Evaluation system and method before control right transfer of autonomous vehicle

Abstract

The invention relates to the field of automatic driving, and discloses an evaluation system and method before control right transfer of an automatic driving vehicle, and the system comprises a detection module which uses an in-vehicle sensor to monitor the activity of a driver before a handover command is issued; the calibration task module is used for generating an instruction for prompting a driver to execute a specific vehicle control operation task; the evaluation module is used for evaluating the response time and accuracy of the driver in the calibration task; the control right handover module determines whether to carry out control right handover according to the evaluation result; the vehicle outputs a corresponding calibration task module to a driver according to information acquired by the detection module, and then the evaluation module evaluates response time and accuracy when the driver executes the calibration task module and outputs an evaluation result to the control right handover module; and the control right handover module executes a handover action or a non-handover action according to an evaluation result of the evaluation module. The system has the advantages of high reliability, high efficiency and the like.

Description

Technical Field

[0001] The present invention relates to the field of autonomous driving, and in particular to an evaluation system and method for an autonomous driving vehicle before handing over control. Background Art

[0002] Advances in autonomous vehicle technology have enabled vehicles to operate without constant driver intervention. However, there are situations where control needs to be handed over to the driver. Studies have shown that drivers take different amounts of time to regain situational awareness, depending on their previous activities. For example, a sleeping driver may take longer to fully wake up than a driver who is only watching the road. For example, CN115140092A discloses a personalized transfer method for human-machine control authority of a co-driving intelligent vehicle, which provides some ideas for transferring control, but its failure to ensure the driver's readiness may lead to delayed response, thereby increasing the risk of accidents. Therefore, there is an urgent need for a system that can not only assess the driver's readiness, but also actively help the driver regain situational awareness before handing over control. Summary of the invention

[0003] In view of the shortcomings of the prior art, the present invention provides an evaluation system for an autonomous driving vehicle before handing over control.

[0004] An evaluation system before an autonomous driving vehicle transfers control rights, comprising a detection module, a calibration task module, an evaluation module, and a control rights transfer module;

[0005] The detection module uses in-vehicle sensors to monitor the driver's activities before issuing a handover command;

[0006] A calibration task module generates instructions to prompt the driver to perform specific vehicle control operation tasks;

[0007] An evaluation module that assesses the driver’s response time and accuracy in the calibration task;

[0008] The control transfer module determines whether to transfer control rights based on the evaluation results;

[0009] The vehicle outputs the corresponding calibration task module to the driver based on the information obtained by the detection module. Then the evaluation module evaluates the driver's response time and accuracy when executing the calibration task module and outputs the evaluation result to the control right handover module. The control right handover module executes the handover action or does not execute the handover action based on the evaluation result of the evaluation module.

[0010] Preferably, a secondary calibration task module is also included. After failing to execute the driving right handover command, the control right handover module transmits a signal to the secondary calibration task module for secondary task calibration. The secondary calibration task module is communicatively connected with the evaluation module.

[0011] Preferably, the detection module includes one or more of a head pitch angle detection unit, a line of sight deviation detection unit, a hand grip strength mean detection unit, a heart rate variability detection unit, a breathing rate detection unit, a voice response delay detection unit, and a lower limb activity frequency detection unit.

[0012] Preferably, the driver's activities are divided into deep sleep, light sleep, reading, video interaction, and observing road conditions based on the information collected by the sensors of the detection module.

[0013] Preferably, the calibration task module matches a corresponding calibration task unit according to the driver activity type output by the detection module, and the calibration task unit includes a steering wheel task and a pedal task;

[0014] The steering wheel task outputs signals through the steering motor torque sensor and gyroscope; the pedal task outputs signals through the brake pressure sensor and pedal displacement sensor, as well as the throttle opening sensor and IMU accelerometer sensor.

[0015] As a preference, the method comprises the following steps:

[0016] Step 1: The vehicle is in the automatic driving state, and the driver makes a request to transfer control;

[0017] Step 2: The detection module of the system evaluates the current state of the driver, and the detection module includes a head pitch angle detection unit, a sight deviation detection unit, a hand grip force mean detection unit, a heart rate variability detection unit, a breathing rate detection unit, a voice response delay detection unit, and a lower limb activity frequency detection unit; the detection unit outputs the current state of the driver according to the above detection signals, and the state is divided into deep sleep, light sleep, reading, video interaction, and observing road conditions;

[0018] Step 3: The system starts the calibration task module, and the calibration task module sets different calibration task units corresponding to different states; the driver performs the calibration task according to the instructions of the calibration task unit;

[0019] Step 4: System evaluation evaluates the driver's performance when executing the calibration task module and transmits it to the control handover module in the form of a signal;

[0020] Step 5: The control right handover module decides whether to hand over the control right or continue the secondary calibration according to the signal output by the evaluation module.

[0021] Preferably, the calibration task module matches a corresponding calibration task unit according to the driver activity type output by the detection module, and the calibration task unit includes a steering wheel task and a pedal task;

[0022] The steering wheel task outputs signals through the steering motor torque sensor and gyroscope; the pedal task outputs signals through the brake pressure sensor and pedal displacement sensor, as well as the throttle opening sensor and IMU accelerometer sensor.

[0023] Compared with the prior art, this solution has the following beneficial effects: The present invention provides a practical solution to the driver readiness problem in the handover of control of an autonomous vehicle. By introducing a vehicle control device calibration task that requires the driver's active participation, the system effectively reawakens the driver, improves safety, and reduces the risk of delayed response in the handover of control. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Figure 1 The control diagram of the system. DETAILED DESCRIPTION

[0025] The present invention is further described in detail below in conjunction with the accompanying drawings and embodiments.

[0026] Example 1

[0027] An evaluation system before an autonomous driving vehicle transfers control rights, comprising a detection module, a calibration task module, an evaluation module, and a control rights transfer module;

[0028] The detection module uses in-vehicle sensors to monitor the driver's activities before issuing a handover command;

[0029] A calibration task module generates instructions to prompt the driver to perform specific vehicle control operation tasks;

[0030] An evaluation module that assesses the driver’s response time and accuracy in the calibration task;

[0031] The control transfer module determines whether to transfer control rights based on the evaluation results;

[0032] The vehicle outputs the corresponding calibration task module to the driver based on the information obtained by the detection module. The evaluation module then evaluates the driver's response time and accuracy when executing the calibration task module and outputs the evaluation result to the control right handover module. The control right handover module executes a handover action or a non-handover action based on the evaluation result of the evaluation module. Executing a non-handover action will perform a secondary task calibration, and the vehicle is still driven by the automatic driving system.

[0033] This solution also includes a secondary calibration task module. After not executing the driving right handover command, the control right handover module transmits a signal to the secondary calibration task module for secondary task calibration. The secondary calibration task module is communicatively connected to the evaluation module. The secondary calibration task module can select the previous first calibration task or a separate calibration task.

[0034] The detection module includes a head pitch angle detection unit - obtained through an RGB-D camera, unit degree, a line of sight deviation detection unit - the horizontal deviation (pixel) between the pupil center and the windshield baseline, a hand grip mean detection unit - steering wheel pressure sensor data, unit Newton, a heart rate variability detection unit - based on a wristband PPG sensor, unit ms, a breathing rate detection unit - a seat-built-in piezoelectric sensor, unit times / minute, a voice response delay detection unit - the time from the system prompt to the driver's first utterance, unit seconds, a lower limb activity frequency detection unit - the number of times the acceleration / brake pedal is touched, unit times / minute. In this solution, all of the above detection units are set.

[0035] In this embodiment, the information collected by the sensors of the detection module is used to classify the driver's activities into deep sleep, light sleep, reading, video interaction, and road condition observation, wherein each activity state will be output with its own unique signal.

[0036] The calibration task module matches the corresponding calibration task unit according to the driver activity type output by the detection module, wherein the task duration and type of the task unit can be calibrated in advance by means of calibration, or the task can be established in real time according to the collected signal. The calibration task unit includes a steering wheel task and a pedal task, wherein the pedal task is further divided into a brake pedal task and an accelerator pedal task;

[0037] The steering wheel task outputs signals through the steering motor sensor and gyroscope; the pedal task outputs signals through the brake pressure sensor and pedal displacement sensor, as well as the throttle opening sensor and IMU accelerometer sensor.

[0038] Example 2

[0039] This embodiment discloses, based on the first embodiment, an evaluation system for an autonomous driving vehicle before handing over control, including the following steps:

[0040] Step 1: The vehicle is in the automatic driving state, and the driver makes a request to transfer control;

[0041] Step 2: The detection module of the system evaluates the current state of the driver, and the detection module includes a head pitch angle detection unit, a sight deviation detection unit, a hand grip force mean detection unit, a heart rate variability detection unit, a breathing rate detection unit, a voice response delay detection unit, and a lower limb activity frequency detection unit; the detection unit outputs the current state of the driver according to the above detection signals, and the state is divided into deep sleep, light sleep, reading, video interaction, and observing road conditions;

[0042] Step 3: The system starts the calibration task module, and the calibration task module sets different calibration task units corresponding to different states; the driver performs the calibration task according to the instructions of the calibration task unit;

[0043] Step 4: System evaluation evaluates the driver's performance when executing the calibration task module and transmits it to the control handover module in the form of a signal;

[0044] Step 5: The control right handover module decides whether to hand over the control right or continue the secondary calibration according to the signal output by the evaluation module.

[0045] In this embodiment, the previous activity is defined as deep sleep, the subsequent steering wheel rotation angle threshold is ±40 degrees, the pedal pressure range is 25-35N, and the longest task is 6s; in light sleep, the steering wheel rotation angle threshold is ±30 degrees, the pedal pressure range is 15-25N, and the longest task is 3.5s; during video interaction, the steering wheel rotation angle threshold is ±20 degrees, the pedal pressure range is 10-20N, and the maximum time window allowed for response is 2.8s; when reading, the steering wheel rotation angle threshold is 15 degrees, the pedal pressure is 8-15N, and the maximum response time is 2.2s.

[0046] Progressive task chain design:

[0047] When the driver fails the first mission, the "stepped wake-up protocol" is started:

[0048] Level 1: Single steering wheel fine adjustment (±5°) + light braking (5N pressure)

[0049] Level 2: Three consecutive direction corrections (±10° alternating) + braking / acceleration alternating operations

[0050] Level 3: Simulates emergency lane change scenarios (90° steering + 80% braking force).

[0051] The calibration task module matches the corresponding calibration task unit according to the driver activity type output by the detection module, and the calibration task unit includes a steering wheel task and a pedal task;

[0052] The steering wheel task outputs signals through the steering motor torque sensor and gyroscope; the pedal task outputs signals through the brake pressure sensor and pedal displacement sensor, as well as the throttle opening sensor and IMU accelerometer sensor;

[0053] The task types are:

[0054]

[0055] The dynamic optimization model of the calibration task unit is:

[0056] The objective function is:

[0057] Output value: the optimal solution x of the function * is a multidimensional vector representing the optimal combination of the following key control parameters:

[0058] x * =[task sequence, single task duration allocation, control input curve] T ,

[0059]

[0060]

[0061] Example of objective function output:

[0062] Assume that the optimized output is:

[0063]

[0064] At this time, the objective function value is:

[0065]

[0066] The comprehensive evaluation objective function value of system efficiency integrates two key dimensions: energy consumption and human-machine collaboration error, reflecting the comprehensive efficiency of the system within a limited time window:

[0067]

[0068] Physical meaning of energy consumption items: including mechanical energy (power consumption of steering / brake motor) and cognitive energy (mental workload of the driver to complete the prompt task), typical value: about 3.2kJ for a single steering task, about 0.9kJ for a voice confirmation task (based on EEG cognitive load model), optimization direction: reduce this value → reduce total system energy consumption, extend battery life, and reduce driver fatigue Physical meaning of human-machine collaboration error item: the degree of match between the driver's actual action and the ideal trajectory (angle error, pressure error, etc.) Safety margin: system intervention is triggered when the steering wheel error is >2° or the braking force error is >5N

[0069] Optimization direction: Reduce this value → Improve the smoothness and safety of control transfer.

[0070] The reliability indicator objective function value of control transfer is directly related to the takeover success rate and emergency response capability, for example:

[0071] Target value < 5.0 → Meets ASIL-D safety requirements, allowing full transfer of control

[0072] 5.0≤Target value<7.5→Auxiliary stability control needs to be activated (such as ESP intervention to compensate for understeer)Target value≥7.5→Trigger emergency braking (AEB) and cancel control transfer.

[0073] Examples:

[0074] When the tire adhesion coefficient decreases due to heavy rain, the target value will be humax -θ target || 2 If the calculated value exceeds the 7.5 threshold, it means that the current environment has exceeded the compensation ability of human drivers, and the system will take over forcibly.

[0075] Preferably, the maximum steering wheel angle is -45° to +45°, and the calculation model for the duration of the steering wheel steering task in a single calibration task unit is:

[0076]

[0077] Duration is the duration, the basic time is t base Obtained from the table, the default basic time is 2s, the EEG θ / β wave ratio reflects the degree of neural arousal, and the environmental complexity env complexity ∈[0,1] is provided by the perception system.

[0078] As a preference, the error term ||θ human -θ target || 2 The following conditions must be met: steering wheel angle error ≤ 2, braking force error ≤ 5N. Exceeding this threshold will trigger task sequence reconstruction.

[0079] Also includes the accuracy index calculation model:

[0080]

[0081] in:

[0082] X1 = Standardized response time (base value 2.5 seconds),

[0083] X2 = Operation accuracy score (steering wheel angle error <5° is 1.0, and 0.2 is deducted for each additional 1°).

[0084] X3 = physiological recovery coefficient (calculated based on EEG β / θ wave power ratio).

[0085] The system presets RI ≥ 85 points as a passing score.

[0086] The calibration task also includes activating the haptic feedback steering wheel.

[0087] Example:

[0088] Precondition: The vehicle is about to exit the highway and the driver requests to switch to manual driving mode.

[0089] Detection phase:

[0090] The millimeter-wave radar detected that the driver's heart rate dropped to 55bpm (the baseline for normal driving conditions is 72bpm±5).

[0091] The infrared camera captured the eyes closed for more than 2 minutes.

[0092] The system determines that the state is "deep sleep", and the estimated recovery time benchmark value is T 0 =9 seconds.

[0093] Calibration task generation:

[0094] Activate haptic feedback steering wheel (applies 5Hz vibration),

[0095] The HUD displays "Please turn the steering wheel 30° to the left and hold for 2 seconds".

[0096] At the same time, it is required to press the brake pedal to the 20N threshold of the pressure sensor.

[0097] Evaluation logic:

[0098] If the response time is ≤3.5 seconds and the steering wheel angle error is <5°, and the RI is ≥85 points, control is transferred.

[0099] If the standard is not met, the Level 2 task chain is initiated: alternating steering ±15° three times + brake / accelerator pedal pulse operation until the handover requirements are met.

[0100] Example 3

[0101] The difference between this embodiment and the above embodiment is that:

[0102] Compared with the prior art, this solution has the following beneficial effects: The present invention provides a practical solution to the driver readiness problem in the handover of control of an autonomous vehicle. By introducing a vehicle control device calibration task that requires the driver's active participation, the system effectively reawakens the driver, improves safety, and reduces the risk of delayed response in the handover of control.

Claims

1. An evaluation system for an autonomous driving vehicle before handing over control, characterized in that: It includes detection module, calibration task module, evaluation module and control right handover module; The detection module uses in-vehicle sensors to monitor the driver's activities before issuing a handover command; A calibration task module generates instructions to prompt the driver to perform specific vehicle control operation tasks; An evaluation module that assesses the driver’s response time and accuracy in the calibration task; The control transfer module determines whether to transfer control rights based on the evaluation results; The vehicle outputs the corresponding calibration task module to the driver based on the information obtained by the detection module. The evaluation module then evaluates the driver's response time and accuracy when executing the calibration task module and outputs the evaluation result to the control right handover module. The control right handover module executes the handover action or does not execute the handover action based on the evaluation result of the evaluation module.

2. The evaluation system for an autonomous driving vehicle before handing over control according to claim 1, characterized in that: It also includes a secondary calibration task module. After the control right handover module does not execute the driving right handover command, it transmits a signal to the secondary calibration task module for secondary task calibration. The secondary calibration task module is communicatively connected with the evaluation module.

3. The evaluation system for an autonomous driving vehicle before handing over control according to claim 1, characterized in that: The detection module includes one or more of a head pitch angle detection unit, a line of sight deviation detection unit, a hand grip force mean detection unit, a heart rate variability detection unit, a breathing rate detection unit, a voice response delay detection unit, and a lower limb activity frequency detection unit.

4. The evaluation system for an autonomous driving vehicle before handing over control according to claim 3, characterized in that: Based on the information collected by the sensors in the detection module, the driver's activities are divided into deep sleep, light sleep, reading, video interaction, and observing road conditions.

5. The evaluation system for an autonomous driving vehicle before handing over control according to claim 4, characterized in that: The calibration task module matches the corresponding calibration task unit according to the driver activity type output by the detection module, and the calibration task unit includes a steering wheel task and a pedal task; The steering wheel task outputs signals through the steering motor torque sensor and gyroscope; The pedal task outputs signals through the brake pressure sensor and the pedal displacement sensor, as well as the throttle opening sensor and the IMU accelerometer sensor.

6. A method for evaluating an autonomous driving vehicle before handing over control, characterized in that: The following steps are involved: Step 1: The vehicle is in the automatic driving state, and the driver makes a request to transfer control; Step 2: The detection module of the system evaluates the current state of the driver, and the detection module includes a head pitch angle detection unit, a sight deviation detection unit, a hand grip force mean detection unit, a heart rate variability detection unit, a breathing rate detection unit, a voice response delay detection unit, and a lower limb activity frequency detection unit; the detection unit outputs the current state of the driver according to the above detection signals, and the state is divided into deep sleep, light sleep, reading, video interaction, and observing road conditions; Step 3: The system starts the calibration task module, and the calibration task module sets different calibration task units corresponding to different states; the driver performs the calibration task according to the instructions of the calibration task unit; Step 4: System evaluation evaluates the driver's performance when executing the calibration task module and transmits it to the control handover module in the form of a signal; Step 5: The control right handover module decides whether to hand over the control right or continue the secondary calibration according to the signal output by the evaluation module.

7. The method for evaluating an autonomous driving vehicle before handing over control according to claim 6, characterized in that: The calibration task module matches the corresponding calibration task unit according to the driver activity type output by the detection module, and the calibration task unit includes a steering wheel task and a pedal task; The steering wheel task outputs signals through the steering motor torque sensor and gyroscope; the pedal task outputs signals through the brake pressure sensor and pedal displacement sensor, as well as the throttle opening sensor and IMU accelerometer sensor; The task types are: The objective function of the dynamic optimization model of the calibration task unit is: Output value: the optimal solution x of the function * is a multidimensional vector representing the optimal combination of the following key control parameters: x * =[task sequence, single task duration allocation, control input curve] T , 8. The method for evaluating an autonomous driving vehicle before handing over control according to claim 7, characterized in that: The maximum steering wheel angle is -45° to +45°. The calculation model for the duration of the steering wheel steering task in a single calibration task unit is: Duration is the duration, the basic time t base Obtained from the table, the default basic time is 2s, the EEG θ / β wave ratio reflects the degree of neural arousal, and the environmental complexity env complexity ∈[0,1] is provided by the perception system.

9. The method for evaluating an autonomous driving vehicle before handing over control according to claim 7, characterized in that: Error term ||θ human -θ target || 2 The following conditions must be met: steering wheel angle error ≤ 2, braking force error ≤ 5N. Exceeding this threshold will trigger task sequence reconstruction.

Images