Safety assessment method and system for unmanned test of remote driving system

Abstract

The invention relates to a safety evaluation method and system for an unmanned test remote driving system, and the method comprises the steps: R1, collecting the data information of a video plug flow state and the data information of a network monitoring state of a remote driving end in a process that the remote driving system takes over an automatic driving vehicle, the data information of fault diagnosis of the vehicle at the automatic driving end and the data information of the surrounding environment are obtained in real time; and R2, based on the data information of the video plug flow state and the data information of the network monitoring state of the remote driving end, predicting the initial take-over state of the remote driving end by adopting a BP neural network regression prediction algorithm based on a fuzzy rule, and obtaining the predicted data information of the initial take-over state of the remote driving end. According to the method, effective safety strategies are formulated for risk problems possibly occurring in a remote driving system, the safety of vehicle driving before and after remote driving take over can be ensured, and timely intervention can be performed when automatic driving is abnormal, so that the safety is ensured.

Description

Technical Field

[0001] The present invention relates to the field of remote driving technology, and in particular to a safety assessment method and system for unmanned testing of a remote driving system. Background Art

[0002] With the vigorous development of science and technology, autonomous driving technology has flourished in recent years. However, based on the current autonomous driving technology, it is still impossible to completely get rid of the assistance and control of manual driving safety officers, that is, the vehicle needs to be monitored by someone, especially in complex road scenes. At this time, if you want to achieve a completely unmanned vehicle, remote driving technology is an effective control solution. In the event of an abnormality in the vehicle's automatic driving, the vehicle can be controlled remotely to get out of trouble. However, in order to ensure the safety of remote driving, a set of feasible and effective safety strategies is an important direction for the development of remote driving technology. Summary of the invention

[0003] In view of the above problems, the present invention provides a safety assessment method and system for unmanned testing of remote driving systems, which not only formulates effective safety strategies for possible risk issues in the remote driving system to ensure the safety of vehicle driving before and after remote driving takes over, but also can intervene in time when abnormalities occur in automatic driving to ensure safety.

[0004] In order to achieve the above-mentioned purpose and other related purposes, the technical solution provided by the present invention is as follows:

[0005] A safety assessment method for unmanned testing of a remote driving system, the method comprising:

[0006] R1. When the remote driving system takes over the autonomous driving vehicle, it collects data information about the video streaming status and network monitoring status of the remote driving end, and obtains data information about the fault diagnosis of the vehicle and the surrounding environment of the autonomous driving end in real time;

[0007] R2. Based on the data information of the video streaming status of the remote driving terminal and the data information of the network monitoring status, the BP neural network regression prediction algorithm based on fuzzy rules is used to predict the initial takeover status of the remote driving terminal, and the data information of the initial takeover status of the remote driving terminal after the prediction is obtained;

[0008] R3. Based on the predicted data information of the preliminary takeover state of the remote driving end, the data information of the fault diagnosis of the vehicle on the automatic driving end, and the data information of the surrounding environment, a remote driving takeover safety situation prediction model based on an improved recursive neural network is constructed to predict the safety situation of the remote driving takeover, and obtain the predicted data information of the safety situation of the remote driving takeover;

[0009] R4. Based on the data information of the predicted safety situation of the remote driving takeover, a safety assessment function G of the remote driving takeover is constructed to evaluate the safety of the remote driving takeover, and obtain data information of the safety assessment of the remote driving takeover.

[0010] Furthermore, the method further comprises:

[0011] R5. Based on the data information of the safety assessment of the remote driving takeover, a preset threshold is set. If the safety assessment of the remote driving takeover is less than the preset threshold, the requirement is not met and remote takeover cannot be performed. If the safety assessment of the remote driving takeover is greater than the preset threshold, the requirement is met and remote takeover of the autonomous driving vehicle is performed.

[0012] Furthermore, in step R2, the use of the fuzzy rule-based BP neural network regression prediction algorithm to predict the initial takeover state of the remote driving terminal includes:

[0013] R21. Based on the data information of the video streaming status of the remote driving terminal and the data information of the network monitoring status, a remote driving terminal state fuzzy rule function Q is established.

[0014]

[0015] Among them, x 1 The data information of the video streaming status of the remote driving end, x 2 is the data information of the network monitoring status of the remote driving end, α 1 , α 2 and α 3 is the fuzzy rule factor of the state of the remote driving terminal, characterizes the fuzzy rule matrix of the state of the remote driving terminal, and obtains data information of the fuzzy rule matrix of the state of the remote driving terminal;

[0016] R22. Input the data information of the fuzzy rule matrix of the state of the remote driving terminal into the BP neural network regression prediction model for training and learning, and determine the prediction function W of the model,

[0017] Among them, y is the data information of the fuzzy rule matrix of the state of the remote driving end, β 1 , β 2 and β 3 is the penalty factor for the model's prediction, and the trained BP neural network regression prediction model is obtained;

[0018] R23. Based on the trained BP neural network regression prediction model, the data information of the fuzzy rule matrix of the state of the remote driving end is input, and the initial takeover state of the remote driving end is predicted to obtain the data information of the predicted initial takeover state of the remote driving end.

[0019] Furthermore, the penalty factor β of the fuzzy prediction is 1 , β 2 and β 3 for,

[0020]

[0021]

[0022] Among them, y is the data information of the fuzzy rule matrix of the state of the remote driving end.

[0023] Furthermore, the fuzzy rule factor α of the state of the remote driving terminal 1 , α 2 and α 3 for,

[0024]

[0025] Among them, x 1 The data information of the video streaming status of the remote driving end, x 2 It is the data information of the network monitoring status of the remote driving terminal.

[0026] Furthermore, in step R3, the construction of a remote driving takeover safety situation prediction model based on an improved recursive neural network to predict the safety situation of the remote driving takeover includes:

[0027] R31. Based on the predicted data information of the initial takeover state of the remote driving end, the data information of the fault diagnosis of the vehicle of the automatic driving end and the data information of the surrounding environment, a time series function P of a recursive neural network is established.

[0028]

[0029] Among them, z 1 is the data information of the initial takeover status of the remote driving end after prediction, 2 Data information for fault diagnosis of vehicles on the autonomous driving side, 3 is the data information of the surrounding environment of the vehicle on the autonomous driving side, δ 1 , δ 2 and δ 3 is the weight factor of the time series of the recurrent neural network, characterizes the time series of the remote driving end taking over the vehicle, and obtains the data information of the time series of the remote driving end taking over the vehicle;

[0030] R32. Input the time series data information of the remote driving end taking over the vehicle into the remote driving takeover safety situation prediction model for training and learning, and determine the remote driving takeover safety situation prediction function S,

[0031]

[0032] Among them, r is the time series data information of the remote takeover vehicle, γ 1 , γ 2 and γ 3 The reward factor for the remote driving takeover safety situation is used to obtain a trained remote driving takeover safety situation prediction model;

[0033] R33. Based on the trained remote driving takeover safety situation prediction model, the time series data information of the remote driving end taking over the vehicle is input to predict the safety situation of the remote driving takeover, and the data information of the predicted safety situation of the remote driving takeover is obtained.

[0034] Furthermore, the reward factor γ for the remote driving to take over the safety situation 1 , γ 2 and γ 3 The constraints are:

[0035]

[0036] Furthermore, the safety assessment function G of the remote driving takeover is:

[0037]

[0038] Among them, h is the data information of the safety situation of remote driving takeover after prediction, η 1 , η 2 and η 3 Safety assessment factors for remote driving takeover.

[0039] In order to achieve the above-mentioned objectives and other related objectives, the present invention also provides a safety assessment system for unmanned testing of remote driving systems, including a computer device, which is programmed or configured to execute the steps of any one of the safety assessment methods for unmanned testing of remote driving systems.

[0040] In order to achieve the above-mentioned purpose and other related purposes, the present invention also provides a computer-readable storage medium, on which is stored a computer program programmed or configured to execute any one of the safety assessment methods for unmanned testing remote driving systems.

[0041] The present invention has the following positive effects:

[0042] 1. The present invention predicts the initial takeover state of the remote driving end by adopting a BP neural network regression prediction algorithm based on fuzzy rules, and combines the construction of a remote driving takeover safety situation prediction model based on an improved recursive neural network to predict the safety situation of the remote driving takeover. It not only formulates effective safety strategies for possible risk problems in the remote driving system, but also ensures the safety of vehicle driving before and after the remote driving takeover, and can intervene in time when abnormalities occur in the automatic driving to ensure safety.

[0043] 2. The present invention evaluates the safety of remote driving takeover by constructing a safety assessment function G for remote driving takeover, which can not only further ensure the safety and reliability of remote driving takeover of autonomous driving vehicles, but also realize unmanned safety testing or operation of vehicles. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 It is a schematic diagram of the method flow of the present invention;

[0045] Figure 2 It is a flow chart of the BP neural network regression prediction algorithm based on fuzzy rules of the present invention;

[0046] Figure 3 The figure is a flow chart of the remote driving takeover safety situation prediction model based on the improved recursive neural network of the present invention. DETAILED DESCRIPTION

[0047] The following is a description of exemplary embodiments of the present disclosure in conjunction with the accompanying drawings, including various details of the embodiments of the present disclosure to facilitate understanding, which should be considered as merely exemplary. Therefore, it should be recognized by those of ordinary skill in the art that various changes and modifications may be made to the embodiments described herein without departing from the scope and spirit of the present disclosure. Similarly, for the sake of clarity and conciseness, descriptions of well-known functions and structures are omitted in the following description.

[0048] Example 1: Figure 1 As shown, a safety assessment method for unmanned testing of a remote driving system, the method comprising:

[0049] R1. When the remote driving system takes over the autonomous driving vehicle, it collects data information about the video streaming status and network monitoring status of the remote driving end, and obtains data information about the fault diagnosis of the vehicle and the surrounding environment of the autonomous driving end in real time;

[0050] R2. Based on the data information of the video streaming status of the remote driving terminal and the data information of the network monitoring status, the BP neural network regression prediction algorithm based on fuzzy rules is used to predict the initial takeover status of the remote driving terminal, and the data information of the initial takeover status of the remote driving terminal after the prediction is obtained;

[0051] R3. Based on the predicted data information of the preliminary takeover state of the remote driving end, the data information of the fault diagnosis of the vehicle on the automatic driving end, and the data information of the surrounding environment, a remote driving takeover safety situation prediction model based on an improved recursive neural network is constructed to predict the safety situation of the remote driving takeover, and obtain the predicted data information of the safety situation of the remote driving takeover;

[0052] R4. Based on the data information of the predicted safety situation of the remote driving takeover, a safety assessment function G of the remote driving takeover is constructed to evaluate the safety of the remote driving takeover, and obtain data information of the safety assessment of the remote driving takeover.

[0053] In this embodiment, the method further includes:

[0054] R5. Based on the data information of the safety assessment of the remote driving takeover, a preset threshold is set. If the safety assessment of the remote driving takeover is less than the preset threshold, the requirement is not met and remote takeover cannot be performed. If the safety assessment of the remote driving takeover is greater than the preset threshold, the requirement is met and remote takeover of the autonomous driving vehicle is performed.

[0055] In this embodiment, if Figure 2 As shown, in step R2, the use of the BP neural network regression prediction algorithm based on fuzzy rules to predict the initial takeover state of the remote driving end includes:

[0056] R21. Based on the data information of the video streaming status of the remote driving terminal and the data information of the network monitoring status, a remote driving terminal state fuzzy rule function Q is established.

[0057]

[0058] Among them, x 1 The data information of the video streaming status of the remote driving end, x 2 is the data information of the network monitoring status of the remote driving end, α 1 , α 2 and α 3 is the fuzzy rule factor of the state of the remote driving terminal, characterizes the fuzzy rule matrix of the state of the remote driving terminal, and obtains data information of the fuzzy rule matrix of the state of the remote driving terminal;

[0059] R22. Input the data information of the fuzzy rule matrix of the state of the remote driving terminal into the BP neural network regression prediction model for training and learning, and determine the prediction function W of the model,

[0060] Among them, y is the data information of the fuzzy rule matrix of the state of the remote driving end, β 1 , β 2 and β 3 is the penalty factor for the model's prediction, and the trained BP neural network regression prediction model is obtained;

[0061] R23. Based on the trained BP neural network regression prediction model, the data information of the fuzzy rule matrix of the state of the remote driving end is input, and the initial takeover state of the remote driving end is predicted to obtain the data information of the predicted initial takeover state of the remote driving end.

[0062] In this embodiment, the penalty factor β of the fuzzy prediction is 1 , β 2 and β 3 for,

[0063]

[0064] Among them, y is the data information of the fuzzy rule matrix of the state of the remote driving end.

[0065] In this embodiment, the fuzzy rule factor α of the state of the remote driving terminal is 1 , α 2 and α 3 for,

[0066]

[0067] Among them, x 1 The data information of the video streaming status of the remote driving end, x 2 It is the data information of the network monitoring status of the remote driving terminal.

[0068] In this embodiment, if Figure 3 As shown, in step R3, the construction of a remote driving takeover safety situation prediction model based on an improved recursive neural network to predict the safety situation of remote driving takeover includes:

[0069] R31. Based on the predicted data information of the initial takeover state of the remote driving end, the data information of the fault diagnosis of the vehicle of the automatic driving end and the data information of the surrounding environment, a time series function P of a recursive neural network is established.

[0070]

[0071] Among them, z 1 is the data information of the initial takeover status of the remote driving end after prediction, 2 Data information for fault diagnosis of vehicles on the autonomous driving side, 3 is the data information of the surrounding environment of the vehicle on the autonomous driving side, δ 1 ,δ 2 and δ 3 is the weight factor of the time series of the recurrent neural network, characterizes the time series of the remote driving end taking over the vehicle, and obtains the data information of the time series of the remote driving end taking over the vehicle;

[0072] R32. Input the time series data information of the remote driving end taking over the vehicle into the remote driving takeover safety situation prediction model for training and learning, and determine the remote driving takeover safety situation prediction function S,

[0073]

[0074] Among them, r is the time series data information of the remote takeover vehicle, γ 1 , γ 2 and γ 3 The reward factor for the remote driving takeover safety situation is used to obtain a trained remote driving takeover safety situation prediction model;

[0075] R33. Based on the trained remote driving takeover safety situation prediction model, the time series data information of the remote driving end taking over the vehicle is input to predict the safety situation of the remote driving takeover, and the data information of the predicted safety situation of the remote driving takeover is obtained.

[0076] In this embodiment, the reward factor γ for the remote driving to take over the safety situation 1 , γ 2 and γ 3 The constraints are:

[0077]

[0078] Example 2: Based on the safety assessment method for unmanned testing of a remote driving system in Example 1, the present invention is further illustrated and described below.

[0079] like Figure 1 As shown, a safety assessment method for unmanned testing of a remote driving system, the method comprising:

[0080] R1. When the remote driving system takes over the autonomous driving vehicle, it collects data information about the video streaming status and network monitoring status of the remote driving end, and obtains data information about the fault diagnosis of the vehicle and the surrounding environment of the autonomous driving end in real time;

[0081] R2. Based on the data information of the video streaming status of the remote driving terminal and the data information of the network monitoring status, the BP neural network regression prediction algorithm based on fuzzy rules is used to predict the initial takeover status of the remote driving terminal, and the data information of the initial takeover status of the remote driving terminal after the prediction is obtained;

[0082] R3. Based on the predicted data information of the preliminary takeover state of the remote driving end, the data information of the fault diagnosis of the vehicle on the automatic driving end, and the data information of the surrounding environment, a remote driving takeover safety situation prediction model based on an improved recursive neural network is constructed to predict the safety situation of the remote driving takeover, and obtain the predicted data information of the safety situation of the remote driving takeover;

[0083] R4. Based on the data information of the predicted safety situation of the remote driving takeover, a safety assessment function G of the remote driving takeover is constructed to evaluate the safety of the remote driving takeover, and obtain data information of the safety assessment of the remote driving takeover.

[0084] In this embodiment, the safety assessment function G of the remote driving takeover is:

[0085]

[0086] Among them, h is the data information of the safety situation of remote driving takeover after prediction, η 1 , η 2 and η 3 Safety assessment factors for remote driving takeover.

[0087] In this embodiment, the present invention provides a safety assessment system for unmanned testing of remote driving systems, including a computer device that is programmed or configured to execute any one of the steps of the safety assessment method for unmanned testing of remote driving systems.

[0088] In this embodiment, the system consists of: (1) an autonomous driving system; (2) a platform system; (3) a remote driving system;

[0089] 1. The autonomous driving system refers to a perception system consisting of radar (ultrasonic radar, millimeter wave radar, lidar), camera, inertial navigation and other sensors, and an autonomous driving controller, which, combined with software algorithms, should realize functions such as environmental perception, planning, positioning and control, and can realize autonomous driving in specific scenarios.

[0090] 2. The platform system: a visual network platform that receives various vehicle information to monitor the current status of the vehicle, including vehicle location, real-time operating status, scheduling status, vehicle operating fault status, etc.; it can receive vehicle automatic driving status information, process it, and send the information to the cabin in the form of a work order, and display it on the cockpit UI interface to inform the remote driver.

[0091] 3. The remote driving system is mainly divided into two system modules: the vehicle side and the remote cockpit side. The vehicle side includes a forward remote driving camera and a remote driving controller. The three fisheye cameras on the left, right and rear of the vehicle form a 360-degree surround view, which can receive the surrounding environmental information, convert and transmit it to the remote test bench cockpit, and display the front, back, left and right four-way video and surround video on the UI interface; the remote driving controller receives and processes environmental information and forwards it to the remote test bench through the 4G / 5G network; the remote cockpit side includes a driving simulator, a computer host and a display screen.

[0092] In order to ensure the safety and reliability of unmanned testing, the specific security strategy is as follows:

[0093] Preventive safety strategies for remote driving:

[0094] (1) Remote driving system self-test: The self-test information results are displayed on the remote test bench display, including the camera video streaming status display; network status monitoring, displaying the real-time network delay status (millisecond level); whether the vehicle is ready and in a takeover state. If abnormal, remote control of the vehicle is not allowed.

[0095] (2) Safety mechanism to prevent accidental touch: To remotely take over the autonomous driving on the test bench side, the control enable key needs to be started. Only after pressing the takeover button can the vehicle side switch from autonomous driving to remote driving. To remotely start the autonomous driving on the test bench side, the control enable key needs to be started. Only after pressing the CAN enable key can the vehicle side autonomous driving be activated.

[0096] Safety strategies during remote driving takeover:

[0097] (1) Limiting the remote driving speed: The maximum speed during remote driving is limited by the knob on the test bench. It is not affected by the remote accelerator pedal opening. That is, when the pedal is pressed to the bottom, the maximum speed remains unchanged. If the speed exceeds the limit, the vehicle will automatically brake.

[0098] (2) Remote driving network delay or disconnection: Network delay during remote driving causes video freezes and screen distortion. If the current video interface freezes and delays for less than 200ms, the vehicle will travel at a speed limit of 10km / h. If the delay exceeds 200ms or even disconnects, the vehicle will trigger automatic braking and parking. At this time, the remote control is in a connected state but cannot be controlled, and the UI will display a network delay warning. The vehicle can be controlled normally after the network is restored. The left and right rearview cameras will drive normally if there is a short freeze delay of less than 200ms. If the connection is disconnected, the speed limit will not exceed 10km / h. If it does not recover for more than 30s, the vehicle will not stop and the remote control is connected but cannot be controlled.

[0099] (3) Remote driving camera failure: If the camera itself fails, resulting in abnormal video streaming, the video screen will be black, and the current view will be black. At this time, the vehicle side triggers emergency braking. The remote control is connected but cannot control the vehicle. If the left, right, or rear video streaming is abnormal, the vehicle speed is limited to 10 km / h. If one or two of the above video streaming is abnormal, the surround view will display other videos normally, and all camera failure information will be displayed on the long UI interface to warn the driver.

[0100] (4) Automatic driving abnormal risk takeover: If the automatic driving has an abnormality, fails to identify or bypasses, and there is a risk of collision, the remote test bench can directly step on the brake pedal to control the vehicle to stop quickly, that is, there is no need to start the takeover and shift gears, thus reducing the reaction time.

[0101] (5) Platform system work order prompt: The platform receives autonomous driving fault diagnosis and perception information through information reception. Through logical judgment, it determines that dangerous scenarios may require takeover and issues different levels of warning work orders to the remote cockpit UI interface to prompt the remote driver whether to take over. Red and orange work orders require takeover, and other non-dangerous faults are only displayed as prompts.

[0102] (6) If a system crash occurs, the vehicle cannot be controlled remotely, or the autonomous driving is abnormal, the nearest test engineer will be contacted for a quick response and on-site investigation and handling.

[0103] In this embodiment, the present invention provides a computer-readable storage medium, on which is stored a computer program programmed or configured to execute any one of the safety assessment methods for unmanned testing of remote driving systems.

[0104] Any reference to memory, storage, database or other media used in the embodiments provided in the present application may include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), 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).

[0105] In summary, the present invention not only formulates effective safety strategies for possible risk issues that may arise in the remote driving system, but can also ensure the safety of vehicle driving before and after remote driving takes over, and can also intervene in time when abnormalities occur in the automatic driving to ensure safety.

[0106] The above specific implementations do not constitute a limitation on the protection scope of the present disclosure. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions can be made according to design requirements and other factors. Any modification, equivalent substitution and improvement made within the spirit and principle of the present disclosure shall be included in the protection scope of the present disclosure.

Claims

1. A safety assessment method for unmanned testing of a remote driving system, characterized in that: The method comprises: R1. When the remote driving system takes over the autonomous driving vehicle, it collects data information about the video streaming status and network monitoring status of the remote driving end, and obtains data information about the fault diagnosis of the vehicle and the surrounding environment of the autonomous driving end in real time; R2. Based on the data information of the video streaming status of the remote driving terminal and the data information of the network monitoring status, the BP neural network regression prediction algorithm based on fuzzy rules is used to predict the initial takeover status of the remote driving terminal, and the data information of the initial takeover status of the remote driving terminal after the prediction is obtained; R3. Based on the predicted data information of the preliminary takeover state of the remote driving end, the data information of the fault diagnosis of the vehicle on the automatic driving end, and the data information of the surrounding environment, a remote driving takeover safety situation prediction model based on an improved recursive neural network is constructed to predict the safety situation of the remote driving takeover, and obtain the predicted data information of the safety situation of the remote driving takeover; R4. Based on the data information of the predicted safety situation of the remote driving takeover, a safety assessment function G of the remote driving takeover is constructed to evaluate the safety of the remote driving takeover, and obtain data information of the safety assessment of the remote driving takeover.

2. The safety assessment method for unmanned testing of a remote driving system according to claim 1, characterized in that: The method further comprises: R5. Based on the data information of the safety assessment of the remote driving takeover, a preset threshold is set. If the safety assessment of the remote driving takeover is less than the preset threshold, the requirement is not met and remote takeover cannot be performed. If the safety assessment of the remote driving takeover is greater than the preset threshold, the requirement is met and remote takeover of the autonomous driving vehicle is performed.

3. The safety assessment method for unmanned testing of a remote driving system according to claim 1, characterized in that: In step R2, the use of the BP neural network regression prediction algorithm based on fuzzy rules to predict the initial takeover state of the remote driving terminal includes: R21. Based on the data information of the video streaming status of the remote driving terminal and the data information of the network monitoring status, a remote driving terminal state fuzzy rule function Q is established. Among them, x1 is the data information of the video streaming status of the remote driving end, x2 is the data information of the network monitoring status of the remote driving end, α1, α2 and α3 are the fuzzy rule factors of the status of the remote driving end, and the fuzzy rule matrix of the status of the remote driving end is characterized to obtain the data information of the fuzzy rule matrix of the status of the remote driving end; R22. Input the data information of the fuzzy rule matrix of the state of the remote driving terminal into the BP neural network regression prediction model for training and learning, and determine the prediction function W of the model, Among them, y is the data information of the fuzzy rule matrix of the state of the remote driving end, β1, β2 and β3 are the penalty factors of the model prediction, and the trained BP neural network regression prediction model is obtained; R23. Based on the trained BP neural network regression prediction model, the data information of the fuzzy rule matrix of the state of the remote driving end is input, and the initial takeover state of the remote driving end is predicted to obtain the data information of the predicted initial takeover state of the remote driving end.

4. The safety assessment method for unmanned testing of a remote driving system according to claim 3 is characterized in that: The penalty factors β1, β2 and β3 of the fuzzy prediction are, Among them, y is the data information of the fuzzy rule matrix of the state of the remote driving end.

5. The safety assessment method for unmanned testing of a remote driving system according to claim 3, characterized in that: The fuzzy rule factors α1, α2 and α3 of the state of the remote driving terminal are, Among them, x1 is the data information of the video streaming status of the remote driving end, and x2 is the data information of the network monitoring status of the remote driving end.

6. The safety assessment method for unmanned testing of a remote driving system according to claim 1, characterized in that: In step R3, the construction of a remote driving takeover safety situation prediction model based on an improved recursive neural network to predict the safety situation of remote driving takeover includes: R31. Based on the predicted data information of the initial takeover state of the remote driving end, the data information of the fault diagnosis of the vehicle of the automatic driving end and the data information of the surrounding environment, a time series function P of a recursive neural network is established. Among them, z1 is the data information of the initial takeover state of the remote driving end after prediction, z2 is the data information of the fault diagnosis of the vehicle on the automatic driving end, z3 is the data information of the surrounding environment of the vehicle on the automatic driving end, δ1, δ2 and δ3 are the weight factors of the time series of the recurrent neural network, which characterize the time series of the remote driving end taking over the vehicle, and obtain the data information of the time series of the remote driving end taking over the vehicle; R32. Input the time series data information of the remote driving end taking over the vehicle into the remote driving takeover safety situation prediction model for training and learning, and determine the remote driving takeover safety situation prediction function S, Among them, r is the time series data information of the remote takeover vehicle, γ1, γ2 and γ3 are the reward factors of the remote driving takeover safety situation, and the trained remote driving takeover safety situation prediction model is obtained; R33. Based on the trained remote driving takeover safety situation prediction model, the time series data information of the remote driving end taking over the vehicle is input to predict the safety situation of the remote driving takeover, and the data information of the predicted safety situation of the remote driving takeover is obtained.

7. The safety assessment method for unmanned testing of a remote driving system according to claim 6, characterized in that: The constraints of the reward factors γ1, γ2 and γ3 for the remote driving to take over the safety situation are:

8. The safety assessment method for unmanned testing of a remote driving system according to claim 1, characterized in that: The safety evaluation function G of the remote driving takeover is: Among them, h is the data information of the predicted safety situation of remote driving takeover, and η1, η2 and η3 are the safety assessment factors of remote driving takeover.

9. A safety assessment system for unmanned testing of a remote driving system, comprising a computer device, characterized in that: The computer device is programmed or configured to execute the steps of the safety assessment method for unmanned testing of a remote driving system as described in any one of claims 1 to 8.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program that is programmed or configured to execute the safety assessment method for unmanned testing of a remote driving system as described in any one of claims 1 to 8.

Images