating an optimal driving strategy including network resource scheduling and vehicle behavior guidance using a pre-trained machine learning model according to the vehicle network connection failure, the vehicle driving task, and the driving risk level specifically includes: generating, using a pre-trained machine learning model, a plurality of alternative driving strategies including network resource scheduling and vehicle behavior guidance and corresponding reward functions based on the vehicle network connection failure, the vehicle's driving task, and the driving risk level; wherein the reward function includes a reward function determined by the degree of impact of executing the alternative driving strategy on the current driving task; The alternative driving strategy with the largest reward function value is selected as the optimal driving strategy.

5. The intelligent driving vehicle driving control method according to claim 4, characterized in that: The alternative driving strategies include: Do not intervene in the current vehicle driving strategy; or Force the vehicle to switch the signal connection path to connect to a neighboring base station with a stable signal; or Reduce vehicle speed and delay entry into areas where vehicle network connectivity issues are occurring.

6. The intelligent driving vehicle driving control method according to claim 1, characterized in that: After controlling the corresponding vehicles to execute the optimal driving strategy according to the priority order of the driving tasks of the vehicles, the method further includes: Continuously monitoring the network connection status of the vehicle to obtain the network connection stability of the vehicle; The vehicle driving information collected each time, the network quality information of the base stations near the vehicle location, the predicted faults, the driving risk level, the generated driving strategy and the network connection status monitored after execution are uniformly processed to obtain structured data; The model is continuously iteratively updated according to the structured data.

7. The intelligent driving vehicle driving control method according to claim 1, characterized in that: The intelligent driving vehicle driving control method further includes: Detecting whether the RSU at the key intersection in the driving path is offline according to the driving path in the vehicle driving information; If an RSU is offline, identify the service range of the RSU, search for a vehicle that is about to enter the service range based on the service range of the RSU and the vehicle driving information, and determine that the driving risk level of the vehicle that is about to enter the service range is high risk; Based on the RSU offline status and the vehicle's driving risk level, a driving strategy including signal source switching and vehicle speed reduction is generated to transmit the traffic light information of the key intersection issued by the urban traffic management center to the corresponding vehicle through the cellular network, control the vehicle to reduce the maximum speed limit, and improve the target detection sensitivity of the on-board sensor.

8. An intelligent driving vehicle control system, characterized in that: The method comprises a unit for executing the intelligent driving vehicle driving control method as described in any one of claims 1 to 7.

9. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the computer program, the steps of the intelligent driving vehicle driving control method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the steps of the intelligent driving vehicle driving control method according to any one of claims 1 to 7 are implemented.

Images