Method for predicting driving risks of intelligent vehicle in different driving areas based on predictive risk field and FGC algorithm

Abstract

The invention discloses a driving risk prediction method for different driving areas of an intelligent vehicle based on a predictive risk field and an FGC algorithm, and belongs to the technical field of artificial intelligence and automatic driving. The driving space around the vehicle is divided into six areas, and a static obstacle and dynamic obstacle risk potential field is constructed by means of an artificial potential field idea; the driving risk quantification problem is converted into field intensity calculation, and the risk degree is judged through an FGC algorithm. The problems that driving risks are difficult to quantify visually and driving risk prediction quantitative standard uncertainty is neglected by adopting threshold value to divide risk levels in a traditional method are solved.

Description

Technical Field

[0001] The present invention relates to the fields of artificial intelligence and autonomous driving technology, and in particular to a method for predicting driving risks in different driving areas of an intelligent vehicle based on a predictive risk field and an FGC algorithm. Background Art

[0002] As the automotive industry evolves toward intelligence, information technology, and digitalization, predicting driving risks for intelligent vehicles is becoming increasingly important. It's not only the core of autonomous driving technology but also a key factor currently hindering its widespread adoption. Existing driving risk prediction methods are insufficient to address the growing complexity of autonomous driving systems. As the level of automation in autonomous driving technology increases, the requirements for system safety are also increasing at an almost exponential rate.

[0003] Currently, there are several main approaches to predicting driving risks for intelligent vehicles. Time-based indicators operate from a temporal perspective, with the core assumption being that collision risk is inversely proportional to the time it takes to reach the potential point of conflict while maintaining the current state of motion. Common indicators of this type include time headway (THW) and time to collision (TTC). Kinematic indicators rely on the vehicle's kinematics and dynamics, assessing risk by analyzing the vehicle's state and its relative motion with other traffic participants. Spatial safety distance is typically used as a criterion for assessment. The study of safety distance can be further categorized into longitudinal and lateral aspects. The advantage of these indicators is their ability to assess potential risks in the driving environment in real time, providing strong support for decision-making and planning. However, their drawbacks are also significant. They require high computational resources and make assumptions about the behavior of traffic participants. If actual behavior does not conform to these assumptions, the accuracy of the results will be significantly reduced. Statistical indicators take future uncertainties into account and apply probability theory to assess the probability of a collision. However, this approach is extremely data-intensive, requiring a large amount of high-quality labeled data. The indicators based on artificial potential fields start from the perspective of potential field theory and describe the various risks faced during driving in the form of potential fields.

[0004] In the field of vehicle driving risk prediction, risk assessment methods based on time indicators and kinematic indicators are often used. Although they are simple and efficient, they deviate from real road driving scenarios due to the assumption of a constant motion state. In contrast, the situation assessment method based on artificial potential fields has obvious advantages. It can integrate various factors, comprehensively consider and provide highly interpretable assessment results. However, traditional methods also have obvious limitations: first, the potential field construction only focuses on the current static risk distribution and does not dynamically model the evolution of potential risks in the future; second, a differentiated risk decoupling mechanism for driving areas is not established, which reduces the reliability of risk perception decisions; third, traditional methods use thresholds to divide risk levels and ignore the uncertainty of driving risk quantitative standards. Therefore, the present invention proposes a predictive driving risk field model, which predicts driving risks in different driving areas by combining vehicle trajectory prediction, driving intention recognition information and the forward Gaussian cloud (FGC) algorithm, aiming to solve the above challenges and effectively improve the adaptability and safety of autonomous driving systems in complex traffic environments. Summary of the Invention

[0005] The purpose of the present invention is to address the technical defects in the prior art and provide a method for predicting driving risks in different driving areas of intelligent vehicles based on predictive risk fields and FGC algorithm.

[0006] The technical solution adopted to achieve the purpose of the present invention is:

[0007] A method for predicting driving risks in different driving areas of an intelligent vehicle based on a predictive risk field and an FGC algorithm includes the following steps:

[0008] Step 1: Divide the driving area around the ego vehicle into six regions r, including the area in the same direction lane directly in front of the ego vehicle, the area in the same direction lane directly behind the ego vehicle, the area in front of the left and right adjacent lanes of the ego vehicle, and the area behind the left and right adjacent lanes of the ego vehicle;

[0009] Step 2: Based on the area r divided in step 1, the ego vehicle senses the relative motion state of the traffic participants in each area and itself. Based on the relative motion state, it first determines whether there is a static obstacle. If there is a static obstacle, the static obstacle risk field strength E is calculated. s,r (k);

[0010] Then, the vehicle determines whether there are any dynamic obstacles in the area in front of the left and right adjacent lanes. If there are dynamic obstacles, the vehicle determines whether the dynamic obstacles in the area in front of the left and right adjacent lanes have the intention to change lanes.

[0011] If the dynamic obstacle has the intention to change lanes, calculate the predictive dynamic obstacle risk field strength E of the vehicle in the lane-changing state in the area in front of the left and right adjacent lanes p,r (k);

[0012] If the dynamic obstacle has no lane-changing intention, then determine whether the ego vehicle has lane-changing intention. If the ego vehicle has lane-changing intention, then calculate the predictive dynamic obstacle risk field strength E of the dynamic obstacles in the rear area of the left and right adjacent lanes when the ego vehicle is in the lane-changing state. p,r (k);

[0013] If the ego vehicle has no intention to change lanes, calculate the predictive dynamic obstacle risk field strength E of the dynamic obstacles in the same direction lane area in front and behind the ego vehicle. p,r (k);

[0014] Finally, the risk field strength E of the vehicle affected by static obstacles and / or dynamic obstacles in each area is calculated. total,r (k), where, for each region, when there are only static obstacles, the E total,r (k) = E s,r (k); When there are static obstacles and dynamic obstacles, E total,r (k) = E s,r (k)+E p,r (k); When there are only dynamic obstacles, E total,r (k) = E p,r (k);

[0015] Step 3: In each area, calculate the degree of certainty μ under the driving danger situation through the FGC algorithm urgent The certainty μ between cloud droplets and driving safety situation safe Cloud droplets, and search for the closest risk field strength E obtained in step 2 total,r (k) of cloud droplets, respectively, are the certainty of the driving danger situation μ urgent,r (k) and the certainty of driving safety situation μ safe,r (k), if μ urgent,r (k)>μ safe,r (k) indicates that the risk level of the area is high at the current k moment, and vice versa.

[0016] In the above technical solution, in step 2, based on the areas divided in step 1, the vehicle senses the current motion state of the traffic participants in each area, and the motion state includes the relative position and relative speed between the vehicle and other traffic participants.

[0017] In the above technical solution, in step 2, the static obstacle risk field strength E is calculated. s,r (k):

[0018]

[0019] Where s and d represent the s-coordinate and d-coordinate of the static obstacle in the Frenet coordinate system, respectively; s0 and d0 represent the source coordinates of the vehicle in the Frenet coordinate system, respectively; m is the position vector of the source coordinate point relative to other vehicles in the Frenet coordinate system; the expression is m = (s-s0, d-d0), k s and k d Represents the risk distribution factor of the vehicle along the s and d directions of the Frenet coordinate system.

[0020] In the above technical solution, the risk distribution factor is expressed as:

[0021]

[0022] Among them, P s and P d is the size of the static obstacle; and They represent the speed of the vehicle in the s and d directions of the Frenet coordinate system, respectively, in m / s; k1 and k2 are unknown constants.

[0023] In the above technical solution, in step 2, if there are no static obstacles, the vehicle senses whether there are dynamic obstacles in the area in front of the left and right adjacent lanes of the vehicle:

[0024] If there is a dynamic obstacle, then according to the vehicle's lane change intention D obj (k) Determine whether the target dynamic obstacle in the area ahead of the left and right adjacent lanes has a lane-changing intention; if the dynamic obstacle has no lane-changing intention, determine the lane-changing intention of the vehicle according to the vehicle's lane-changing intention D. ego (k) Determine whether the vehicle intends to change lanes.

[0025] If the target dynamic obstacle has the intention to change lanes, then the predicted trajectory of the target vehicle is combined Assuming the ego vehicle has a constant speed, calculate the predictive dynamic obstacle risk field strength E of the target vehicle in the area to the ego vehicle. p,r (k);

[0026] If the target dynamic obstacle has no lane-changing intention, according to the lane-changing intention of the ego vehicle D ego (k) Determine whether the vehicle intends to change lanes.

[0027] If the vehicle intends to change lanes, then the predicted trajectory of the vehicle is combined with the Assuming that the target dynamic obstacles in the surrounding area have a constant speed, calculate the predictive dynamic obstacle risk field strength E of the target vehicles in the surrounding area to the ego vehicle. p,r (k);

[0028] If the vehicle has no intention to change lanes, calculate the predictive dynamic obstacle risk field strength E of the vehicle from the dynamic obstacles in the area p,r (k), the lane-changing intention influencing factor k d is 1, which means there is no lane-changing intention.

[0029] In the above technical solution, the predictive dynamic obstacle risk field strength E p,r (k):

[0030]

[0031] Where s p,i and d p,i The s coordinate and d coordinate of the predicted trajectory of the vehicle in the Frenet coordinate system at the i-th second or s p,i and d p,i The predicted trajectory of the target vehicle is The predicted coordinates at the i-th second, k p,i is the impact factor of the predictive risk field at the i-th second, θ is the angle between the relative position vector m of the dynamic obstacle target vehicle and the ego vehicle and the direction of the moving speed, k d is the influencing factor of lane-changing intention.

[0032] In the above technical solution, the step 3 specifically includes the following steps:

[0033] Step 3.1: First calculate N under dangerous driving situation and safe driving situation cloud The degree of certainty μ i (x i ), including the degree of certainty μ under dangerous driving situations urgent and the certainty μ of driving safety situation safe , the degree of certainty μ under the dangerous driving situation urgent The expected value Ex of the driving danger situation urgent , Entropy of dangerous driving situation En urgent and the super entropy He of dangerous driving situations urgent Determine the degree of certainty μ of the driving safety situation safe The expected value Ex of driving safety situation safe , Entropy of driving safety situation En safe and the super entropy He of driving safety situation safe Sure;

[0034] Step 3.2: Determination of the degree μ under the driving danger situation urgent and the certainty μ of driving safety situation safe Search the cloud droplets to find the closest risk field strength E in each area obtained in step 2 total,rThe cloud droplets of (k) correspond to the certainty of the driving danger situation μ urgent,r (k) and the certainty of driving safety situation μ safe,r (k), if μ urgent,r (k)>μ safe,r (k) indicates that the risk level of the area is high at the current k moment, and vice versa.

[0035] In the above technical solution, in step 3.1, the N cloud The degree of certainty μ i (x i ) is expressed as:

[0036]

[0037] En' i =NORM(En,He 2 ) (4)

[0038] x i =NORM(Ex,En′ i 2 )

[0039] Among them, x i N cloud The ith cloud droplet in the cloud droplets x, μ i (x i ) represents cloud droplet x i The degree of certainty, Ex is the expected value Ex of the driving danger situation urgent Or the expected value of driving safety situation Ex safe , En is the entropy of dangerous driving situation En urgent Or the entropy of driving safety situation En safe , He is the super entropy He of the dangerous driving situation urgent Or the super entropy He of driving safety situation safe .

[0040] In the above technical solution, the expected value Ex of the driving danger situation urgent , Entropy of dangerous driving situation En urgent , the super entropy He of dangerous driving situations urgent , the expected value of driving safety situation Ex safe , Entropy of driving safety situation En safe and the super entropy He of driving safety situation safe Respectively expressed as:

[0041]

[0042] Here, CD represents the degree of mixing.

[0043] In the above technical solution, the E urgent For E urgent1 or E urgent2 , E urgent1 It is a quantitative value of the driving risk of the vehicle being affected by dynamic obstacles in the same-direction lane area in front and behind. When the longitudinal distance between the two vehicles reaches the minimum safe distance d during braking, brake_min When facing rear-end collision risk, d brake_min =t TTC ·v x_rear , t TTC Indicates the minimum collision time in seconds; v x_rear Indicates the longitudinal speed of the vehicle behind at the current sampling period k, in meters per second; according to d brake_min Convert the coordinates Δs and Δd into the Frenet coordinate system, and calculate the dynamic obstacle risk field strength E according to formula (6) as E urgent1 :

[0044]

[0045] The E urgent2 It is the quantitative value of the driving risk of the vehicle being affected by dynamic obstacles in the front area of the left and right adjacent lanes when changing lanes. When the longitudinal distance between the two vehicles is less than the minimum safe distance d during the lane change process, change_min When the vehicle is in the same direction, it poses a safety threat to the two vehicles, including:

[0046]

[0047] Where, t h is the headway, d cg_car d cg_rear or d cg_front ;d cg_rear For the rear vehicle, the maximum comfortable deceleration is a b_max The distance required to stop from braking; d cg_rear The vehicle ahead is decelerated at the maximum comfortable speed a b_max The distance required to stop from braking; T s is the sampling time; v b_max is the speed reduction in each sampling period; t car For the maximum comfortable deceleration a b_max The number of sampling cycles required to brake to stop; according to d change_min Convert the coordinates Δs and Δd into the Frenet coordinate system, and calculate the dynamic obstacle risk field strength E according to formula (6) as E urgent2 .

[0048] In the above technical solution, the driving risk quantification value E of the vehicle in the lane-changing state due to dynamic obstacles in the front area of the same lane or the left and right adjacent lanes issafe , when the longitudinal distance between the vehicle in the adjacent lane and the vehicle reaches d des It will not pose a threat to the driving safety of the two vehicles.

[0049]

[0050] Where, d min is the minimum safe distance between the two vehicles, which is the minimum safe distance d during the braking process of the two vehicles. brake_min The minimum safe distance d during the following process follow_min Decide and take the maximum value; L car is the vehicle body length; d0 is a fixed distance value; b1, b2 and b3 are parameters greater than 0; v x_rel is the longitudinal relative speed of the vehicle in the current sampling period; a x_front is the longitudinal acceleration of the vehicle ahead in the current sampling period, according to d des Convert the coordinates Δs and Δd into the Frenet coordinate system, and calculate the dynamic obstacle risk field strength E according to formula (6) as E urgent2 .

[0051] Compared with the prior art, the present invention has the following beneficial effects:

[0052] 1. This invention considers the size of the static obstacle and the fact that the risk potential field value increases faster as the ego-vehicle approaches the static obstacle. It then uses field theory and introduces a two-dimensional normal distribution model to describe the static risk field strength of the static obstacle at (s, d).

[0053] 2. This invention divides the driving space around the vehicle into six zones and constructs static and dynamic obstacle risk potential fields using the concept of artificial potential fields. This converts the driving risk quantification problem into field strength calculation, solving the problem of driving risk being difficult to intuitively quantify.

[0054] 3. This invention integrates temporal dynamic features within a six-area driving risk estimation framework, combines lane change intention recognition with vehicle trajectory prediction results, and further enables precise quantitative risk assessment of the six areas surrounding the vehicle. This improves the accuracy and timeliness of driving risk assessment, providing stronger support for autonomous driving decision-making.

[0055] 4. This invention addresses the uncertainty problem of the quantitative standard for driving risk prediction by refining the concept of driving risk into two sub-situations: danger and safety. Based on the FGC algorithm, quantitative values and qualitative concepts are converted into each other, making the driving risk prediction more comprehensive and accurate. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] Figure 1 Divide the driving area around the vehicle into different areas.

[0057] Figure 2 Schematic diagram of the risk field distribution of static obstacles.

[0058] Figure 3 Schematic diagram of the risk field distribution of dynamic obstacles.

[0059] Figure 4 Schematic diagram of the algorithm flow of the present invention.

[0060] Figure 5 This is the influence analysis of the static obstacle model parameters, where (a) is the influence of k1 on s and b, and (b) is the influence of k2 on s and b.

[0061] Figure 6 Analysis of the impact of dynamic obstacle model parameters, where (a) is the impact of k1 on s and b, and (b) is the impact of k2 on s and b.

[0062] Figure 7 These are the Gaussian cloud distribution diagrams of the two situations of the vehicle in the current state. DETAILED DESCRIPTION

[0063] The present invention will be further described in detail below with reference to specific embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0064] like Figure 2-Figure 4 As shown, a method for predicting driving risks in different driving areas of an intelligent vehicle based on a predictive risk field and an FGC algorithm includes the following steps:

[0065] Step 1: Driving area division: Since the attention points and response methods of the ego vehicle to vehicles in different directions around the road are different, different areas are divided according to the driving situation, mainly considering the traffic participants in the lane and adjacent lanes, and the driving area around the ego vehicle is divided into six areas r, such as Figure 1 As shown:

[0066] Area 2 is the lane in the same direction directly in front of the vehicle, and is the key area for driving observation. When there is a vehicle driving or cutting in front, there is a risk of rear-end collision with the vehicle.

[0067] Zone 5 is located directly behind the ego vehicle in the same direction. Drivers typically pay little attention to this area, but the autonomous vehicle can perceive it. When the ego vehicle decelerates, it must consider the risks in this area to prevent rear-end collisions.

[0068] Areas 1 and 3 are located in front of the adjacent lanes on the left and right of the ego vehicle. If both the ego vehicle and the vehicles in these areas maintain their lanes, the threat is relatively small due to the lane markings. However, if a vehicle in these areas intends to change lanes, the threat increases. This also requires special attention when the ego vehicle changes lanes.

[0069] Areas 4 and 6 are located behind the left and right adjacent lanes of the vehicle. The risks involved need to be considered when the vehicle changes lanes.

[0070] Step 2: Based on the area r divided in step 1, the vehicle senses whether there are static obstacles in each area of the traffic participants (specifically vehicles in this embodiment). If there are static obstacles, the static obstacle risk field strength E is calculated. s,r (k) Then, it is determined whether there are dynamic obstacles in the area in front of the left and right adjacent lanes (i.e., area 1 and area 3). If there are dynamic obstacles, the vehicle determines whether the dynamic obstacles in the area in front of the left and right adjacent lanes have the intention to change lanes. If the dynamic obstacles have the intention to change lanes, the predictive dynamic obstacle risk field strength E of the vehicle in the area in front of the left and right adjacent lanes is calculated. p,r (k); If the dynamic obstacle has no lane-changing intention, determine whether the ego vehicle has lane-changing intention. If the ego vehicle has lane-changing intention, calculate the predictive dynamic obstacle risk field strength E of the dynamic obstacles in the rear area of the left and right adjacent lanes when the ego vehicle is in the lane-changing state. p,r (k) If the ego vehicle has no intention to change lanes, calculate the predictive dynamic obstacle risk field strength E of the dynamic obstacles in the same direction lane area in front and behind the ego vehicle. p,r (k); Finally, calculate the risk field strength E of the vehicle affected by static obstacles and / or dynamic obstacles in each area total,r (k).

[0071] The specific steps include:

[0072] Step 2.1: Check whether the speed of the target vehicle in each area is zero, that is, whether there is a static obstacle. If so, calculate the static obstacle risk field strength E s,r (k):

[0073]

[0074] Where k is the current sampling period, s and d represent the s-coordinate and d-coordinate of the static obstacle in the Frenet coordinate system respectively; s0 and d0 represent the field source coordinates of the vehicle in the Frenet coordinate system respectively; m is the position vector of the field source coordinate point relative to other vehicles in the Frenet coordinate system; the expression is m = (s-s0, d-d0), k s and k d Represents the risk distribution factor of the vehicle along the s and d directions of the Frenet coordinate system.

[0075] For the risk distribution of static obstacles, the motion state of the vehicle itself will also affect it, so the risk distribution factor is expressed as:

[0076]

[0077] Among them, P s and P d is the size of the static obstacle; and They represent the speed of the vehicle in the s and d directions of the Frenet coordinate system, respectively, in m / s; k1 and k2 are unknown constants.

[0078] like Figure 5 As shown in Figure 2, the influence of parameters k1 and k2 on the static obstacle risk model. Figure 5 (a) shows that as k1 increases, the influence range of the risk potential field in the s direction increases. When k1 increases a times, the influence range increases. The influence range in the d direction has not changed; Figure 5 As shown in (b), the risk potential field increases in the d direction as k2 increases. When k2 increases by b times, the impact range increases. The influence range in the s direction remains unchanged, wherein the influence range is defined as the maximum continuous coverage length when the field strength in the s direction or the d direction is greater than 0.05. In this embodiment, k1=4 and k2=8.

[0079] Step 2.2: If there are no static obstacles, check whether there is a target vehicle with a non-zero speed in area 1 or area 3, i.e., a dynamic obstacle. If so, proceed to step 2.3; otherwise, proceed to step 2.4.

[0080] Step 2.3: Identify D based on the vehicle’s lane change intention obj (k) (Based on the vehicle's own detection or perception) Determine whether the target vehicle in area 1 or area 3 intends to change lanes:

[0081] If the target vehicle has no intention to change lanes, then the predicted trajectory of the target vehicle (Based on the vehicle's own detection or perception), assuming the vehicle's constant speed, calculate the predictive dynamic obstacle risk field strength E of the target vehicle in the area to which the vehicle is subject. p,r (k):

[0082]

[0083] Where s p,i and d p,i are the s coordinate and d coordinate of the predicted trajectory of the vehicle in the Frenet coordinate system at the i-th second; k p,i is the impact factor of the predictive risk field at the i-th second, and the trajectory prediction time domain is set to 5s, s p,i and d p,i The predicted trajectory of the target vehicle is The predicted coordinates at the i-th second. Since the longer the prediction time, the greater the uncertainty of the trajectory prediction, the influence factor k p,i Set as θ is the angle between the relative position vector m of the dynamic obstacle target vehicle and the ego vehicle and the direction of motion speed, k d is the influencing factor of lane-changing intention.

[0084] If the target vehicle has the intention to change lanes, execute step 2.4;

[0085] Furthermore, the predictive dynamic obstacle risk field strength E of the target vehicle to which the ego vehicle is subjected in this embodiment is p,r (k) is in the dynamic obstacle risk field E d The dynamic obstacle risk field potential E is established on the basis of d :

[0086] Dynamic obstacles include motor vehicles, non-motor vehicles, and pedestrians. These obstacles pose a greater threat to the vehicle and may cause serious traffic accidents if not handled properly. Dynamic obstacles have the following characteristics:

[0087] (1) The severity of a collision between a vehicle and a dynamic obstacle is related to the obstacle's own properties and its motion state.

[0088] (2) The closer the distance between the vehicle and the dynamic obstacle, the higher the probability of collision, and the probability of collision does not increase linearly with the damage caused by the collision. As the distance gets closer, the driving risk increases significantly.

[0089] (3) The potential risk of a dynamic obstacle to the ego vehicle is not only related to the distance between the two, but also to the speed and relative position vector of the dynamic obstacle. At the same distance, the smaller the angle between the ego vehicle's direction vector relative to the dynamic obstacle and the object's speed, the greater the risk to the ego vehicle.

[0090] Therefore, the dynamic risk E of the dynamic obstacle at (s, d) d It can be expressed as:

[0091]

[0092] Among them, θ is the angle between the relative position vector m of the dynamic obstacle and the direction of motion speed.

[0093] like Figure 6 As shown in Figure 2, the influence of parameters k1 and k2 on the model. Figure 6 (a) shows that the risk potential field increases with k d The influence range in the s direction increases with the increase, showing a linear relationship, while the influence range in the d direction does not change; Figure 6As can be seen from (b), as k2 increases, the impact range of the risk potential field in the d direction increases and the impact degree gradually decreases, while the impact range in the s direction does not change. In this embodiment, k1=6.1 and k2=0.5 are taken.

[0094] Step 2.4: Based on the lane-changing intention D of the vehicle ego (k) Determine whether the vehicle intends to change lanes;

[0095] If the vehicle intends to change lanes, the predicted trajectory of the vehicle is combined with the (Detection or perception based on the vehicle itself), assuming that the target vehicles in the surrounding area have a constant speed, calculate the potential dynamic obstacle risk force field strength E of the target vehicles in the surrounding area to the vehicle according to Formula 5 p,r (k) (calculated according to formula (3));

[0096] If the ego vehicle has no lane-changing intention, proceed to step 2.5.

[0097] Step 2.5: Calculate the predicted dynamic obstacle risk field strength E of the vehicle from the vehicles in the area. p,r (k) (calculated according to formula (3)), the lane change intention influencing factor k d is 1, which means there is no lane-changing intention.

[0098] Step 2.6, calculate the driving risk potential field strength E of each driving area to which the vehicle is subject total,r (k), where, for each region, when there are only static obstacles, the E total,r (k) = E s,r (k); When there are static obstacles and dynamic obstacles, the E total,r (k) = E s,r (k)+E p,r (k); When there are only dynamic obstacles, the E total,r (k) = E p,r (k);

[0099] Step 3: In each area, calculate the degree of certainty μ under the driving danger situation through the FGC algorithm urgent The certainty μ between cloud droplets and driving safety situation safe Cloud droplets, and search for the closest risk field strength E obtained in step 2 total,r (k) of cloud droplets, respectively, are the certainty of the driving danger situation μ urgent,r (k) and the certainty of driving safety situation μ safe,r (k), if μ urgent,r (k)>μ safe,r (k) indicates that the risk level of the area at the current k moment is high, otherwise it is low. Specifically, the following steps are included:

[0100] Step 3.1, first calculate N cloud The degree of certainty μ i (x i ):

[0101]

[0102] Among them, x i N cloud The ith cloud droplet in the cloud droplets x, μ i (x i ) represents cloud droplet x i The degree of certainty, including the degree of certainty μ under dangerous driving situations urgent and the certainty μ of driving safety situation safe , Ex is the expected value Ex of the driving danger situation urgent Or the expected value of driving safety situation Ex safe , En is the entropy of dangerous driving situation En urgent Or the entropy of driving safety situation En safe , He is the super entropy He of the dangerous driving situation urgent Or the super entropy He of driving safety situation safe :

[0103]

[0104]

[0105] Among them, CD represents the degree of mixing. When CD is close to 0, the super entropy is small, indicating that the concept extension is convergent and it is easy to form a consensus. When CD>=1, the super entropy increases, indicating that the concept extension is dispersed and ambiguous, and it is difficult to form a consensus.

[0106] The E urgent For E urgent1 or E urgent2 The driving risk quantification value of the vehicle affected by dynamic obstacles in the same direction lane area in front and behind is defined as E urgent1 , when the longitudinal distance between the two vehicles reaches the minimum safe distance d during braking brake_min When facing rear-end collision risk, d brake_min =t TTC ·v x_rear , t TTC Indicates the minimum collision time in seconds; v x_rear Indicates the longitudinal speed of the vehicle behind the current sampling period k, in meters per second; according to d brake_min Convert the coordinates Δs and Δd into the Frenet coordinate system, and calculate the dynamic obstacle risk field strength E according to formula (6) as E urgent1 .

[0107]

[0108] Define the driving risk quantification value E of the vehicle when it is facing dynamic obstacles in the front area of the left and right adjacent lanes during lane change urgent2 , when the longitudinal distance between the two vehicles is less than the minimum safe distance d during lane change change_min When the vehicle is in the same direction, it poses a safety threat to the two vehicles, including:

[0109]

[0110] Where, t h is the headway, d cg_car d cg_rear or d cg_front ;d cg_rear For the rear vehicle, the maximum comfortable deceleration is a b_max The distance required to stop from braking; d cg_rear The vehicle ahead is decelerated at the maximum comfortable speed a b_max The distance required to stop from braking; T s is the sampling time; v b_max is the speed reduction in each sampling period; t car For the maximum comfortable deceleration a b_max The number of sampling cycles required to brake to stop; according to d change_min Convert the coordinates Δs and Δd into the Frenet coordinate system, and calculate the dynamic obstacle risk field strength E according to formula (6) as E urgent2 .

[0111] Among them, the driving risk quantification value E of the vehicle when it is subject to dynamic obstacles in the same lane or the area in front of the left and right adjacent lanes in the lane change state is safe , when the longitudinal distance between the vehicle in the adjacent lane and the vehicle reaches d des It will not pose a threat to the driving safety of the two vehicles.

[0112]

[0113] Where, d min is the minimum safe distance between the two vehicles, which is the minimum safe distance d during the braking process of the two vehicles. brake_min The minimum safe distance d during the following process follow_min Decide and take the maximum value; L car is the vehicle body length; d0 is a fixed distance value; b1, b2 and b3 are parameters greater than 0; v x_rel is the longitudinal relative speed of the vehicle in the current sampling period; a x_front is the longitudinal acceleration of the vehicle ahead in the current sampling period. desConvert the coordinates Δs and Δd into the Frenet coordinate system, and calculate the dynamic obstacle risk field strength E according to formula (6) as E urgent2 .

[0114] Step 3.2, such as Figure 7 As shown in the figure, the schematic diagram of the Gaussian cloud droplet distribution calculated by step 3.1 is shown. The horizontal axis represents the field strength value E, and the vertical axis represents the certainty μ of the corresponding situation when the field strength is E in the current state. The cloud droplets with the horizontal axis closest to the risk field strength E obtained in step 2 are searched in the driving danger situation and driving safety situation cloud droplets. total,r The vertical coordinate of the cloud droplet (k) is the degree of certainty μ of the driving danger situation. urgent,r (k) and the certainty of driving safety situation μ safe,r (k), if μ urgent,r (k)>μ safe,r (k) indicates that the risk level of the area is high at the current k moment, and vice versa.

[0115] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A method for predicting driving risks in different driving areas for intelligent vehicles based on predictive risk fields and FGC algorithm, characterized by: The following steps are involved: Step 1: Divide the driving area around the ego vehicle into six regions r, including the area in the same direction lane directly in front of the ego vehicle, the area in the same direction lane directly behind the ego vehicle, the area in front of the left and right adjacent lanes of the ego vehicle, and the area behind the left and right adjacent lanes of the ego vehicle; Step 2: Based on the area r divided in step 1, the ego vehicle senses the relative motion state of the traffic participants in each area and itself. Based on the relative motion state, it first determines whether there is a static obstacle. If there is a static obstacle, the static obstacle risk field strength E is calculated. s,r (k); Then, the vehicle determines whether there are any dynamic obstacles in the area in front of the left and right adjacent lanes. If there are dynamic obstacles, the vehicle determines whether the dynamic obstacles in the area in front of the left and right adjacent lanes have the intention to change lanes. If the dynamic obstacle has the intention to change lanes, calculate the predictive dynamic obstacle risk field strength E of the vehicle in the lane-changing state in the area in front of the left and right adjacent lanes p,r (k); If the dynamic obstacle has no lane-changing intention, then determine whether the ego vehicle has lane-changing intention. If the ego vehicle has lane-changing intention, then calculate the predictive dynamic obstacle risk field strength E of the dynamic obstacles in the rear area of the left and right adjacent lanes when the ego vehicle is in the lane-changing state. p,r (k) If the ego vehicle has no intention to change lanes, calculate the predictive dynamic obstacle risk field strength E of the dynamic obstacles in the same direction lane area in front and behind the ego vehicle. p,r (k); Finally, the risk field strength E of the vehicle affected by static obstacles and / or dynamic obstacles in each area is calculated. total,r (k), where, for each region, when there are only static obstacles, the E total,r (k) = E s,r (k); When there are static obstacles and dynamic obstacles, the E total,r (k) = E s,r (k)+E p,r (k); When there are only dynamic obstacles, the E total,r (k) = E p,r (k); Step 3: In each area, calculate the degree of certainty μ under the driving danger situation through the FGC algorithm urgent The certainty μ between cloud droplets and driving safety situation safe Cloud droplets, and search for the closest risk field strength E obtained in step 2 total,r (k) of cloud droplets, respectively, are the certainty of the driving danger situation μ urgent,r (k) and the certainty of driving safety situation μ safe,r (k), if μ urgent,r (k)>μ safe,r (k) indicates that the risk level of the area is high at the current k moment, and vice versa.

2. The prediction method according to claim 1, characterized in that In step 2, based on the areas divided in step 1, the vehicle senses the current motion state of the traffic participants in each area, where the motion state includes the relative position and relative speed of the vehicle and other traffic participants.

3. The prediction method according to claim 1, wherein: In step 2, calculate the static obstacle risk field strength E s,r (k): Where s and d represent the s-coordinate and d-coordinate of the static obstacle in the Frenet coordinate system, respectively; s0 and d0 represent the source coordinates of the vehicle in the Frenet coordinate system, respectively; m is the position vector of the source coordinate point relative to other vehicles in the Frenet coordinate system; the expression is m = (s-s0, d-d0), k s and k d Represents the risk distribution factor of the vehicle along the s and d directions of the Frenet coordinate system.

4. The prediction method according to claim 3, characterized in that The risk distribution factor is expressed as: Among them, P s and P d is the size of the static obstacle; and They represent the speed of the vehicle in the s and d directions of the Frenet coordinate system, respectively, in m / s; k1 and k2 are unknown constants.

5. The prediction method according to claim 1, wherein: The predictive dynamic obstacle risk field strength E p,r (k): Where s p,i and d p,i The s coordinate and d coordinate of the predicted trajectory of the vehicle in the Frenet coordinate system at the i-th second or s p,i and d p,i The predicted trajectory of the target vehicle is The predicted coordinates at the i-th second, k p,i is the impact factor of the predictive risk field at the i-th second, θ is the angle between the relative position vector m of the dynamic obstacle target vehicle and the ego vehicle and the direction of the moving speed, k d is the influencing factor of lane-changing intention.

6. The prediction method according to claim 1, characterized in that The step 3 specifically includes the following steps: Step 3.1: First calculate N under dangerous driving situation and safe driving situation cloud The degree of certainty μ i (x i ), including the degree of certainty μ under dangerous driving situations urgent and the certainty μ of driving safety situation safe , the degree of certainty μ under the dangerous driving situation urgent The expected value Ex of the driving danger situation urgent , Entropy of dangerous driving situation En urgent and the super entropy He of dangerous driving situations urgent Determine the degree of certainty μ of the driving safety situation safe The expected value Ex of driving safety situation safe , Entropy of driving safety situation En safe and the super entropy He of driving safety situation safe Sure; Step 3.2: Determination of the degree μ under the driving danger situation urgent and the certainty μ of driving safety situation safe Search the cloud droplets to find the closest risk field strength E in each area obtained in step 2 total,r The cloud droplets of (k) correspond to the certainty of the driving danger situation μ urgent,r (k) and the certainty of driving safety situation μ safe,r (k), if μ urgent,r (k)>μ safe,r (k) indicates that the risk level of the area is high at the current k moment, and vice versa.

7. The prediction method according to claim 6, characterized in that In step 3.1, the N cloud The degree of certainty μ i (x i ) is expressed as: Among them, x i N cloud The ith cloud droplet in the cloud droplets x, μ i (x i ) represents cloud droplet x i The degree of certainty, Ex is the expected value Ex of the driving danger situation urgent Or the expected value of driving safety situation Ex safe , En is the entropy of dangerous driving situation En urgent Or the entropy of driving safety situation En safe , He is the super entropy He of the dangerous driving situation urgent Or the super entropy He of driving safety situation safe .

8. The prediction method according to claim 7, characterized in that The expected value Ex of the driving danger situation urgent , Entropy of dangerous driving situation En urgent , the super entropy He of dangerous driving situations urgent , the expected value of driving safety situation Ex safe , Entropy of driving safety situation En safe and the super entropy He of driving safety situation safe Respectively expressed as: Here, CD represents the degree of mixing.

9. The prediction method according to claim 8, characterized in that The E urgent For E urgent1 or E urgent2 , E urgent1 It is a quantitative value of the driving risk of the vehicle being affected by dynamic obstacles in the same-direction lane area in front and behind. When the longitudinal distance between the two vehicles reaches the minimum safe distance d during braking, brake_min When facing rear-end collision risk, d brake_min =t TTC ·v x_rear , t TTC Indicates the minimum collision time in seconds; v x_rear Indicates the longitudinal speed of the vehicle behind at the current sampling period k, in meters per second; according to d brake_min Convert the coordinates Δs and Δd into the Frenet coordinate system, and calculate the dynamic obstacle risk field strength E according to formula (6) as E urgent1 : The E urgent2 It is the quantitative value of the driving risk of the vehicle being affected by dynamic obstacles in the front area of the left and right adjacent lanes when changing lanes. When the longitudinal distance between the two vehicles is less than the minimum safe distance d during the lane change process, change_min When the vehicle is in the same direction, it poses a safety threat to the two vehicles, including: Where, t h is the headway, d cg_car d cg_rear or d cg_front ;d cg_rear For the rear vehicle, the maximum comfortable deceleration is a b_max The distance required to stop from braking; d cg_rear The vehicle ahead is decelerated at the maximum comfortable speed a b_max The distance required to stop from braking; T s is the sampling time; v b_max is the speed reduction in each sampling period; t car For the maximum comfortable deceleration a b_max The number of sampling cycles required to brake to stop; according to d change_min Convert the coordinates Δs and Δd into the Frenet coordinate system, and calculate the dynamic obstacle risk field strength E according to formula (6) as E urgent2 .

10. The prediction method according to claim 9, characterized in that The quantified value E of the driving risk of the ego vehicle being affected by dynamic obstacles in the same lane or the area in front of the left or right adjacent lane when changing lanes safe , when the longitudinal distance between the vehicle in the adjacent lane and the vehicle reaches d des It will not pose a threat to the driving safety of the two vehicles. Where, d min is the minimum safe distance between the two vehicles, which is the minimum safe distance d during the braking process of the two vehicles. brake_min The minimum safe distance d during the following process follow_min Decide and take the maximum value; L car is the vehicle body length; d0 is a fixed distance value; b1, b2 and b3 are parameters greater than 0; v x_rel is the longitudinal relative speed of the vehicle in the current sampling period; a x_front is the longitudinal acceleration of the vehicle ahead in the current sampling period, according to d des Convert the coordinates Δs and Δd into the Frenet coordinate system, and calculate the dynamic obstacle risk field strength E according to formula (6) as E urgent2 .

Images