Road multi-source risk factor fusion method and device based on automatic driving

Abstract

The invention provides a road multi-source risk factor fusion method based on automatic driving. The method comprises the steps that 1, traffic elements in an automatic driving environment are divided into traffic marking lines, static obstacles and dynamic obstacles; 2, projecting traffic marking type traffic elements into a traffic marking grid map; step 3, projecting the static obstacle traffic elements into the static obstacle grid map; 4, projecting the dynamic obstacle type traffic elements into the dynamic obstacle grid map; 5, fusing the traffic marking grid map, the static obstacle grid map and the dynamic obstacle grid map into a unified occupation grid map; step 6, in the uniformly occupied grid map, the danger coefficient and the influence range of any grid point are # imgabs0 #; and step 7, calculating according to the function # imgabs1 # to obtain a collision risk value of each grid in the unified grid map to the own vehicle.

Description

Technical Field

[0001] This application relates to the field of autonomous driving technology, and in particular, to a method and device for fusing multi-source risk factors of roads based on autonomous driving. Background Art

[0002] With the rapid development of the new energy industry and intelligent vehicles, autonomous driving vehicles have also achieved explosive progress. Major manufacturers have started road tests of autonomous driving vehicles at different levels, and L2+-level autonomous driving has gradually been implemented in new energy vehicles. How to ensure the safe operation of autonomous driving vehicles on the road has become a core issue. There are a large number of risk items from different sources in urban road traffic, including elements such as road environment, traffic participants, and traffic rules. These different risk items will all affect the safe operation of autonomous driving vehicles, but they are different in category, source, and dimension. Different traffic risk items pose different threats to vehicles. For example, traffic markings are regulatory constraints on vehicles and will not cause substantial collisions; fixed obstacles such as guardrails and flower beds will collide with vehicles, but they are static and completely predictable; other traffic participants such as moving vehicles and pedestrians are moving targets, and their movement trajectories are uncertain, posing the greatest risk to the safe operation of autonomous driving vehicles. The avoidance priority levels among these different risk items are also different, and there will be problems of priority conflicts in some extreme cases. How to classify, fuse, and condense these different risk information into information in the same dimension has become a key issue in the field of autonomous driving. Summary of the Invention

[0003] In view of this, the present invention provides a method and device for fusing multi-source risk factors of roads based on autonomous driving to solve the technical problems in the prior art that do not elaborate in detail the calculation methods of the collision risk coefficients and influence ranges of different traffic participants, cannot quantify the risks on the road, do not consider risk information such as lane lines, dynamic and static obstacles on the real-time road, and cannot fuse and uniformly express the risks of different traffic risk items with different categories, sources, and dimensions on the road.

[0004] The present invention provides a method for fusing multi-source risk factors of roads based on autonomous driving. The method includes: Step 1, classifying traffic elements in the autonomous driving environment into three categories: traffic markings, static obstacles, and dynamic obstacles; Step 2, projecting traffic elements of the traffic marking category onto a traffic marking grid map. Traffic markings are divided into single dotted lines, single solid lines, and double yellow lines. Determine the risk coefficient Y1 and influence range N1 of the single dotted line, and the risk coefficient Y2 and influence range N2 of the single solid line and double yellow line; Step 3, projecting traffic elements of the static obstacle category onto a static obstacle grid map, and determining its risk coefficient Y3 and influence range N3; Step 4, projecting traffic elements of the dynamic obstacle category onto a dynamic obstacle grid map, and determining its risk coefficient Y4 and influence range N4; Step 5, fusing the traffic marking grid map, static obstacle grid map, and dynamic obstacle grid map into a unified occupancy grid map; Step 6, in the unified occupancy grid map, the risk coefficient and influence range of any grid point are: Yn = γ attribute , Nn = N attribute , where γ attribute and N attribute respectively represent the target types of the area where the grid is located; Step 7, according to the function calculate the collision risk value generated by each grid in the unified grid map for the ego vehicle, where Risk n represents the collision risk value generated by the nth grid p n in the unified grid map for the ego vehicle, δ represents the scene coefficient, Yn represents the risk coefficient of the grid p n , Nn represents the influence range of the grid, and d(p n , ego) represents the distance between the grid and the ego vehicle.

[0005] Further, the risk coefficient Y1 = 0, the influence range N1 = 0, the risk coefficient Y2 = Ymax, and the influence range N2 = 0.

[0006] Further, the functional formula for determining its risk coefficient Y3 and influence range N3 is: Y3 = γ max , if d(obs, ego) < dis min , d(obs, ego) = dis min , where, represents the lateral / normal influence range of the obstacle in the vehicle driving direction, represents the longitudinal / tangential influence range of the obstacle in the vehicle driving direction, α lat represents the coefficient of the static obstacle in the vehicle driving lateral / normal direction, α lonIndicates the coefficient of the static obstacle in the longitudinal / tangential direction of the vehicle's travel. Monte_Carlo(polygon_area) represents the area of the irregular static obstacle obtained by Monte Carlo sampling method. obs in d(obs,ego) represents the static obstacle, ego represents the host vehicle, and d(obs,ego) represents the distance between the host vehicle and the obstacle. Represents the lateral deviation vector of the static obstacle from the travel direction of the host vehicle. Represents the longitudinal deviation vector of the static obstacle from the travel direction of the host vehicle.

[0007] Furthermore, the functional expressions for determining its risk coefficient Y4 and influence range N4 are: Y4 = γ max ,

[0008] v n =(v ego ×(v obs -v ego ))), v t =(v ego ·

[0009] (v obs -v ego ))), where Represents the lateral / normal influence range of the dynamic obstacle in the direction of its own travel speed. Represents the longitudinal / tangential influence range of the dynamic obstacle in the direction of its own travel speed. α lat Represents the coefficient of the dynamic obstacle in the lateral / normal direction of its own speed. α lon Represents the coefficient of the dynamic obstacle in the longitudinal / tangential direction of the vehicle's travel. Monte_Carlo(polygon_area) represents the area of the irregular dynamic obstacle obtained by Monte Carlo sampling method. obs in d(obs,ego) represents the dynamic obstacle, ego represents the host vehicle, and d(obs,ego) represents the distance between the host vehicle and the obstacle. Represents the lateral deviation vector of the dynamic obstacle in the direction of its own travel. Represents the longitudinal deviation vector of the dynamic obstacle in the direction of its own travel. ‖v obs ‖ represents the magnitude of the obstacle speed vector, which is the value of its speed. v n Calculates the normal vector difference between the host vehicle speed and the obstacle speed of the two speed vectors. v t Calculates the tangential vector difference between the host vehicle speed and the obstacle speed of the two speed vectors. ‖v n ‖ and ‖vt ‖ represent their moduli respectively.

[0010] Furthermore, when the distance between the host vehicle and the static obstacle is less than the preset minimum distance threshold dis min d(obs,ego) = dis min .

[0011] Furthermore, the influence range is proportional to the area of the static obstacle and inversely proportional to the distance from the static obstacle to the host vehicle.

[0012] Furthermore, the influence range is proportional to the area of the dynamic obstacle, inversely proportional to the distance from the dynamic obstacle to the host vehicle, proportional to the speed of the dynamic obstacle, and inversely proportional to the speed difference between the dynamic obstacle and the host vehicle.

[0013] Furthermore, step 6 further includes: when a grid point contains multiple target types, select the target type with the highest priority according to the following priority sorting, Y4 > Y3 > Y2 > Y1, N4 > N3 > N2 > N1.

[0014] The present invention also provides a device for fusing multi-source risk factors of roads based on autonomous driving. The device includes: a classification module, arranged on an intelligent driving vehicle, for classifying traffic elements in the autonomous driving environment into three categories: traffic markings, static obstacles, and dynamic obstacles; a projection module, connected to the classification module, for projecting the three types of traffic elements, namely traffic markings, static obstacles, and dynamic obstacles, into a traffic marking grid map, a static obstacle grid map, and a dynamic obstacle grid map respectively, and obtaining their danger coefficients and influence ranges; a fusion module, connected to the projection module, for fusing the traffic marking grid map, the static obstacle grid map, and the dynamic obstacle grid map into a unified occupancy grid map. In the unified occupancy grid map, the danger coefficient and influence range of any grid point are: Yn = γ attribute , Nn = N attribute , where γ attribute and N attribute respectively represent the target types of the area where the grid is located. Calculate the collision risk value generated by each grid in the unified grid map for the host vehicle according to the function . Among them, Risk n represents the collision risk value generated by the nth grid p n in the unified grid map for the host vehicle. δ is a scenario coefficient. Yn represents the danger coefficient of the grid p n , and Nn represents the influence range of the grid. d(p n ,ego) represents the distance between the grid and the host vehicle.

[0015] Further, when a same grid point contains multiple target types, the fusion module is further configured to select the target type with the highest priority according to the following priority sorting: Y4 > Y3 > Y2 > Y1, N4 > N3 > N2 > N1.

[0016] The present invention provides a method and device for fusing multi-source risk factors of a road based on autonomous driving, mainly used to solve the problem in the prior art that traffic risk elements (road environment, traffic participants, traffic rules, etc.) with different categories, different sources, and different dimensions encountered by autonomous vehicles on the road cannot be quantified. By calculating the danger coefficients and influence ranges of different traffic risk elements, the above different traffic risk elements are fused and expressed on a grid map with a unified quantitative standard, so as to calculate the collision risk value for the host vehicle. Brief Description of the Drawings

[0017] Figure 1 is a schematic flowchart of a method for fusing multi-source risk factors of a road based on autonomous driving provided by the present invention;

[0018] Figure 2 is an example diagram of road traffic elements provided by the present invention;

[0019] Figure 3 is another schematic flowchart of a method for fusing multi-source risk factors of a road based on autonomous driving provided by the present invention. Detailed Embodiments

[0020] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0021] Embodiment 1

[0022] The present invention provides a method and device for fusing multi-source risk factors of a road based on autonomous driving. The device includes a classification module, a projection module, and a fusion module, and operates through the following steps, as Figure 1 shown.

[0023] Step 1: Classify traffic elements in the autonomous driving environment into three categories: traffic markings, static obstacles, and dynamic obstacles;

[0024] When an autonomous vehicle is driving on an open road, the vehicle not only has to avoid the threat of obstacle collisions but also needs to comply with traffic regulations and drive according to traffic markings. There are various restrictive conditions on the road, such as road boundaries, road markings, traffic rules, and various traffic participants. All these restrictive conditions must be considered in the autonomous driving system, otherwise, the safety and stability of autonomous driving cannot be guaranteed. In urban road traffic, the traffic elements that pose safety / violation risks to autonomous vehicles mainly include boundary information in the road (such as curbs, guardrails, flower beds, etc.), traffic marking information (single solid lines, single dashed lines, double yellow lines, boundary lines, road turning signs, etc.), static obstacles (road construction, roadside parking, etc.), and dynamic traffic participants (motor vehicles, non-motor vehicles, special-shaped vehicles, pedestrians, etc.). These elements are as Figure 2 shown. Therefore, it is necessary to classify these traffic elements first, that is, to classify the traffic elements in the autonomous driving environment into three categories: traffic markings, static obstacles, and dynamic obstacles through a classification module installed on the intelligent driving vehicle.

[0025] Step 2: Project the traffic element of the traffic marking type into the traffic marking grid map. Traffic markings are divided into single dashed lines, single solid lines, and double yellow lines. Determine the risk coefficient Y1 and the influence range N1 of the single dashed line, and the risk coefficient Y2 and the influence range N2 of the single solid line and the double yellow line;

[0026] Step 3: Project the traffic element of the static obstacle type into the static obstacle grid map and determine its risk coefficient Y3 and influence range N3;

[0027] Step 4: Project the traffic element of the dynamic obstacle type into the dynamic obstacle grid map and determine its risk coefficient Y4 and influence range N4;

[0028] These traffic elements will have different degrees of influence on the driving vehicle. For example, traffic markings are a regulatory constraint on the vehicle and will not cause a substantial collision; fixed obstacles such as guardrails and flower beds will collide with the vehicle, but they are static and completely predictable; other traffic participants such as moving vehicles and pedestrians are moving targets, and their movement trajectories are uncertain, posing the greatest risk to the safe operation of autonomous vehicles. Therefore, it is necessary to uniformly divide the risks that these different elements may pose to the vehicle into two parts: the risk coefficient and the influence range. In this application, a projection module connected to the classification module projects these three types of traffic elements, namely traffic markings, static obstacles, and dynamic obstacles, into the traffic marking grid map, the static obstacle grid map, and the dynamic obstacle grid map respectively to obtain their risk coefficients and influence ranges. At the same time, the maximum value of the risk coefficient can be set as Ymax, and the minimum value as 0; the maximum value of the influence range is Nmax, and the minimum value is 0.

[0029] Step 5, fuse the traffic lane marking raster map, the static obstacle raster map, and the dynamic obstacle raster map into a unified occupancy raster map;

[0030] Step 6, in the unified occupancy raster map, the risk coefficient and influence range of any raster point are: Yn = γ attribute , Nn = N attribute , where γ attribute and N attribute respectively represent the target types of the area where the raster is located;

[0031] Step 7, according to the function calculate the collision risk value generated by each grid in the unified raster map for the ego vehicle, where Risk n represents the collision risk value generated by the nth grid p n in the unified raster map for the ego vehicle, δ is a scenario coefficient, Yn represents the risk coefficient of the grid p n , Nn represents the influence range of the grid, and d(p n , ego) represents the distance between the grid and the ego vehicle.

[0032] The fusion module, connected to the projection module, is used to fuse the traffic lane marking raster map, the static obstacle raster map, and the dynamic obstacle raster map into a unified occupancy raster map. In the unified occupancy raster map, the risk coefficient and influence range of any raster point are: Yn = γ attribute , Nn = N attribute , where γ attribute and N attribute respectively represent the target types of the area where the raster is located. According to the function calculate the collision risk value generated by each grid in the unified raster map for the ego vehicle, where Risk n represents the collision risk value generated by the nth grid p n in the unified raster map for the ego vehicle, δ is a scenario coefficient, Yn represents the risk coefficient of the grid p n , Nn represents the influence range of the grid, and d(p n , ego) represents the distance between the grid and the ego vehicle.

[0033] The present invention provides a method and device for fusing multi-source risk factors of roads based on autonomous driving. This technical solution classifies traffic elements in the autonomous driving environment into three categories: traffic markings, static obstacles, and dynamic obstacles. Then, the data of these three categories are respectively projected onto a traffic marking grid map, a static obstacle grid map, and a dynamic obstacle grid map. The risk coefficients and influence ranges of traffic elements in the traffic marking grid map, the static obstacle grid map, and the dynamic obstacle grid map are respectively solved. Subsequently, the results of the traffic marking grid map, the static obstacle grid map, and the dynamic obstacle grid map are fused into the same occupancy grid map as a unified expression of multi-source traffic risk factors. This technical solution is mainly used to solve the problem in the prior art that traffic risk elements with different categories, different sources, and different dimensions encountered by autonomous vehicles on the road cannot be quantified.

[0034] Embodiment 2

[0035] The present invention provides a method and device for fusing multi-source risk factors of roads based on autonomous driving. The device includes a classification module, a projection module, and a fusion module, and operates through the following steps, as Figure 1 shown.

[0036] Step 1: Classify traffic elements in the autonomous driving environment into three categories: traffic markings, static obstacles, and dynamic obstacles;

[0037] Step 2: Project traffic elements of the traffic marking category onto the traffic marking grid map. Traffic markings are divided into single dotted lines, single solid lines, and double yellow lines. Determine the risk coefficient Y1 and influence range N1 of the single dotted line, and the risk coefficient Y2 and influence range N2 of the single solid line and the double yellow line;

[0038] The passable area of the road does not pose a risk to the vehicle, and both the risk coefficient and the influence range are 0. The single dotted line on the road is a marking that separates different lanes. It restricts the vehicle to drive within the lane but does not limit the vehicle from crossing. Therefore, both the risk coefficient Y1 and the influence range N1 of the lane line are 0. The double yellow line and the single solid line are separating lines for opposite lanes in urban roads. Vehicles cannot cross them, but they do not directly pose a collision risk themselves. They only limit the driving range of the vehicle. Therefore, the vehicle can approach infinitely. The risk coefficient Y2 and the influence range N2 of the double yellow line and the single solid line. Thus, the risk coefficient Y1 = 0, the influence range N1 = 0, the risk coefficient Y2 = Ymax, and the influence range N2 = 0.

[0039] Step 3: Project traffic elements of the static obstacle category onto the static obstacle grid map and determine its risk coefficient Y3 and influence range N3;

[0040] Static obstacles include parked vehicles, flower beds, curbs, construction, etc. on the roadside, which are all classified as static obstacles without distinction. Moreover, static obstacles themselves cannot be collided with, vehicles cannot pass through them, and they will pose a collision risk to the surrounding due to perception errors. For traffic safety, vehicles need to approach these obstacles carefully. The functional formulas for determining their risk coefficient Y3 and influence range N3 are as follows:

[0041] Y3 = γ max ,

[0042]

[0043] if d(obs,ego) < dis min , d(obs,ego) = dis min ,

[0044] wherein, represents the lateral / normal influence range of the obstacle in the vehicle driving direction, represents the longitudinal / tangential influence range of the obstacle in the vehicle driving direction, α lat represents the coefficient of the static obstacle in the lateral / normal direction of the vehicle driving, α lon represents the coefficient of the static obstacle in the longitudinal / tangential direction of the vehicle driving, Monte_Carlo(polygon_area) represents the area of the irregular static obstacle obtained by Monte Carlo sampling method, obs in d(obs,ego) represents the static obstacle, ego represents the host vehicle, and d(obs,ego) represents the distance between the host vehicle and the obstacle, represents the lateral deviation vector of the static obstacle from the host vehicle driving direction, represents the longitudinal deviation vector of the static obstacle from the host vehicle driving direction. When the distance between the host vehicle and the static obstacle is less than the preset minimum distance threshold dis min , d(obs,ego) = dis min .

[0045] Step 4: Project the traffic elements of the dynamic obstacle category onto the dynamic obstacle grid map, and determine its risk coefficient Y4 and influence range N4;

[0046] Dynamic obstacles include vehicles, pedestrians, non-motor vehicles, etc. Since dynamic obstacles have speeds, there are collision risks in the spaces in their speed directions, and the collision risk range will be amplified due to detection and prediction errors in perception. The functional formulas for determining their risk coefficient Y4 and influence range N4 are as follows:

[0047] Y4 = γ max ,

[0048]

[0049] v n = (v ego × (v obs - v ego )),

[0050] v t = (v ego · (v obs - v ego )),

[0051] wherein, represents the lateral / normal influence range of the dynamic obstacle in the direction of its own driving speed, represents the longitudinal / tangential influence range of the dynamic obstacle in the direction of its own driving speed, α lat represents the coefficient of the dynamic obstacle in the lateral / normal direction of its own speed, α lon represents the coefficient of the dynamic obstacle in the longitudinal / tangential direction of the vehicle driving, Monte_Carlo(polygon_area) represents the area of the irregular dynamic obstacle obtained by the Monte Carlo sampling method, obs in d(obs,ego) represents the dynamic obstacle, ego represents the host vehicle, and d(obs,ego) represents the distance between the host vehicle and the obstacle. represents the lateral deviation vector of the dynamic obstacle in the direction of its own driving, represents the longitudinal deviation vector of the dynamic obstacle in the direction of its own driving, ‖v obs ‖ represents the modulus of the obstacle speed vector, that is, its speed value, v n calculates the normal vector difference between the two speed vectors of the host vehicle speed and the obstacle speed, v t calculates the tangential vector difference between the two speed vectors of the host vehicle speed and the obstacle speed, ‖v n ‖ and ‖v t ‖ respectively represent their moduli. When the distance between the host vehicle and the static obstacle is less than the preset minimum distance threshold dis min d(obs,ego) = dis min .

[0052] Step 5, fuse the traffic marking grid map, the static obstacle grid map, and the dynamic obstacle grid map into a unified occupancy grid map;

[0053] Step 6, in the unified occupancy grid map, the risk coefficient and influence range of any grid point are: Yn = γ attribute , Nn = N attribute , wherein γ attribute and N attributerespectively represent the target types of the areas where the grids are located;

[0054] Step 7, according to the function calculate the collision risk value generated by each grid in the unified grid map for the ego vehicle, where Risk n represents the collision risk value generated by the nth grid p n in the unified grid map for the ego vehicle, δ is a scenario coefficient, Yn represents the risk coefficient of the grid p n , and Nn represents the influence range of the grid, d(p n , ego) represents the distance between the grid and the ego vehicle.

[0055] The present invention provides a method and device for fusing multi-source risk factors on roads based on autonomous driving. This technical solution classifies traffic elements in the autonomous driving environment into three categories: traffic markings, static obstacles, and dynamic obstacles, and projects these three types of data onto a traffic marking grid map, a static obstacle grid map, and a dynamic obstacle grid map respectively. Then, it separately solves the risk coefficients and influence ranges of traffic elements in the traffic marking grid map, the static obstacle grid map, and the dynamic obstacle grid map, and finally fuses the results of the traffic marking grid map, the static obstacle grid map, and the dynamic obstacle grid map into the same occupancy grid map as a unified expression of multi-source traffic risk factors. This technical solution is mainly used to solve the problem in the prior art that traffic risk elements with different categories, different sources, and different dimensions encountered by autonomous driving vehicles on roads cannot be quantified.

[0056] Embodiment III

[0057] The present invention provides a method and device for fusing multi-source risk factors on roads based on autonomous driving. The device includes a classification module, a projection module, and a fusion module, and operates through the following steps, as Figure 3 shown.

[0058] Step 1, classify traffic elements in the autonomous driving environment into three categories: traffic markings, static obstacles, and dynamic obstacles;

[0059] Step 2, project traffic marking traffic elements onto a traffic marking grid map. Traffic markings are divided into single dotted lines, single solid lines, and double yellow lines. Determine the risk coefficient Y1 and influence range N1 of the single dotted line, and the risk coefficient Y2 and influence range N2 of the single solid line and the double yellow line;

[0060] Step 3, project static obstacle traffic elements onto a static obstacle grid map and determine their risk coefficient Y3 and influence range N3;

[0061] As can be seen from the formulas for the risk coefficient Y3 and the influence range N3 described above, the influence range is directly proportional to the area of the static obstacle and inversely proportional to the distance from the static obstacle to the host vehicle. That is to say, the larger the area of the static obstacle, the larger its corresponding influence range. The closer the static obstacle is to the host vehicle, the larger its corresponding influence range; the farther the distance, the smaller its corresponding influence range.

[0062] Step 4: Project the dynamic obstacle type traffic elements onto the dynamic obstacle grid map, and determine its risk coefficient Y4 and influence range N4;

[0063] As can be seen from the formulas for the risk coefficient Y4 and the influence range N4 described above, the influence range is directly proportional to the area of the dynamic obstacle, inversely proportional to the distance from the dynamic obstacle to the host vehicle, directly proportional to the speed of the dynamic obstacle, and inversely proportional to the speed difference between the dynamic obstacle and the host vehicle. That is to say, the larger the area of the dynamic obstacle, the larger its corresponding influence range; the closer the dynamic obstacle is to the host vehicle, the larger its corresponding influence range; the farther the distance, the smaller its corresponding influence range; the larger the speed value of the dynamic obstacle, the larger its influence range.

[0064] The greater the speed difference between the dynamic obstacle and the host vehicle, the greater the influence range in its corresponding speed direction.

[0065] Step 5: Integrate the traffic marking grid map, the static obstacle grid map, and the dynamic obstacle grid map into a unified occupancy grid map;

[0066] Step 6: In the unified occupancy grid map, the risk coefficient and influence range of any grid point are: Yn = γ attribute , Nn = N attribute , where γ attribute and N attribute respectively represent the target types of the area where the grid is located;

[0067] Since the target types of the area where the grid is located include dashed lines and passable areas, single solid lines and double yellow lines, static obstacles, and dynamic obstacles, the said Step 6 further includes: when the same grid point contains multiple target types, the fusion module selects the target type with the highest priority according to the following priority order, Y4>Y3>Y2>Y1, N4>N3>N2>N1.

[0068] Step 7: Calculate the collision risk value generated by each grid in the unified grid map for the host vehicle according to the function , where Risk n represents the collision risk value generated by the nth grid p n in the unified grid map for the host vehicle, δ is a scenario coefficient, Yn represents the risk coefficient of the grid p n , and Nn represents the influence range of the grid, d(pn , ego) represents the distance between the grid and the vehicle itself.

[0069] The present invention provides a method and device for fusing multi-source risk factors of roads based on autonomous driving. This technical solution classifies traffic elements in the autonomous driving environment into three categories: traffic markings, static obstacles, and dynamic obstacles, and projects the data of these three categories into a traffic marking grid map, a static obstacle grid map, and a dynamic obstacle grid map respectively. Then, it calculates the danger coefficients and influence ranges of traffic elements in the traffic marking grid map, the static obstacle grid map, and the dynamic obstacle grid map, and finally fuses the results of the traffic marking grid map, the static obstacle grid map, and the dynamic obstacle grid map into the same occupancy grid map as a unified expression of multi-source traffic risk factors. This technical solution is mainly used to solve the problem in the prior art that traffic risk elements with different categories, different sources, and different dimensions encountered by autonomous driving vehicles on the road cannot be quantified.

[0070] In summary, the embodiments of the present invention provide a method and device for fusing multi-source risk factors of roads based on autonomous driving, which quantify traffic risk elements with different categories, different sources, and different dimensions encountered by autonomous driving vehicles on the road; calculate the danger coefficients and influence ranges of different traffic risk elements; fuse these different traffic risk elements, express them on the grid map with a unified quantification standard, and calculate the collision risk value for the vehicle itself.

[0071] The above are only the preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present invention shall be included within the scope of protection of the present invention.

Claims

1. A road multi-source risk factor fusion method based on autonomous driving, characterized in that: The method comprises: Step 1: Classify traffic elements in the autonomous driving environment into three categories: traffic markings, static obstacles, and dynamic obstacles; Step 2: Project traffic elements of traffic markings into a traffic marking grid map. Traffic markings are divided into single dashed lines, single solid lines, and double yellow lines. Determine the hazard coefficient Y1 and impact range N1 of single dashed lines, and the hazard coefficient Y2 and impact range N2 of single solid lines and double yellow lines. Step 3, project the static obstacle traffic elements into the static obstacle grid map to determine their risk factor Y3 and impact range N3; Step 4, project the dynamic obstacle traffic elements into the dynamic obstacle grid map to determine its risk factor Y4 and impact range N4; Step 5, integrating the traffic line grid map, the static obstacle grid map, and the dynamic obstacle grid map into a unified occupied grid map; Step 6: In the uniformly occupied grid map, the risk factor and influence range of any grid point is: Yn = γ attribute , Nn=N attribute , where γ attribute and N attribute They represent the risk factor and impact range of the target type in the area where the grid is located; Step 7, according to the function The collision risk value of each grid in the unified grid map to the vehicle is calculated, where Risk n Represents the nth grid p in the unified grid map n For the collision risk value generated by the vehicle, δ represents the scene coefficient, and Yn represents the grid p n The risk factor is Nn, which represents the influence range of the grid. n ,ego) represents the distance between the grid and the ego vehicle.

2. According to the method for fusing multi-source risk factors of roads based on autonomous driving according to claim 1, it is characterized in that: The risk coefficient Y1=0, the impact range N1=0, the risk coefficient Y2=Ymax, the impact range N2=0.

3. According to the method for fusing multi-source risk factors of roads based on autonomous driving in claim 1, it is characterized in that: The function formula for determining the risk factor Y3 and the impact range N3 is: Y3=γ max , if d(obs,ego)<dis min ,d(obs,ego)=dis min , in, Indicates the lateral / normal influence range of the obstacle in the direction of vehicle travel. Indicates the longitudinal / tangential influence range of the obstacle in the vehicle's travel direction, α lat Represents the coefficient of the static obstacle in the lateral / normal direction of the vehicle, α lon represents the coefficient of the static obstacle in the longitudinal / tangential direction of the vehicle. Monte_Carlo(polygon_area) represents the area of ​​the irregular static obstacle obtained by Monte Carlo sampling method. In d(obs,ego), obs represents the static obstacle, ego represents the ego vehicle, and d(obs,ego) represents the distance between the ego vehicle and the obstacle. It represents the lateral deviation vector between the static obstacle and the vehicle’s travel direction. It represents the longitudinal deviation vector between the static obstacle and the vehicle's travel direction.

4. According to the method for fusing multi-source risk factors of roads based on autonomous driving in claim 1, it is characterized in that: The function formula for determining the risk factor Y4 and the impact range N4 is: Y4=γ max , push min ,d(obs,ego)=dis min , v n =(v ego ×(v obs -v ego )), v t =(v ego ·(v obs -v ego )), in, Indicates the lateral / normal influence range of the dynamic obstacle in the direction of its own driving speed. Indicates the longitudinal / tangential influence range of the dynamic obstacle in the direction of its own driving speed, α lat Represents the coefficient of the dynamic obstacle in the lateral / normal direction of its own velocity, α lon It represents the coefficient of the dynamic obstacle in the longitudinal / tangential direction of the vehicle. Monte_Carlo(polygon_area) represents the area of ​​the irregular dynamic obstacle obtained by Monte Carlo sampling method. In d(obs,ego), obs represents the dynamic obstacle, ego represents the ego vehicle, and d(obs,ego) represents the distance between the ego vehicle and the obstacle. It represents the lateral deflection vector of the dynamic obstacle in its own traveling direction. represents the longitudinal deflection vector of the dynamic obstacle in its own driving direction, ‖v obs ‖ represents the modulus of the obstacle’s velocity vector, which is the magnitude of its velocity, v n The normal vector difference between the vehicle speed and the obstacle speed is calculated, v t The tangent vector difference between the vehicle speed and the obstacle speed is calculated, ‖v n ‖ and ‖v t ‖ represent their modules respectively.

5. A method for fusing multi-source risk factors of roads based on autonomous driving according to claim 3 or 4, characterized in that: The distance between the vehicle and the static obstacle is less than the preset minimum distance threshold value dis min When d(obs,ego)=dis min .

6. According to the method for fusing multi-source risk factors of roads based on autonomous driving in claim 1, it is characterized in that: The influence range is proportional to the area of ​​the static obstacle and inversely proportional to the distance from the static obstacle to the vehicle.

7. According to claim 1, a method for fusing multi-source risk factors of roads based on autonomous driving is characterized in that: The influence range is proportional to the area of ​​the dynamic obstacle, inversely proportional to the distance from the dynamic obstacle to the vehicle, proportional to the speed of the dynamic obstacle, and inversely proportional to the speed difference between the dynamic obstacle and the vehicle.

8. According to the method for fusing multi-source risk factors of roads based on autonomous driving in claim 1, it is characterized in that: The step 6 also includes: when the same grid point contains multiple target types, the target type with the highest priority is selected according to the following priority order: Y4>Y3>Y2>Y1, N4>N3>N2>N1.

9. A device for implementing the method for fusing multi-source road risk factors based on autonomous driving according to claims 1-8, characterized in that: The device comprises: The classification module is set on the intelligent driving vehicle and is used to classify the traffic elements in the autonomous driving environment into three categories: traffic markings, static obstacles, and dynamic obstacles; The projection module is connected to the classification module and is used to project the three types of traffic elements, namely, traffic markings, static obstacles, and dynamic obstacles, into the traffic marking grid map, the static obstacle grid map, and the dynamic obstacle grid map, respectively, and obtain their hazard coefficients and impact ranges; The fusion module is connected to the projection module and is used to fuse the traffic marking grid map, the static obstacle grid map and the dynamic obstacle grid map into a unified occupied grid map. In the unified occupied grid map, the hazard coefficient and influence range of any grid point are: Yn = γ attribute , Nn=N attribute , where γ attribute and N attribute Respectively represent the risk factor and impact range of the target type in the area where the grid is located. According to the function The collision risk value of each grid in the unified grid map to the vehicle is calculated, where Risk n Represents the nth grid p in the unified grid map n For the collision risk value generated by the vehicle, δ is a scenario coefficient, and Yn represents the grid p n The risk factor is Nn, which represents the influence range of the grid. n ,ego) represents the distance between the grid and the ego vehicle.

10. The device for fusing multi-source road risk factors based on autonomous driving according to claim 9, characterized in that: The fusion module is also used to select the target type with the highest priority according to the following priority order when the same grid point contains multiple target types: Y4>Y3>Y2>Y1, N4>N3>N2>N1.

Images