A safety processing method for the collision risk of autonomous driving

Abstract

The present invention relates to a safety processing method for the collision risk of autonomous driving. The method includes: P1: The vehicle independently receives perception, positioning, and map information; P2: Based on the above information, the perception module of the vehicle screens dangerous targets for perception of targets to obtain potential dangerous targets; P3: Conduct threat judgment on the potential dangerous targets, and based on the decision-making algorithm, delay control algorithm, adjusted decision threshold, and emergency braking algorithm, output the decision result corresponding to the emergency situation; P4: Output the decision result corresponding to the emergency situation to the control module of the vehicle. The present invention analyzes the safety processing mechanism when the perception information is abnormal and the decision-making algorithm fails, and maximally ensures driving safety.

Description

Technical Field

[0001] The present invention relates to the field of automatic collision avoidance technology, and in particular to a method for safely handling collision risks in automatic driving. Background Art

[0002] With the continuous development of science and technology, cars have become an indispensable means of transportation in our lives, and the car's automatic driving technology has also been slowly integrated into our lives. Automatic emergency braking is an active safety technology that measures the distance to the vehicle in front or obstacles, and then uses the data analysis module to compare the measured distance with the alarm distance and safety distance. When the system detects that a collision is very likely to occur, it will actively brake to reduce the vehicle speed and avoid the collision as much as possible. Manual takeover is used as a safeguard in emergency situations or when the intelligent driving system sensor sends a serious failure or misjudgment.

[0003] The commonly used active collision avoidance scheme is to continuously detect the braking condition of the vehicle's brake pedal. If the brake pedal is braked quickly, it is inferred that there is a collision risk. The system detects the position and movement of objects around the vehicle relative to the vehicle and evaluates the collision risk. When the system determines that braking alone cannot avoid the collision but steering avoidance can avoid the collision, the system takes over the steering wheel to implement active avoidance operations. The system is prone to misjudgment by inferring the collision risk through rapid braking of the brake pedal; the driver's operation of the brake pedal makes it difficult for the system to estimate the change in vehicle speed, and the actual avoidance trajectory of the vehicle may deviate greatly from the theoretical trajectory, and the reliability of collision avoidance is low. Therefore, finding a method to solve the above problems is an urgent problem that we need to solve. Summary of the invention

[0004] In view of the above deficiencies in the prior art, the present invention provides a safe handling method for autonomous driving collision risks, which not only analyzes the safety handling mechanism when perception information is abnormal and decision-making algorithm fails, but also ensures driving safety to the greatest extent.

[0005] In order to achieve the above-mentioned purpose and other related purposes, the technical solution provided by the present invention is as follows: A method for safely handling collision risks of autonomous driving, comprising the following steps:

[0006] P1: The vehicle independently receives perception, positioning, and map information;

[0007] P2: Based on the above information, the perception module of the vehicle senses the target and screens the dangerous target to obtain the potential dangerous target;

[0008] P3: Perform threat judgment on the potential dangerous target, and output the decision result corresponding to the emergency situation based on the decision algorithm, delay control algorithm, adjustment of decision threshold and emergency braking algorithm;

[0009] P4: Output the decision result corresponding to the emergency situation to the control module of the vehicle.

[0010] Further, in step P2, when the perception information of the vehicle's perception module cannot be obtained, the following processing methods are included:

[0011] f1. Determine that the perception module cannot work properly, and the vehicle performs emergency braking and the HMI broadcasts.

[0012] f2. Determine that the perception module can work properly. Based on the previous frame of the vehicle's global trajectory, it is concluded that the perception module has lost multiple frames and cannot send perception data. As a result, the vehicle decelerates and stops and the HMI broadcasts.

[0013] Further, in step P3, when accurate positioning, map, and perception information cannot be obtained, the screening of the potential dangerous targets cannot be performed, and the following processing methods are included:

[0014] a1. When the positioning signal is unstable, resulting in a low positioning accuracy and a small error judgment, the vehicle adaptively adjusts based on the decision algorithm threshold and the HMI broadcasts; or when the error judgment is large, the vehicle decelerates and stops and the HMI broadcasts.

[0015] a2. When the positioning signal is unstable and the positioning signal is lost, based on the previous frame of data of the vehicle, it is concluded that the vehicle's positioning sensor cannot work properly or the positioning algorithm cannot work properly, and the vehicle decelerates and stops and the HMI broadcasts.

[0016] a3. When the perception signal is unstable, for the input of the perception accuracy error judgment, the input of the perception missed detection situation, and the input of the perception type error judgment of the vehicle, if the output error is allowed, the vehicle adaptively adjusts based on the decision algorithm threshold and the HMI broadcasts; or if the error is large, the vehicle decelerates and stops and the HMI broadcasts.

[0017] Further, in step P4, the decision failure records of the decision result corresponding to the emergency situation include: missed detection and error of static obstacles, missed detection and error of braking obstacles, and missed detection and error of pedestrian blind areas; when the adjustment of the decision algorithm for obtaining the decision result is too large, resulting in an input error exceeding the decision threshold, the vehicle decelerates and stops and the HMI broadcasts; when the delay of the decision result exceeds the threshold, the recorded data is output and waiting for the engineer to modify, and as a result, the vehicle decelerates and stops and the HMI broadcasts.

[0018] Further, the emergency braking algorithm includes:

[0019] d1. Based on the criterion of two vehicles moving in the same direction or in the same lane, obtain the minimum distance d for emergency braking min :

[0020] ,

[0021] where d min is the minimum vehicle distance, a min,brake is the minimum acceleration, and the vehicle speed v r The maximum acceleration a within the reaction time ρ max,accel , and the vehicle speed v f The maximum acceleration a at max,brake ;

[0022] d2. Based on the lane change of two vehicles, the minimum distance d for emergency braking is obtained min :

[0023] ,

[0024] where μ is the distance between the doors of the two vehicles, v1 and v2 are the speeds of the two vehicles, and v 1,ρ , v 2,ρ are the speeds of the two vehicles at the reaction time ρ respectively, and a lat min,brake is the braking acceleration of the two vehicles.

[0025] Furthermore, the delay control algorithm includes the following steps:

[0026] Q1. Obtain multiple detection sequences of the potential dangerous targets, and each detection sequence of the potential dangerous targets includes the target state data of multiple target objects;

[0027] Q2. Based on the multiple detection sequences of the potential dangerous targets with time stamps, predict the target state data in the j detection sequences of the potential dangerous targets to obtain the (j + 1)-th detection sequence of the dangerous targets and the corresponding target state data;

[0028] Q3. Based on the actual vehicle test in the actual scenario, obtain the target state data corresponding to the (j + 1)-th detection sequence of the potential dangerous targets;

[0029] Q4. Remove duplicates from the target state data corresponding to the detection sequences of the potential dangerous targets obtained in steps Q2 and Q3, eliminate the data errors caused by calculation and signal delay, and fuse them to obtain the (j + 1)-th fused potential dangerous target sequence.

[0030] The present invention has the following positive effects:

[0031] The present invention proposes an emergency active algorithm model for an autonomous vehicle in two common scenarios: driving in its own lane and changing lanes. According to the scenario and the dangerous situation, the actions taken by the vehicle are to change the input of the RSS formula, output a normal parking instruction, and directly apply maximum braking. Finally, the safety handling mechanism in case of abnormal perception information and failure of the decision-making algorithm is analyzed to ensure driving safety to the greatest extent. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] Figure 1 It is a flowchart of the algorithm of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0033] In order to make the objectives, technical solutions and advantages of the present invention more clearly understood, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0034] Embodiment: A safety handling method for autonomous driving collision risk, comprising the following steps:

[0035] P1: The vehicle independently receives perception, positioning, and map information.

[0036] P2: Based on the above information, the perception module of the vehicle screens dangerous targets to obtain potential dangerous targets.

[0037] P3: Perform a threat judgment on the potential dangerous targets, and output the decision result corresponding to the emergency situation based on the decision-making algorithm, delay control algorithm, adjusted decision threshold, and emergency braking algorithm.

[0038] P4: Output the decision result corresponding to the emergency situation to the control module of the vehicle.

[0039] Further, in step P2, when the perception information of the perception module of the vehicle cannot be obtained, the following processing methods are included:

[0040] f1. Determine that the perception module cannot work properly, and the vehicle performs emergency braking and the HMI broadcasts.

[0041] f2. Determine that the perception module can work properly, and based on the previous frame of the global trajectory of the vehicle, it is concluded that the perception module has lost multiple frames and cannot send perception data, so the vehicle decelerates and stops and the HMI broadcasts. Further, in step P3, when accurate positioning, map, and perception information cannot be obtained, the screening of the potential dangerous targets cannot be performed, and the following processing methods are included:

[0042] a1. When the positioning signal is unstable, resulting in a low positioning accuracy and a small error judgment, the vehicle adaptively adjusts based on the decision algorithm threshold and gives an HMI broadcast; or when the error judgment is large, the vehicle decelerates and stops and gives an HMI broadcast.

[0043] a2. When the positioning signal is unstable and the positioning signal is lost, based on the previous frame data of the vehicle, it is concluded that the positioning sensor of the vehicle cannot work properly or the positioning algorithm cannot work properly. The vehicle decelerates and stops and gives an HMI broadcast.

[0044] a3. When the sensing signal is unstable, for the input of the sensing accuracy error judgment, the input of the sensing missed detection situation, and the input of the sensing type error judgment of the vehicle, if the output error is allowed, the vehicle adaptively adjusts based on the decision algorithm threshold and gives an HMI broadcast; or if the error is large, the vehicle decelerates and stops and gives an HMI broadcast.

[0045] Further, in step P4, the decision failure records corresponding to the decision results in the emergency situation include: missed detection and error of static obstacles, missed detection and error of braking obstacles, and missed detection and error of pedestrian blind spots; when the adjustment of the decision algorithm for obtaining the decision result is too large, resulting in the input error exceeding the decision threshold, the vehicle decelerates and stops and gives an HMI broadcast; when the delay of the decision result exceeds the threshold, the recorded data is output for the engineer to modify, and thus the vehicle decelerates and stops and gives an HMI broadcast.

[0046] Further, the emergency braking algorithm includes:

[0047] d1. Based on the criterion of two vehicles moving in the same direction or in the same lane, the minimum distance d for emergency braking is obtained min :

[0048] ,

[0049] where d min is the minimum vehicle-limiting distance, a min,brake is the minimum acceleration, the vehicle speed is v r the maximum acceleration a within the reaction time ρ max,accel , the vehicle speed is v f the maximum acceleration a at max,brake ;

[0050] d2. Based on the lane change of two vehicles, the minimum distance d for emergency braking is obtained min :

[0051] ,

[0052] where μ is the distance between the doors of the two vehicles, v1, v2 are the speeds of the two vehicles, v 1,ρ , v 2,ρThey are the speeds of the two vehicles at the reaction time ρ, respectively, a lat min,brake are the braking accelerations of the two vehicles.

[0053] Furthermore, the delay control algorithm includes the following steps:

[0054] Q1. Obtain multiple detection sequences of the potential dangerous targets. Each detection sequence of the potential dangerous target includes target state data of multiple target objects;

[0055] Q2. Based on the multiple detection sequences of the potential dangerous targets with timestamps, predict the target state data in the j detection sequences of the potential dangerous targets to obtain the (j + 1)-th detection sequence of the dangerous target and the corresponding target state data;

[0056] Q3. Based on the real vehicle test in the actual scenario, obtain the target state data corresponding to the (j + 1)-th detection sequence of the potential dangerous target;

[0057] Q4. Remove duplicates from the target state data corresponding to the detection sequences of the potential dangerous targets obtained in steps Q2 and Q3, eliminate the data errors caused by calculation and signal delay, and fuse them to obtain the (j + 1)-th fused potential dangerous target sequence.

[0058] Specifically, as Figure 1 shown, the failure diagnosis mainly relies on the automatic information release at the sensor end, the input information state of the decision-making algorithm, and the dynamic prediction model to independently judge whether there is a delay or signal loss in the perception information. When the final perception input accuracy has no impact and the scenario does not pose a danger, the system remains unchanged. When the failure degree of the sensor has a certain impact on the accuracy and there is a possibility of collision, adjust the algorithm input amount or the corresponding threshold to ensure no collision occurs. When the input signal error is large, the decision calculation will be based on the most dangerous signal input within the high-confidence range, and the decision-making algorithm will be calculated with a conservative input signal. When the input signal error affects other normal decision-making behaviors of the emergency braking algorithm, ensure safe parking according to the environmental situation judged from the input results of the previous few frames. When there is a loss of the input signal or the input information is completely distorted, perform emergency braking. And output the failure situation to the HMI.

[0059] The emergency braking algorithm makes decision calculations based on lane, obstacle position, speed, and acceleration information to ensure no collision occurs. In terms of algorithm details, decision-making delay problems may be caused by computing power or computing methods. At the same time, problems such as the decision-making algorithm having no solution or not covering special scenarios due to special scenarios not being considered may lead to large deviations in the decision-making results or the non-occurrence of decision-making behaviors.

[0060] The failure diagnosis is based on the real-time monitoring of the input and output timestamps in the algorithm and the actual vehicle tests in the actual scenario. The decision delay time series is accurately output to the failure handling unit and tested in the actual scenario, and manual judgment is made to determine whether there are invalid scenarios. For invalid scenarios, the algorithm is improved in advance to eliminate uncovered scenarios. For the decision delay situation, combined with the perception delay information, three countermeasures are taken: adjusting the delay control algorithm, adjusting the decision threshold, and emergency braking, and an HMI prompt is provided. Within the controllable delay range, a hybrid strategy of delay control algorithm and threshold adjustment is adopted. First, the delay control algorithm is used to reduce the impact of delay signals, and at the same time, the emergency avoidance threshold is adjusted to ensure driving safety. When the delay amount poses a greater threat, emergency braking is performed. In complex scenarios, the problem of no solution or slow operation in the optimization algorithm solving process caused by the limitation of computing power or computing method is solved.

[0061] In summary, the present invention analyzes the safety handling mechanism when the perception information is abnormal and the decision algorithm fails, and maximally ensures driving safety.

[0062] The above is a specific description of the preferred embodiment of the present invention, but the present invention is not limited to the described embodiment. Those skilled in the art can make various equivalent deformations or substitutions without departing from the spirit of the present invention, and these equivalent deformations or substitutions are all included in the scope defined by the claims of this application.

Claims

1. A safety processing method for autonomous driving collision risk, characterized in that, Including the following steps: P1. The vehicle independently and dynamically receives perception, positioning, and map information; P2. Based on the above information, the perception module of the vehicle perceives the target to screen for dangerous targets to obtain potential dangerous targets; P3. Conduct a threat judgment on the potential dangerous targets. Based on the emergency collision algorithm and the delay control algorithm, set corresponding decision thresholds, and then output the decision results corresponding to the emergency situation according to the decision algorithm; P4. Output the decision results corresponding to the emergency situation to the control module of the vehicle; In step P3, the emergency braking algorithm includes: d1. Based on the criterion that two vehicles are moving in the same direction or in the same lane, the minimum distance d for emergency braking is obtained min , ， Among them, d min is the minimum vehicle distance, a min,brake is the minimum acceleration, the vehicle speed v r The maximum acceleration a within the reaction time ρ max,accel , the vehicle speed v f The maximum acceleration a under max,brake ; d2. Based on the lane change of two vehicles, the minimum distance d for emergency braking is obtained min , ， where μ is the distance between the two vehicle doors, v1 and v2 are the speeds of the two vehicles, and v 1,ρ , v 2,ρ are the speeds of the two vehicles at the reaction time ρ, and a lat min,brake is the braking acceleration of the two vehicles; In step P3, the delay control algorithm includes the following steps: Q1. Obtain multiple detection sequences of the potential dangerous targets. Each detection sequence of the potential dangerous targets includes the target state data of multiple target objects; Q2. Based on the multiple detection sequences of the potential dangerous targets with time stamps, predict the target state data in j detection sequences of the potential dangerous targets to obtain the (j + 1)-th detection sequence of the dangerous targets and the corresponding target state data; Q3. Based on the actual vehicle test in the actual scenario, obtain the target state data corresponding to the (j + 1)-th detection sequence of the potential dangerous targets; Q4. Perform duplicate removal processing on the target state data corresponding to the detection sequences of the potential dangerous targets obtained in steps Q2 and Q3, eliminate the data errors caused by calculation and signal delay, and fuse them to obtain the (j + 1)-th fused sequence of the potential dangerous targets.

2. The safety processing method for the automatic driving collision risk according to claim 1, characterized in that In step P2, when the perception information of the perception module of the vehicle cannot be obtained, the following processing methods are included: f1. Judge that the perception module cannot work properly, and the vehicle performs emergency braking and the HMI broadcasts; f2. Judge that the perception module can work properly. Based on the previous frame of the global trajectory of the vehicle, it is obtained that the perception module has lost multiple frames and cannot send perception data. Therefore, the vehicle decelerates and stops and the HMI broadcasts.

3. The safety processing method for the collision risk of an autonomous vehicle according to claim 1, wherein, In step P3, when accurate positioning, map, and perception information cannot be obtained and the decision results corresponding to the emergency situation cannot be output, the following processing methods are included: a1. When the positioning signal is unstable, resulting in a low positioning accuracy and a small error judgment, the vehicle adaptively adjusts based on the decision algorithm threshold and the HMI broadcasts; or when the error judgment is large, the vehicle decelerates and stops and the HMI broadcasts; a2. When the positioning signal is unstable and the positioning signal is lost, based on the previous frame of data of the vehicle, it is obtained that the positioning sensor of the vehicle cannot work properly or the positioning algorithm cannot work properly, and the vehicle decelerates and stops and the HMI broadcasts; a3. When the perception signal is unstable, analyze the input of the perception accuracy error judgment, the input of the perception missed detection situation, and the input of the perception type error judgment with the data of the perception module of the vehicle. If the output error is allowed, the vehicle adaptively adjusts based on the decision algorithm threshold and the HMI broadcasts; or if the error is large, the vehicle decelerates and stops and the HMI broadcasts.

4. The safety processing method for the automatic driving collision risk according to claim 1, characterized in that In step P4, the decision failure records corresponding to the decision results of the emergency situation include: undetected static obstacles and errors, undetected braking obstacles and errors, and undetected pedestrian blind spots and errors; for the decision failure records, the decision algorithm is updated in advance to eliminate and overwrite them; when the adjustment of the decision algorithm for obtaining the decision result is too large, resulting in an input error exceeding the decision threshold, the vehicle decelerates and stops and the HMI broadcasts; when the decision result is delayed beyond the threshold, the recorded data is output for the engineer to modify, and thus the vehicle decelerates and stops and the HMI broadcasts.

Images