Method and device for predicting and deciding autonomous driving of vehicle

Abstract

The application relates to the technical field of automatic driving decision planning, in particular to an automatic driving prediction decision method and device for a vehicle, wherein the method comprises the following steps: acquiring state information of obstacles around the vehicle; acquiring center lines of a lane where the vehicle is located and surrounding lanes; obtaining points closest to the center lines of the lanes corresponding to the surrounding obstacles through traversal; obtaining coordinate information of a signal lamp in combination with position information of the signal lamp and a remaining time length of a green light; generating a predicted trajectory of the surrounding obstacles according to a predicted acceleration and a lane-changing intention; calculating probabilities generated by all interaction scenes according to an intention of a vehicle of interest and the predicted trajectory; calculating a planning trajectory in parallel to calculate a scene occurrence probability and a trajectory quality score, and obtaining a comprehensive evaluation index; and then selecting a final optimal trajectory. Therefore, the problems that in the related art, the intention of an obstacle is difficult to be non-deterministically predicted, it is difficult to adapt to complex traffic environments, and safe, reliable and stable decision suggestions cannot be provided are solved.

Description

Technical Field

[0001] This application relates to the field of decision planning technology, and in particular to a predictive decision-making method and apparatus for autonomous driving of vehicles. Background Technology

[0002] Autonomous driving refers to vehicles that achieve safe and automatic operation through technologies such as artificial intelligence without any active human intervention. While autonomous driving brings convenience to people's lives, it also causes a large number of traffic accidents, thus increasing investment in research on trajectory prediction of surrounding obstacles and vehicle behavior decision-making in real life.

[0003] Among related technologies, independent research on trajectory prediction and behavioral decision-making, rule-based prediction methods, and deep learning-based prediction methods can be adopted. Among them, independent research on trajectory prediction and behavioral decision-making mainly involves not considering the impact of the vehicle's decision-making behavior on the trajectory prediction of obstacles, and not considering the trajectory of obstacles when making behavioral decisions for the vehicle. Rule-based prediction methods mainly involve simulating motion based on the vehicle's physical model to generate a stable avoidance strategy. Deep learning-based prediction methods continuously improve the accuracy of the model by training and testing on a large dataset.

[0004] However, in related technologies, it is difficult to make nondeterministic predictions about the intention of obstacles, and special scenarios such as traffic lights are not considered. Furthermore, they are highly dependent on models and datasets, resulting in poor interpretability of prediction results. They are difficult to adapt to complex traffic environments and cannot provide safe, reliable, and stable decision-making suggestions, which urgently need improvement. Summary of the Invention

[0005] This application provides a predictive decision-making method and apparatus for autonomous driving of vehicles to solve the problems in related technologies, such as the difficulty in making uncertain predictions of the intentions of obstacles, the lack of consideration for special scenarios such as traffic lights, the difficulty in adapting to complex traffic environments, and the inability to provide safe, reliable, and stable decision-making suggestions.

[0006] The first aspect of this application provides a predictive decision-making method for autonomous driving of a vehicle, comprising: acquiring the state information of obstacles around the vehicle at the current moment; acquiring lane information of the lane currently occupied by the vehicle and surrounding lane information to obtain the centerline of each lane edge; using the centerline as a reference line, traversing each obstacle around the vehicle, and finding the closest point on the centerline of the corresponding lane based on the state information of each obstacle around the vehicle; acquiring the position information and remaining green light duration of a traffic light, and obtaining the coordinate information of the traffic light based on the closest point on the centerline; predicting the acceleration of each obstacle around the vehicle based on the remaining green light duration and the coordinate information, and identifying the change in the acceleration of each obstacle around the vehicle. The system calculates the lane change intention, generates a predicted trajectory for obstacles around each vehicle based on the acceleration and lane change intention; acquires at least one vehicle of interest based on the actual position of the vehicle, calculates the probability of occurrence of different interaction scenarios based on the intention and predicted trajectory of the at least one vehicle of interest, and calculates the probability of generation of all interaction scenarios; calculates the planned trajectory of the vehicle in multiple scenarios in parallel to evaluate the quality score of the planned trajectory in different scenarios, calculates the scenario occurrence probability and trajectory quality score in each scenario as a comprehensive evaluation index for decision planning in the corresponding scenario; obtains the largest comprehensive evaluation index among the multiple scenarios based on the comprehensive evaluation index, and takes the planned trajectory corresponding to the largest comprehensive evaluation index as the final optimal trajectory.

[0007] Optionally, in one embodiment of this application, the formula for calculating the centerline of each lane edge line can be:

[0008]

[0009] in, The x-coordinate, y-coordinate, angle with the x-axis, curvature, and cumulative distance of the i-th point on the lane centerline are respectively given. The x-coordinate, y-coordinate, angle with the x-axis, curvature, and cumulative distance of the i-th point on the left side of the lane are respectively given. The x-coordinate, y-coordinate, angle with the x-axis, curvature, and cumulative distance of the i-th point on the left side of the lane are respectively given.

[0010] Optionally, in one embodiment of this application, the formula for predicting the acceleration of each obstacle around the vehicle can be:

[0011]

[0012] Among them, a max v is the estimated maximum acceleration of the obstacle. i Let v be the velocity of the i-th obstacle, v0 be the velocity of the obstacle in front of the obstacle in the lane where the obstacle is located, and d be the velocity of the obstacle.* (v i Δv) is the desired distance between the obstacle and the vehicle in front, Δv is the speed difference between the obstacle and the vehicle in front, and d is the distance between the obstacle and the obstacle in front in the lane. g I represents the current distance between the obstacle and the traffic light. A (d g v max , t g ) is an indicator function, v max For lane speed limits, t g θ1 represents the remaining green light duration, θ2 represents the following vehicle weight, and θ2 represents the traffic light weight.

[0013] Optionally, in one embodiment of this application, identifying the lane-changing intention of each obstacle around the vehicle includes: traversing the obstacles around the vehicle in each lane and obtaining the coordinate positions of the obstacles around the vehicle; predicting the lane-changing intention of the obstacles based on the coordinate positions of the obstacles around the vehicle using a preset priority, wherein the lane-changing benefit value of the corresponding lane is obtained according to the number of obstacles around each vehicle and the corresponding courtesy factor to determine the lane-changing intention.

[0014] Optionally, in one embodiment of this application, the formula for calculating the lane-changing profit value can be:

[0015]

[0016]

[0017] u ke =a th ,

[0018]

[0019]

[0020] in, The original calculated value for profit from left lane changing. a is the expected acceleration value of the obstacle after changing lanes to the left. i The expected acceleration value of the obstacle in the current lane. a is the expected acceleration value of the vehicle following in the left lane after a left lane change. lfv This represents the expected acceleration value of the vehicle following in the left lane before the left lane change. a is the expected acceleration value of the vehicle following in the current lane after a left lane change. fv Let p be the expected acceleration of the vehicle behind in the current lane before the left lane change, and p be the yield factor. The original calculated value for profit from right lane changing. a is the expected acceleration value of the obstacle after changing lanes to the right. iThe expected acceleration value of the obstacle in the current lane. This represents the expected acceleration value of the vehicle following in the right lane after a right lane change. Let a be the expected acceleration value of the vehicle following in the current lane after a right lane change, ReLU be the activation function, and a be the expected acceleration value of the vehicle following in the current lane. th To determine the profit threshold for lane changing, u le The final value of the profit from a left lane change, u re The final value of the profit from right lane change, u ke Maintain profitability for the lane.

[0021] Optionally, in one embodiment of this application, the formula for calculating the comprehensive evaluation index can be:

[0022] U i =p i *F i ,

[0023] Where, p i F represents the probability of behavior for each obstacle. i Let be the quality score of the i-th trajectory.

[0024] A second aspect of this application provides an autonomous driving prediction and decision-making device for a vehicle, characterized by comprising: a first acquisition module, configured to acquire state information of obstacles surrounding the vehicle at the current moment; a second acquisition module, configured to acquire lane information of the current lane of the vehicle and surrounding lane information to obtain the centerline of each lane edge; a traversal module, configured to traverse each obstacle surrounding the vehicle using the centerline as a reference line, and find the nearest point on the centerline of the corresponding lane based on the state information of each obstacle surrounding the vehicle; a third acquisition module, configured to acquire the position information and remaining green light duration of a traffic light, and obtain the coordinate information of the traffic light based on the nearest point on the centerline; and a prediction module, configured to predict the acceleration of each obstacle surrounding the vehicle based on the remaining green light duration and the coordinate information, and identify the... The system calculates the lane-changing intentions of obstacles around each vehicle to generate predicted trajectories for each obstacle based on the acceleration and the lane-changing intentions. A first calculation module is used to acquire at least one vehicle of interest based on the actual position of the vehicle, calculate the probability of occurrence of different interaction scenarios based on the intentions and predicted trajectories of the at least one vehicle of interest, and calculate the probability of generation of all interaction scenarios. A second calculation module is used to calculate the planned trajectories of the vehicle in multiple scenarios in parallel to evaluate the quality scores of the planned trajectories in different scenarios, and to calculate the scenario occurrence probability and trajectory quality score for each scenario as a comprehensive evaluation index for decision-making and planning in the corresponding scenario. A generation module is used to obtain the largest comprehensive evaluation index among the multiple scenarios based on the comprehensive evaluation index, and to use the planned trajectory corresponding to the largest comprehensive evaluation index as the final optimal trajectory.

[0025] Optionally, in one embodiment of this application, the formula for calculating the centerline of each lane edge line can be:

[0026]

[0027] in, The x-coordinate, y-coordinate, angle with the x-axis, curvature, and cumulative distance of the i-th point on the lane centerline are respectively given. The x-coordinate, y-coordinate, angle with the x-axis, curvature, and cumulative distance of the i-th point on the left side of the lane are respectively given. The x-coordinate, y-coordinate, angle with the x-axis, curvature, and cumulative distance of the i-th point on the left side of the lane are respectively given.

[0028] Optionally, in one embodiment of this application, the formula for predicting the acceleration of each obstacle around the vehicle can be:

[0029]

[0030] Among them, amax v is the estimated maximum acceleration of the obstacle. i Let v be the velocity of the i-th obstacle, v0 be the velocity of the obstacle in front of the obstacle in the lane where the obstacle is located, and d be the velocity of the obstacle. * (v i Δv) represents the desired distance between the obstacle and the vehicle in front, where Δv is the speed difference between the obstacle and the vehicle in front. d d is the distance between the obstacle and the obstacle in front in the lane. g I represents the current distance between the obstacle and the traffic light. A (d g v max , t g ) is an indicator function, v max For lane speed limits, t g θ1 represents the remaining green light duration, θ2 represents the following vehicle weight, and θ2 represents the traffic light weight.

[0031] Optionally, in one embodiment of this application, the prediction module includes: an acquisition unit, configured to traverse the obstacles around the vehicle in each lane and acquire the coordinate positions of the obstacles around the vehicle; and a prediction unit, configured to predict the lane-changing intention of the obstacles based on the coordinate positions of the obstacles around the vehicle using a preset priority, wherein the lane-changing benefit value of the corresponding lane is acquired according to the number of obstacles around each vehicle and the corresponding courtesy factor to determine the lane-changing intention.

[0032] Optionally, in one embodiment of this application, the formula for calculating the lane-changing profit value can be:

[0033]

[0034]

[0035] u ke =a th ,

[0036]

[0037]

[0038] in, The original calculated value for profit from left lane changing. a is the expected acceleration value of the obstacle after changing lanes to the left. i The expected acceleration value of the obstacle in the current lane. a is the expected acceleration value of the vehicle following in the left lane after a left lane change. lfv This represents the expected acceleration value of the vehicle following in the left lane before the left lane change. a is the expected acceleration value of the vehicle following in the current lane after a left lane change. fvLet p be the expected acceleration of the vehicle behind in the current lane before the left lane change, and p be the yield factor. The original calculated value for profit from right lane changing. a is the expected acceleration value of the obstacle after changing lanes to the right. i The expected acceleration value of the obstacle in the current lane. This represents the expected acceleration value of the vehicle following in the right lane after a right lane change. Let a be the expected acceleration value of the vehicle following in the current lane after a right lane change, ReLU be the activation function, and a be the expected acceleration value of the vehicle following in the current lane. th To determine the profit threshold for lane changing, u le The final value of the profit from a left lane change, u re The final value of the profit from right lane change, u ke Maintain profitability for the lane.

[0039] Optionally, in one embodiment of this application, the formula for calculating the comprehensive evaluation index can be:

[0040] U i =p i *F i ,

[0041] Where, p i F represents the probability of behavior for each obstacle. i Let be the quality score of the i-th trajectory.

[0042] A third aspect of this application provides an electronic device, including: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the autonomous driving prediction and decision-making method for a vehicle as described in the above embodiments.

[0043] A fourth aspect of this application provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described method for predictive decision-making of autonomous driving in a vehicle.

[0044] A fifth aspect of this application provides a computer program product, including a computer program that, when executed, implements the above-described vehicle autonomous driving prediction and decision-making method.

[0045] This application embodiment can traverse the obstacles around each vehicle to find the nearest point on the centerline of the corresponding lane, obtain the coordinate information of the corresponding traffic light, and then predict the acceleration of the obstacles around each vehicle. Combined with lane-changing intentions, a predicted trajectory is generated. The probability of different interaction scenarios is calculated based on the intentions of the vehicles of interest, thus obtaining the probability of all interaction scenarios. By parallel computing the planned trajectories of the vehicle under multiple scenarios, the probability of scenario occurrence and trajectory quality score for each scenario are obtained, achieving the highest comprehensive evaluation index among multiple scenarios, and the corresponding planned trajectory is taken as the final optimal trajectory. This application embodiment can confirm the non-deterministic intentions of surrounding obstacles during autonomous driving decision-making, and through parallel spatiotemporal methods, deduce possible interaction scenarios, making reasonable decisions in each corresponding scenario, thereby achieving vehicle driving behavior decision-making through comprehensive evaluation. Therefore, it solves the problems in related technologies, such as the difficulty in non-deterministic prediction of obstacle intentions, the lack of consideration for special scenarios such as traffic lights, the difficulty in adapting to complex traffic environments, and the inability to provide safe, reliable, and stable decision-making suggestions.

[0046] Additional aspects and advantages of this application will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of this application. Attached Figure Description

[0047] The above and / or additional aspects and advantages of this application will become apparent and readily understood from the following description of the embodiments taken in conjunction with the accompanying drawings, wherein:

[0048] Figure 1 This is a flowchart of an autonomous driving prediction and decision-making method for a vehicle according to an embodiment of this application;

[0049] Figure 2 This is a block diagram illustrating a scenario where there is only one vehicle in front of a vehicle according to an embodiment of this application;

[0050] Figure 3 This is a block diagram illustrating the filtering of obstacles of interest when a vehicle has lanes on both the left and right sides, according to an embodiment of this application.

[0051] Figure 4 This is a block diagram illustrating the filtering of obstacles of interest for a vehicle that exists only in the left lane, according to one embodiment of this application.

[0052] Figure 5 This is a block diagram illustrating the filtering of obstacles of interest for a vehicle that exists only in the right lane, according to one embodiment of this application.

[0053] Figure 6This is a block diagram of a vehicle parallel spatiotemporal decision planning processing framework provided according to an embodiment of this application;

[0054] Figure 7 A flowchart illustrating the working principle of an autonomous driving prediction and decision-making method for vehicles according to an embodiment of this application;

[0055] Figure 8 This is a block diagram of an autonomous driving prediction and decision-making device for a vehicle provided according to an embodiment of this application;

[0056] Figure 9 This is a schematic diagram of the structure of an electronic device provided according to an embodiment of this application. Detailed Implementation

[0057] The embodiments of this application are described in detail below. Examples of these embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain this application, and should not be construed as limiting this application.

[0058] The following describes a vehicle autonomous driving prediction and decision-making method and apparatus according to embodiments of this application, with reference to the accompanying drawings. Addressing the problems mentioned in the background art, such as the difficulty in making non-deterministic predictions of the intentions of obstacles, the lack of consideration for special scenarios like traffic lights, the difficulty in adapting to complex traffic environments, and the inability to provide safe, reliable, and stable decision-making suggestions, this application provides an autonomous driving prediction and decision-making method for vehicles. In this method, the system traverses each obstacle surrounding the vehicle to find the closest point on the centerline of the corresponding lane, obtains the coordinate information of the corresponding traffic light, and then predicts the acceleration of each obstacle surrounding the vehicle. Combined with lane-changing intentions, a predicted trajectory is generated. The probability of occurrence of different interaction scenarios is calculated based on the intentions of the vehicle of interest, thus obtaining the probability of generation of all interaction scenarios. By parallel computing the planned trajectories of the vehicle under multiple scenarios, the probability of scenario occurrence and trajectory quality score for each scenario are obtained, achieving the acquisition of the highest comprehensive evaluation index among multiple scenarios, and using its corresponding planned trajectory as the final optimal trajectory. Embodiments of this application can confirm the non-deterministic intentions of surrounding obstacles during autonomous driving decision-making and, through parallel spatiotemporal methods, deduce possible interaction scenarios, making reasonable decisions in the corresponding scenarios, thereby achieving a comprehensive evaluation-based decision on the vehicle's driving behavior. This solves the problems in related technologies, such as the difficulty in making uncertain predictions about the intentions of obstacles, the lack of consideration for special scenarios such as traffic lights, the difficulty in adapting to complex traffic environments, and the inability to provide safe, reliable, and stable decision-making suggestions.

[0059] Specifically, Figure 1This is a flowchart of an autonomous driving prediction and decision-making method for a vehicle provided according to an embodiment of this application.

[0060] like Figure 1 As shown, the autonomous driving prediction and decision-making method for this vehicle includes the following steps:

[0061] In step S101, the state information of obstacles around the vehicle at the current moment is obtained.

[0062] It is understood that obstacles around the vehicle can be obstacles within a certain range (e.g., 1 meter, 30 centimeters, this application does not make specific limitations) on the left, right, front, and rear sides of the vehicle, or obstacles in a certain direction within a certain range of the vehicle, such as obstacles in front. The specific setting can be determined by those skilled in the art according to the actual situation, and this application does not make specific limitations. Obstacles can be, but are not limited to, other vehicles, traffic warning lines, animals, etc., and the specific setting can be determined by those skilled in the art according to the actual situation, and this application does not make specific limitations.

[0063] Furthermore, in this embodiment, the obstacle's state information may include, but is not limited to, the obstacle's location information (x i y i ), velocity v i , heading angle θ i and the obstacle's identification ID i This application does not impose specific limitations. Assume that there are a total of m obstacles in this embodiment, and i represents the i-th obstacle, then i∈[1,m].

[0064] In some embodiments, the present application can use sensing devices such as cameras and millimeter-wave radar to acquire the state information of obstacles, or other devices. The specific settings can be made by those skilled in the art according to the actual situation, and the present application does not impose any specific limitations.

[0065] As one possible approach, embodiments of this application can use relevant equipment, such as millimeter-wave radar (this application does not impose specific limitations), to obtain the status information of all obstacles within a certain range around a vehicle, whether it is moving or stationary.

[0066] For example, such as Figure 2 As shown in the embodiment, this application assumes that the vehicle is traveling in the left lane of a two-lane road, and there is an obstacle vehicle in front of it. Information about the obstacle is obtained through relevant equipment such as millimeter-wave radar (this application does not impose specific limitations), which may include the location of the obstacle (x). l y l ), velocity v l , heading angle θ l and the obstacle's identification IDl .

[0067] In step S102, lane information of the vehicle's current lane and surrounding lane information are obtained to calculate the centerline of each lane edge. The formula for calculating the centerline of each lane edge can be:

[0068]

[0069] in, The x-coordinate, y-coordinate, angle with the x-axis, curvature, and cumulative distance of the i-th point on the lane centerline are respectively given. The x-coordinate, y-coordinate, angle with the x-axis, curvature, and cumulative distance of the i-th point on the left side of the lane are respectively given. The x-coordinate, y-coordinate, angle with the x-axis, curvature, and cumulative distance of the i-th point on the left side of the lane are respectively given.

[0070] It is understood that the embodiments of this application may, but are not limited to, using high-precision maps, vehicle navigation, BeiDou, GPS (Global Positioning System), etc., and this application does not impose specific limitations, to obtain lane information of the vehicle's current lane and surrounding lane information. Furthermore, in the embodiments of this application, if there are no lanes around the vehicle, the surrounding vehicle information is empty. In addition, the embodiments of this application can obtain lane line information for up to three lanes: the current lane, the left lane, and the right lane.

[0071] In actual implementation, the embodiments of this application can use high-precision maps or other methods, and this application does not impose specific limitations. The lane information of the vehicle's current lane and the surrounding lane information are obtained, and then the center line of each lane edge is obtained using the calculation formula of the center line of each lane edge.

[0072] In other words, in this embodiment, each lane can be represented by two discrete point sequences. For example, the left lane can be represented by two discrete point sequences: the left edge of the left lane and the right edge of the left lane. Each point can contain corresponding position information, angle information with the x-axis, curvature information, and cumulative length information. This embodiment can utilize... The x-coordinate, y-coordinate, angle with the x-axis, curvature, and cumulative distance of the i-th point on the lane centerline are respectively given. The x-coordinate, y-coordinate, angle with the x-axis, curvature, and cumulative distance of the i-th point on the left side of the lane are respectively given. Let x represent the x-coordinate, y-coordinate, angle with the x-axis, curvature, and cumulative distance of the i-th point on the left edge of the corresponding lane. Then, the formula for calculating the centerline of each lane edge can be:

[0073]

[0074] For example, such as Figure 2 As shown, in this embodiment of the application, the lane information of the vehicle’s current lane and the lane information of the right lane can be obtained through a high-precision map (or other devices, which are not specifically limited in this application). Each lane can be represented by two discrete point sequences, and then the center line information of each lane edge can be obtained by using the calculation formula of the center line of each lane edge.

[0075] In step S103, using the center line as a reference line, each obstacle around the vehicle is traversed, and the point closest to the center line of the corresponding lane is found based on the status information of each obstacle around the vehicle.

[0076] Those skilled in the art will understand that, in order to find the closest point on the centerline of the lane where the obstacle is located, the embodiments of this application can use the lane centerline as a reference line to traverse every obstacle around the vehicle, and project the coordinates of the obstacles to the Frenet coordinate system using relevant coordinate transformation formulas, thereby obtaining the corresponding s-coordinates and l-coordinates. The coordinate transformation expression can be:

[0077]

[0078] s i =s r

[0079] Where, x i y i Let x be the coordinate of the i-th obstacle. r y r θ r s r For point p r x and y coordinates, angle with the x-axis, and cumulative distance.

[0080] For example, such as Figure 2 As shown, in this embodiment of the application, the lane centerline can be used as a reference line to find the point on the lane centerline closest to the obstacle based on its position, denoted as p. r The obstacle's coordinates are then projected onto the Frenet coordinate system according to the coordinate transformation expression, thus obtaining the corresponding s-coordinates and l-coordinates.

[0081] In step S104, the location information of the traffic light and the remaining green light duration are obtained, and the coordinate information of the traffic light is obtained based on the nearest point on the center line.

[0082] It is understood that the embodiments of this application may, but are not limited to, using cameras, image processing, computer vision, and possible traffic management system interfaces to obtain the location information of traffic lights and the remaining green light duration. The specific settings can be made by those skilled in the art according to the actual situation, and this application does not impose any specific limitations.

[0083] In some embodiments, the present application embodiments can combine technologies such as cameras and image processing to obtain the location information of traffic lights and the remaining green light duration, and combine the coordinate transformation expression in step S103 to obtain the s coordinate and l coordinate corresponding to the traffic lights.

[0084] For example, such as Figure 2 As shown, this application embodiment can combine technologies such as cameras and image processing to obtain the location information of traffic lights and the remaining green light duration t. g And by combining the coordinate transformation expression in step S103, the s-coordinate of the traffic light is obtained. g and the l-coordinate l g .

[0085] In step S105, the acceleration of obstacles around each vehicle is predicted based on the remaining green light duration and coordinate information, and the lane-changing intention of each obstacle around the vehicle is identified, so as to generate a predicted trajectory of each obstacle around the vehicle based on the acceleration and lane-changing intention.

[0086] As one possible implementation, embodiments of this application can combine the predicted acceleration of obstacles around each vehicle with the lane-changing intention to generate the predicted trajectory of each obstacle. The generation process can be as follows:

[0087] In some embodiments of this application, if the obstacle indicates an intention to change lanes to the left, the Frenet coordinates of the lane change endpoint can be obtained using the formula for calculating the Frenet coordinates of the left lane change endpoint, which can be expressed as:

[0088]

[0089] l end =l lc ,

[0090] Among them, s end Let s be the coordinate of the end point of the lane change. start Let v be the current obstacle's s-coordinate. start Let T be the current speed of the obstacle. c The lane change time is manually given, and a(τ) can be calculated from the coordinate transformation expression. end Let l be the coordinate of the lane change endpoint, which is equal to the center line s of the left lane. end The corresponding l-coordinates l lc .

[0091] In some other embodiments, if the obstacle in this application presents an intention to change lanes to the right, the Frenet coordinates of the lane change endpoint can be obtained by the formula for calculating the Frenet coordinates of the right lane change endpoint, which can be expressed as:

[0092]

[0093] l end =l rc ,

[0094] Among them, s end Let s be the coordinate of the end point of the lane change. start Let v be the current obstacle's s-coordinate. start Let T be the current speed of the obstacle. c The lane change time is manually given, and a(τ) is calculated from the coordinate transformation expression. end Let l be the coordinate of the lane change endpoint, which is equal to the center line s of the right lane. end The corresponding l-coordinates l rc .

[0095] In some other embodiments, if the obstacle in the present application does not change lanes, the predicted Frenet coordinates of the endpoint can be obtained by the formula for calculating the Frenet coordinates of the endpoint without changing lanes, which can be expressed as:

[0096]

[0097] l end =l kc ,

[0098] Among them, s end To predict the s-coordinate of the endpoint, s start Let v be the current obstacle's s-coordinate. start Let l be the current speed of the obstacle, T be the prediction time (given manually), and a(τ) be calculated from the coordinate transformation expression. end The coordinate l of the predicted endpoint is equal to the current lane centerline s. end The corresponding l-coordinates l kc .

[0099] Furthermore, in this embodiment, a fifth-order polynomial can be used to connect the starting point and the lane change endpoint or the predicted endpoint to generate a predicted trajectory for each obstacle around the vehicle. Additionally, in this embodiment, if an obstacle has multiple possible intentions, multiple predicted trajectories can be generated. Other methods can also be used to generate the predicted trajectory for each obstacle around the vehicle; this application does not impose specific limitations.

[0100] For example, such as Figure 2As shown, embodiments of this application can generate predicted trajectories for obstacles, and the process can be as follows:

[0101] In the embodiments of this application, if the obstacle indicates an intention to change lanes to the right, the Frenet coordinates of the end point of the right lane change can be obtained from the Frenet coordinates calculation formula of the end point of the right lane change; if the obstacle does not change lanes, the Frenet coordinates of the predicted end point can be obtained from the Frenet coordinates calculation formula of the end point of the non-lane change.

[0102] Furthermore, in this embodiment, a fifth-order polynomial can be used to connect the starting point and the lane change endpoint or the predicted endpoint to generate a predicted trajectory for each obstacle surrounding the vehicle. Additionally, in this embodiment, if the obstacle has multiple possible intentions, multiple predicted trajectories can be generated.

[0103] Optionally, in one embodiment of this application, the formula for predicting the acceleration of each obstacle around the vehicle can be:

[0104]

[0105] Among them, a max v is the estimated maximum acceleration of the obstacle. i Let v be the velocity of the i-th obstacle, v0 be the velocity of the obstacle in front of the obstacle in the lane where the obstacle is located, and d be the velocity of the obstacle. * (v i Δv) is the desired distance between the obstacle and the vehicle in front, Δv is the speed difference between the obstacle and the vehicle in front, and d is the distance between the obstacle and the obstacle in front in the lane. g I represents the current distance between the obstacle and the traffic light. A (d g v max , t g ) is an indicator function, v max For lane speed limits, t g θ1 represents the remaining green light duration, θ2 represents the following vehicle weight, and θ2 represents the traffic light weight.

[0106] As one possible implementation, in order to generate the predicted trajectory of each obstacle around the vehicle, this embodiment of the application can predict the acceleration of each obstacle around the vehicle according to the prediction formula of the acceleration of each obstacle around the vehicle, thereby obtaining the predicted acceleration of the obstacle at the current moment. The prediction formula for the acceleration of each obstacle around the vehicle can be:

[0107]

[0108] Among them, a max v is the estimated maximum acceleration of the obstacle. i Let v be the velocity of the i-th obstacle, v0 be the velocity of the obstacle in front of the obstacle in the lane where the obstacle is located, and d be the velocity of the obstacle.* (v i Δv) is the desired distance between the obstacle and the vehicle in front, Δv is the speed difference between the obstacle and the vehicle in front, and d is the distance between the obstacle and the obstacle in front in the lane. g I represents the current distance between the obstacle and the traffic light. A (d g v max , t g ) is an indicator function, v max For lane speed limits, t g Let θ1 be the remaining green light duration, θ2 be the following weight, and θ1 be the traffic light weight. Furthermore, in this embodiment, θ1∈[0,1], θ2∈[0,1], and θ1+θ2=1.

[0109] Furthermore, in actual implementation, embodiments of this application can predict the acceleration of obstacles based on their s-coordinates, following certain rules (e.g., from largest to smallest, without specific limitations). Where there are no obstacles ahead, v0 represents the lane speed limit, d represents the distance to the traffic light, and a... max Estimates can be made based on vehicle model; this application does not impose specific limitations.

[0110] Furthermore, in the embodiments of this application, d * (v i ,Δv) and I A The expressions can be as follows:

[0111]

[0112]

[0113] In this embodiment of the application, if there is no vehicle in front, Δv is the speed difference from the road speed limit, s0 is the manually set minimum distance between vehicles, T is the manually set safe headway, and b is the comfort deceleration.

[0114] For example, such as Figure 2 As shown, in this embodiment of the application, the acceleration of obstacles can be predicted based on the prediction formula of the acceleration of each obstacle around the vehicle, and finally the predicted acceleration at the current moment can be obtained.

[0115] Optionally, in one embodiment of this application, identifying the lane-changing intention of each obstacle surrounding the vehicle includes: traversing the obstacles surrounding the vehicle in each lane and obtaining the coordinate positions of the obstacles; predicting the lane-changing intention of the obstacles based on the coordinate positions of the obstacles and a preset priority, wherein the lane-changing benefit value of the corresponding lane is obtained according to the number of obstacles surrounding each vehicle and the corresponding courtesy factor to determine the lane-changing intention. The formula for calculating the lane-changing benefit value can be:

[0116]

[0117]

[0118] u ke =a th ,

[0119]

[0120]

[0121] in, The original calculated value for profit from left lane changing. a is the expected acceleration value of the obstacle after changing lanes to the left. i The expected acceleration value of the obstacle in the current lane. a is the expected acceleration value of the vehicle following in the left lane after a left lane change. lfv This represents the expected acceleration value of the vehicle following in the left lane before the left lane change. a is the expected acceleration value of the vehicle following in the current lane after a left lane change. fv Let p be the expected acceleration of the vehicle behind in the current lane before the left lane change, and p be the yield factor. The original calculated value for profit from right lane changing. a is the expected acceleration value of the obstacle after changing lanes to the right. i The expected acceleration value of the obstacle in the current lane. This represents the expected acceleration value of the vehicle following in the right lane after a right lane change. Let a be the expected acceleration value of the vehicle following in the current lane after a right lane change, ReLU be the activation function, and a be the expected acceleration value of the vehicle following in the current lane. th To determine the profit threshold for lane changing, u le The final value of the profit from a left lane change, u re The final value of the profit from right lane change, u ke Maintain profitability for the lane.

[0122] As one possible implementation, embodiments of this application can predict the lane-changing intentions of obstacles, thereby identifying the lane-changing intentions of each obstacle around the vehicle. The prediction process can be as follows:

[0123] First, in this embodiment, the obstacles surrounding the vehicle in each lane are traversed, and the s-coordinates of the obstacles are taken. These obstacles are then prioritized (e.g., from largest to smallest, though this application does not impose specific limitations) to predict the lane-changing intentions of the obstacles sequentially, and the lane-changing intention is determined using a formula for calculating the lane-changing profit value. In this embodiment, The original calculated value for profit from left lane changing. a is the expected acceleration value of the obstacle after changing lanes to the left. iThe expected acceleration value of the obstacle in the current lane. a is the expected acceleration value of the vehicle following in the left lane after a left lane change. lfv This represents the expected acceleration value of the vehicle following in the left lane before the left lane change. a is the expected acceleration value of the vehicle following in the current lane after a left lane change. fv Let be the expected acceleration of the vehicle behind in the current lane before the left lane change, and p be the courtesy factor, which represents the impact of other vehicles on the predicted vehicle. This value can be manually set. The final profit from the left lane change can be calculated using the ReLU function. Processed, The original calculated value for profit from right lane changing. a is the expected acceleration value of the obstacle after changing lanes to the right. i The expected acceleration value of the obstacle in the current lane. a is the expected acceleration value of the vehicle following in the right lane after a right lane change. rfv This represents the expected acceleration of the vehicle following in the right lane before a right lane change. a is the expected acceleration value of the vehicle following in the current lane after a right lane change. fv Given the expected acceleration of the vehicle behind in the current lane before the right lane change, the final gain from the right lane change can be calculated using the ReLU function. The processed values, including the expected acceleration values, can be calculated using the coordinate transformation expression. The lane-keeping profit value is equal to the lane-changing profit threshold 'a'. th This can be set manually. In this case, the formula for calculating the profit value of lane changing in this embodiment of the application can be:

[0124]

[0125]

[0126] u ke =a th ,

[0127]

[0128]

[0129] Furthermore, embodiments of this application can be categorized into the following cases based on the driving environment in which the obstacle to be predicted is located:

[0130] If there are lanes on both the left and right sides of the obstacle, then the embodiments of this application can be based on u le u ke u re The calculation formula calculates the expected lane-changing profit, and then obtains the probability of a left lane change. Lane keeping probability Right lane change probability

[0131] If there is no lane to the left of the obstacle and a lane to the right, then the embodiments of this application can be based on u ke u re The formula is used to calculate the expected lane-changing profit, and then to obtain the left lane-changing probability p. lc =0, lane keeping probability Right lane change probability

[0132] If there is no lane to the right of the obstacle and a lane to the left, then the embodiments of this application can be based on u le u ke The calculation formula calculates the expected lane-changing profit, and then obtains the probability of a left lane change. Lane keeping probability The probability of changing lanes to the right, p rc =0.

[0133] Furthermore, in the embodiments of this application, the left lane change probability p lc Lane keeping probability p ∈[0,1] kc ∈[0,1], right lane change probability p rc ∈[0,1], and p lc +p kc +p rc =1.

[0134] For example, such as Figure 2 As shown, embodiments of this application can predict lane-changing intentions of obstacles in front of the vehicle, by... Figure 2 As can be seen, since there is no lane on the left side of the obstacle and a lane on the right side, this embodiment of the application can be based on u ke u re The formula is used to calculate the expected lane-changing profit, and then to obtain the left lane-changing probability p. lc =0, lane keeping probability Right lane change probability

[0135] In step S106, at least one vehicle of interest is obtained based on the actual location of the vehicle, and the probability of occurrence of different interaction scenarios is calculated based on the intent and predicted trajectory of the at least one vehicle of interest, and the probability of generation of all interaction scenarios is calculated.

[0136] As one possible implementation, embodiments of this application can first determine the vehicle of interest, using the vehicle's actual position as a reference, and then filter the vehicle of interest based on the positions of obstacles around the vehicle. Embodiments of this application can filter the vehicle of interest for different situations, which can be:

[0137] In some embodiments of this application, if lanes exist on both the left and right sides of the vehicle, the selection of vehicles of interest can be as follows: Figure 3 As shown in the diagram. Red vehicles represent vehicles. In this case, the vehicles of interest in this embodiment may include the vehicle in front in the left lane, the vehicle behind in the left lane, the vehicle in front in the current lane, the vehicle in front in the right lane, and the vehicle behind in the right lane.

[0138] In some other embodiments, if the vehicle only has a lane on its left, the selection of vehicles of interest can be as follows: Figure 4 As shown in the diagram. Red vehicles represent vehicles. In this case, the vehicles of interest in this embodiment may include the vehicle in front in the left lane, the vehicle behind in the left lane, and the vehicle in front in the current lane.

[0139] In some other embodiments, if the vehicle only has a lane on the right, the selection of vehicles of interest can be as follows: Figure 5 As shown in the diagram. Red vehicles represent vehicles. In this case, the vehicles of interest in this embodiment may include the vehicle in front in the right lane, the vehicle behind in the right lane, and the vehicle in front in the current lane.

[0140] Then, in this embodiment of the application, the probability of occurrence of different interaction scenarios can be calculated based on the intent of the vehicle of interest and the predicted trajectory. The calculation method can be to calculate the product of the probabilities of the intent of the vehicle of interest, which can be expressed as:

[0141]

[0142] Where N is the number of vehicles of interest, p i The probability of behavior corresponding to each obstacle.

[0143] Furthermore, in the embodiments of this application, such as Figure 1 In the scenario shown, the probability that all vehicles stay in their current lane is the product of the probabilities that all five vehicles of interest stay in their current lane.

[0144] Finally, in this embodiment of the application, the probability of generating all interactive scenarios can be calculated, the M scenarios with the highest probability can be selected, and their corresponding probabilities p can be recorded.

[0145] For example, such as Figure 2 As shown, in this embodiment of the application, the vehicle of interest can be determined first, and then, based on the vehicle's position, the vehicle of interest can be filtered according to the positions of surrounding obstacles. Figure 2In the scenario shown, the vehicle of interest in this embodiment is only one vehicle ahead. Then, based on the intent of the vehicle of interest and the predicted trajectory, this embodiment calculates the probability of different interaction scenarios, which is calculated as the product of the intent probabilities of the vehicle of interest. Finally, this embodiment calculates the probability of all interaction scenarios and selects the two scenarios with the highest probability, recording their corresponding probabilities p1 and p2.

[0146] In step S107, the planned trajectories of the vehicle under multiple scenarios are calculated in parallel to evaluate the quality scores of the planned trajectories under different scenarios. The scenario occurrence probability and trajectory quality score for each scenario are calculated as a comprehensive evaluation index for decision-making and planning in the corresponding scenario. The formula for calculating the comprehensive evaluation index can be:

[0147] U i =p i *F i ,

[0148] Where, p i F represents the probability of behavior for each obstacle. i Let be the quality score of the i-th trajectory.

[0149] As one possible implementation, embodiments of this application can employ parallel computing. That is, embodiments of this application can use M threads to compute the planned trajectory of the vehicle in M ​​scenarios in parallel, such as... Figure 6 As shown, the planned trajectory C(t) in the corresponding scenario is obtained. The method for calculating the planned trajectory can be any decision planning algorithm, such as the lattice algorithm. This application does not impose any specific restrictions.

[0150] Furthermore, embodiments of this application can calculate the scenario occurrence probability and trajectory quality score for each scenario based on the quality scores of different planned trajectories, serving as a comprehensive evaluation index for decision-making and planning in the corresponding scenario. In other words, embodiments of this application can use the product of the scenario occurrence probability and the trajectory quality score in the current scenario as a comprehensive evaluation index for decision-making and planning in that scenario. The formula for calculating the comprehensive evaluation index can be:

[0151] U i =p i *F i ,

[0152] Where, p i F represents the probability of behavior for each obstacle. i Let F be the quality score of the i-th trajectory. i It can be represented as:

[0153]

[0154] Where, ω s For safety weights, ω e For efficiency weights, t0 is the trajectory start time, t f Let s be the trajectory termination time. end The s-coordinate of the trajectory endpoint, s start Let s be the coordinate of the starting point of the trajectory.

[0155] For example, such as Figure 2 As shown, this embodiment of the application can employ parallel computing, using two threads to calculate the planned trajectory of the vehicle in two scenarios in parallel, thereby obtaining the planned trajectories C1(t) and C2(t) for the corresponding scenarios. The method for calculating the planned trajectory can employ the lattice algorithm, and this application does not impose specific limitations. Furthermore, this embodiment of the application can evaluate the quality of the planned trajectories in different scenarios based on the expression for quality scores, thereby obtaining the quality score F1 for the first trajectory and the quality score F2 for the second trajectory. Furthermore, this embodiment of the application calculates the probability p of each scenario's occurrence using the calculation formula of the comprehensive evaluation index. i The trajectory quality score F in the current scenario i The product U i This serves as a comprehensive evaluation indicator for decision-making and planning in this scenario.

[0156] In step S108, the largest comprehensive evaluation index among multiple scenarios is obtained based on the comprehensive evaluation index, and the planning trajectory corresponding to the largest comprehensive evaluation index is taken as the final optimal trajectory.

[0157] As one possible approach, embodiments of this application can find the largest comprehensive evaluation index across multiple scenarios, which can be represented as U. max The corresponding planned trajectory is taken as the final optimal trajectory, which can be represented as C. * (t).

[0158] For example, such as Figure 2 As shown in the embodiment of this application, the largest comprehensive evaluation index U can be found in two scenarios. max And its corresponding planned trajectory is taken as the final optimal trajectory C. * (t).

[0159] The working principle of the vehicle autonomous driving prediction and decision-making method proposed in this application will be explained in detail below with a specific embodiment.

[0160] like Figure 2 As shown in the embodiment of this application, the vehicle is traveling in the left lane of a two-lane road, and there is an obstacle vehicle in front of it.

[0161] Step S701: Obtain information about vehicle obstacles. In this embodiment, information about vehicle obstacles can be obtained using relevant devices such as millimeter-wave radar (this application does not impose specific limitations). This information may include the location of the obstacles (x...). l y l ), velocity v l , heading angle θ l and the obstacle's identification ID l .

[0162] Step S702: Calculate the centerline of each lane edge. In this embodiment of the application, the lane information of the current lane of the vehicle and the lane information of the right lane can be obtained through a high-precision map (or other devices, which are not specifically limited in this application). Each lane can be represented by two discrete point sequences, and then the centerline information of each lane edge can be obtained by using the calculation formula of the centerline of each lane edge.

[0163] Step S703: Coordinate Projection. In this embodiment, the lane centerline can be used as a reference line to find the point on the lane centerline closest to the obstacle, denoted as p. r The obstacle's coordinates are then projected onto the Frenet coordinate system according to the coordinate transformation expression, thus obtaining the corresponding s-coordinates and l-coordinates.

[0164] Step S704: Obtain traffic light coordinate information. This embodiment of the application can combine technologies such as cameras and image processing to obtain the location information of the traffic lights and the remaining green light duration t. g And by combining the coordinate transformation expression, the s-coordinate of the traffic light can be obtained. g and the l-coordinate l g .

[0165] Step S705: Obstacle Acceleration Prediction. In this embodiment, the acceleration of obstacles can be predicted based on a prediction formula for the acceleration of each obstacle around the vehicle, ultimately obtaining the predicted acceleration at the current moment.

[0166] Step S706: Obstacle lane change intention prediction. The application embodiment can predict the lane change intention of obstacles in front of the vehicle, by... Figure 2 As can be seen, since there is no lane on the left side of the obstacle and a lane on the right side, this embodiment of the application can be based on u ke u re The formula is used to calculate the expected lane-changing profit, and then to obtain the left lane-changing probability p. lc =0, lane keeping probability Right lane change probability

[0167] Step S707: Generating the predicted trajectory of the obstacle. In this embodiment, if the obstacle intends to change lanes to the right, the Frenet coordinates of the right lane change endpoint can be obtained using the Frenet coordinate calculation formula of the right lane change endpoint; if the obstacle does not change lanes, the Frenet coordinates of the predicted endpoint can be obtained using the Frenet coordinate calculation formula of the non-lane change endpoint. Furthermore, in this embodiment, a fifth-order polynomial can be used to connect the starting point and the lane change endpoint or the predicted endpoint to generate the predicted trajectory of each obstacle around the vehicle. In addition, in this embodiment, if the obstacle has multiple intentions, multiple predicted trajectories can be generated.

[0168] Step S708: Obtain the vehicle of interest. For example... Figure 2 As shown, this embodiment of the application can filter vehicles of interest based on the position of the vehicle and the positions of surrounding obstacles. Figure 2 In the scenario shown, the vehicle of interest in this embodiment is only one vehicle in front.

[0169] Step S709: Calculate the probability of occurrence of different interaction scenarios. In this embodiment, the probability of occurrence of different interaction scenarios can be calculated based on the intent of the vehicle of interest and the predicted trajectory. The calculation method is the product of the intent probabilities of the vehicle of interest.

[0170] Step S710: Calculate the probability of generating all interactive scenarios. This embodiment of the application can calculate the probability of generating all interactive scenarios, and select the two scenarios with the highest probability, recording their corresponding probabilities p1 and p2.

[0171] Step S711: Parallel computation of planned trajectories for multiple scenarios. This embodiment of the application can employ parallel computation, using two threads to compute the planned trajectories of the vehicle in two scenarios in parallel, thereby obtaining the corresponding planned trajectories C1(t) and C2(t). The method for calculating the planned trajectories can employ the lattice algorithm; this application does not impose specific limitations.

[0172] Step S712: Evaluate the quality score of the planned trajectory under different scenarios. In this embodiment, the quality of the planned trajectory under different scenarios can be evaluated based on the expression for the quality score, thereby obtaining the quality score F1 of the first trajectory and the quality score F2 of the second trajectory.

[0173] Step S713: Obtain the comprehensive evaluation index. In this embodiment, the probability p of each scenario occurring is calculated using the formula for the comprehensive evaluation index. i The trajectory quality score F in the current scenario i The product U i This serves as a comprehensive evaluation indicator for decision-making and planning in this scenario.

[0174] Step S714: Obtain the final optimal trajectory. This embodiment of the application can find the largest comprehensive evaluation index U in two scenarios. max And its corresponding planned trajectory is taken as the final optimal trajectory C. * (t).

[0175] According to the autonomous driving prediction and decision-making method for vehicles proposed in this application, the method can traverse the obstacles around each vehicle to find the nearest point on the centerline of the corresponding lane, obtain the coordinate information of the corresponding traffic light, and then predict the acceleration of each obstacle around the vehicle. Combined with lane-changing intentions, a predicted trajectory is generated. The probability of different interaction scenarios is calculated based on the intentions of the vehicles of interest, thus obtaining the probability of all interaction scenarios. By parallel computing the planned trajectories of the vehicle under multiple scenarios, the probability of scenario occurrence and trajectory quality score for each scenario are obtained, achieving the acquisition of the highest comprehensive evaluation index among multiple scenarios, and using its corresponding planned trajectory as the final optimal trajectory. This application embodiment can confirm the non-deterministic intentions of surrounding obstacles during autonomous driving decision-making, and through parallel spatiotemporal methods, deduce possible interaction scenarios, making reasonable decisions in each corresponding scenario, thereby achieving a comprehensive evaluation of the vehicle's driving behavior. Therefore, it solves the problems in related technologies, such as the difficulty in non-deterministic prediction of obstacle intentions, the lack of consideration for special scenarios such as traffic lights, the difficulty in adapting to complex traffic environments, and the inability to provide safe, reliable, and stable decision-making suggestions.

[0176] Next, referring to the accompanying drawings, an autonomous driving prediction and decision-making device for a vehicle according to an embodiment of this application is described.

[0177] Figure 8 This is a block diagram of an autonomous driving prediction and decision-making device for a vehicle according to an embodiment of this application.

[0178] like Figure 8 As shown, the autonomous driving prediction and decision-making device 10 of the vehicle includes: a first acquisition module 100, a second acquisition module 200, a traversal module 300, a third acquisition module 400, a prediction module 500, a first calculation module 600, a second calculation module 700, and a generation module 800.

[0179] The first acquisition module 100 is used to acquire the status information of obstacles around the vehicle at the current moment.

[0180] The second acquisition module 200 is used to acquire lane information of the current lane of the vehicle and surrounding lane information in order to obtain the center line of each lane edge.

[0181] Traversal module 300 is used to traverse the obstacles around each vehicle with the center line as the reference line, and find the point closest to the center line of the corresponding lane based on the status information of the obstacles around each vehicle.

[0182] The third acquisition module 400 is used to acquire the location information of the traffic light and the remaining green light duration, and to obtain the coordinate information of the traffic light based on the nearest point on the center line.

[0183] The prediction module 500 is used to predict the acceleration of obstacles around each vehicle based on the remaining green light duration and coordinate information, and to identify the lane-changing intention of obstacles around each vehicle, so as to generate a predicted trajectory of obstacles around each vehicle based on the acceleration and lane-changing intention.

[0184] The first calculation module 600 is used to acquire at least one vehicle of interest based on the actual location of the vehicle, calculate the probability of occurrence of different interaction scenarios based on the intent and predicted trajectory of at least one vehicle of interest, and calculate the probability of generation of all interaction scenarios.

[0185] The second calculation module 700 is used to calculate the planned trajectories of vehicles in multiple scenarios in parallel, to evaluate the quality scores of the planned trajectories in different scenarios, and to calculate the scenario occurrence probability and trajectory quality score in each scenario, which serve as a comprehensive evaluation index for decision planning in the corresponding scenario.

[0186] The generation module 800 is used to obtain the largest comprehensive evaluation index among multiple scenarios based on the comprehensive evaluation index, so as to take the planning trajectory corresponding to the largest comprehensive evaluation index as the final optimal trajectory.

[0187] Optionally, in one embodiment of this application, the formula for calculating the centerline of each lane edge can be:

[0188]

[0189] in, The x-coordinate, y-coordinate, angle with the x-axis, curvature, and cumulative distance of the i-th point on the lane centerline are respectively given. The x-coordinate, y-coordinate, angle with the x-axis, curvature, and cumulative distance of the i-th point on the left side of the lane are respectively given. The x-coordinate, y-coordinate, angle with the x-axis, curvature, and cumulative distance of the i-th point on the left side of the lane are respectively given.

[0190] Optionally, in one embodiment of this application, the formula for predicting the acceleration of each obstacle around the vehicle can be:

[0191]

[0192] Among them, a maxv is the estimated maximum acceleration of the obstacle. i Let v be the velocity of the i-th obstacle, v0 be the velocity of the obstacle in front of the obstacle in the lane where the obstacle is located, and d be the velocity of the obstacle. * (v i Δv) is the desired distance between the obstacle and the vehicle in front, Δv is the speed difference between the obstacle and the vehicle in front, and d is the distance between the obstacle and the obstacle in front in the lane. g I represents the current distance between the obstacle and the traffic light. A (d g v max , t g ) is an indicator function, v max For lane speed limits, t g θ1 represents the remaining green light duration, θ2 represents the following vehicle weight, and θ2 represents the traffic light weight.

[0193] Optionally, in one embodiment of this application, the prediction module 500 includes: an acquisition unit and a prediction unit.

[0194] The acquisition unit is used to traverse the obstacles around the vehicle in each lane and obtain the coordinates of the obstacles around the vehicle.

[0195] The prediction unit is used to predict the lane-changing intention of obstacles based on the coordinate positions of obstacles around the vehicle and a preset priority. Specifically, the lane-changing benefit value of the corresponding lane is obtained according to the number of obstacles around each vehicle and the corresponding courtesy factor to determine the lane-changing intention.

[0196] Optionally, in one embodiment of this application, the formula for calculating the profit from lane changing can be:

[0197]

[0198]

[0199] u ke =a th ,

[0200]

[0201]

[0202] in, The original calculated value for profit from left lane changing. a is the expected acceleration value of the obstacle after changing lanes to the left. i The expected acceleration value of the obstacle in the current lane. a is the expected acceleration value of the vehicle following in the left lane after a left lane change. lfv This represents the expected acceleration value of the vehicle following in the left lane before the left lane change. a is the expected acceleration value of the vehicle following in the current lane after a left lane change. fv Let p be the expected acceleration of the vehicle behind in the current lane before the left lane change, and p be the yield factor. The original calculated value for profit from right lane changing. a is the expected acceleration value of the obstacle after changing lanes to the right. i The expected acceleration value of the obstacle in the current lane. This represents the expected acceleration value of the vehicle following in the right lane after a right lane change. Let a be the expected acceleration value of the vehicle following in the current lane after a right lane change, ReLU be the activation function, and a be the expected acceleration value of the vehicle following in the current lane. th To determine the profit threshold for lane changing, u le The final value of the profit from a left lane change, u re The final value of the profit from right lane change, u ke Maintain profitability for the lane.

[0203] Optionally, in one embodiment of this application, the formula for calculating the comprehensive evaluation index can be:

[0204] U i =p i *F i ,

[0205] Where, p i F represents the probability of behavior for each obstacle. i Let be the quality score of the i-th trajectory.

[0206] It should be noted that the foregoing explanation of the vehicle autonomous driving prediction and decision-making method embodiment also applies to the vehicle autonomous driving prediction and decision-making device of this embodiment, and will not be repeated here.

[0207] The autonomous driving prediction and decision-making device for vehicles proposed in this application can traverse the obstacles around each vehicle to find the closest point on the centerline of the corresponding lane, obtain the coordinate information of the corresponding traffic light, and then predict the acceleration of each obstacle around the vehicle. Combined with lane-changing intentions, a predicted trajectory is generated. The probability of different interaction scenarios is calculated based on the intentions of the vehicle of interest, thus obtaining the probability of all interaction scenarios. By parallel computing the planned trajectories of the vehicle under multiple scenarios, the probability of scenario occurrence and trajectory quality score for each scenario are obtained, achieving the acquisition of the highest comprehensive evaluation index among multiple scenarios, and using its corresponding planned trajectory as the final optimal trajectory. This application embodiment can confirm the non-deterministic intentions of surrounding obstacles during autonomous driving decision-making, and through parallel spatiotemporal methods, deduce possible interaction scenarios, making reasonable decisions in each corresponding scenario, thereby achieving a comprehensive evaluation of the vehicle's driving behavior. Therefore, it solves the problems in related technologies, such as the difficulty in non-deterministic prediction of obstacle intentions, the lack of consideration for special scenarios such as traffic lights, the difficulty in adapting to complex traffic environments, and the inability to provide safe, reliable, and stable decision-making suggestions.

[0208] Figure 9 A schematic diagram of the structure of an electronic device provided in an embodiment of this application. The electronic device may include:

[0209] The memory 901, the processor 902, and the computer program stored on the memory 901 and capable of running on the processor 902.

[0210] When the processor 902 executes the program, it implements the vehicle autonomous driving prediction and decision-making method provided in the above embodiments.

[0211] Furthermore, electronic devices also include:

[0212] Communication interface 903 is used for communication between memory 901 and processor 902.

[0213] The memory 901 is used to store computer programs that can run on the processor 902.

[0214] The memory 901 may include high-speed RAM memory, and may also include non-volatile memory, such as at least one disk storage device.

[0215] If the memory 901, processor 902, and communication interface 903 are implemented independently, then the communication interface 903, memory 901, and processor 902 can be interconnected via a bus to complete communication between them. The bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, or an Extended Industry Standard Architecture (EISA) bus, etc. The bus can be divided into address bus, data bus, control bus, etc. For ease of representation, Figure 9 The bus is represented by a single thick line, but this does not mean that there is only one bus or one type of bus.

[0216] Optionally, in a specific implementation, if the memory 901, processor 902, and communication interface 903 are integrated on a single chip, then the memory 901, processor 902, and communication interface 903 can communicate with each other through an internal interface.

[0217] The processor 902 may be a central processing unit (CPU), an application specific integrated circuit (ASIC), or one or more integrated circuits configured to implement the embodiments of this application.

[0218] This application also provides a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the above-described vehicle autonomous driving prediction and decision-making method.

[0219] This application also provides a computer program product, including a computer program that, when executed by a processor, implements the above-described vehicle autonomous driving prediction and decision-making method.

[0220] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of this application. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0221] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this application, "N" means at least two, such as two, three, etc., unless otherwise explicitly specified.

[0222] Any process or method described in the flowchart or otherwise herein can be understood as representing a module, segment, or portion of code comprising one or N executable instructions for implementing custom logic functions or processes, and the scope of the preferred embodiments of this application includes additional implementations in which functions may be performed not in the order shown or discussed, including substantially simultaneously or in reverse order depending on the functions involved, as should be understood by those skilled in the art to which embodiments of this application pertain.

[0223] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-included system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device. More specific examples (a non-exhaustive list) of computer-readable media include: an electrical connection having one or more wires (electronic device), a portable computer disk drive (magnetic device), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Alternatively, the computer-readable medium may be paper or other suitable media on which the program can be printed, since the program can be obtained electronically by optically scanning the paper or other medium, followed by editing, interpreting, or otherwise processing as necessary, and then stored in a computer memory.

[0224] It should be understood that the various parts of this application can be implemented using hardware, software, firmware, or a combination thereof. In the above embodiments, the N steps or methods can be implemented using software or firmware stored in memory and executed by a suitable instruction execution system. If implemented in hardware, as in another embodiment, it can be implemented using any one or more of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0225] Those skilled in the art will understand that all or part of the steps of the methods in the above embodiments can be implemented by a program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, the program includes one or a combination of the steps of the method embodiments.

[0226] Furthermore, the functional units in the various embodiments of this application can be integrated into a processing module, or each unit can exist physically separately, or two or more units can be integrated into a module. The integrated module can be implemented in hardware or as a software functional module. If the integrated module is implemented as a software functional module and sold or used as an independent product, it can also be stored in a computer-readable storage medium.

[0227] The storage medium mentioned above can be a read-only memory, a disk, or an optical disk, etc. Although embodiments of this application have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting this application. Those skilled in the art can make changes, modifications, substitutions, and variations to the above embodiments within the scope of this application.

Claims

1. A predictive decision-making method for autonomous driving of a vehicle, characterized in that, include: Obtain the status information of obstacles around the vehicle at the current moment; Obtain lane information of the current lane of the vehicle and information of the surrounding lanes to determine the centerline of each lane edge. Using the center line as a reference line, traverse the obstacles around each vehicle, and find the point closest to the center line of the corresponding lane based on the state information of the obstacles around each vehicle. Obtain the location information and remaining green light duration of the traffic light, and obtain the coordinate information of the traffic light based on the nearest point on the center line; Based on the remaining green light duration and the coordinate information, the acceleration of the obstacles around each vehicle is predicted, and the lane-changing intention of the obstacles around each vehicle is identified, so as to generate a predicted trajectory of the obstacles around each vehicle based on the acceleration and the lane-changing intention. At least one vehicle of interest is obtained based on the actual location of the vehicle, and the probability of occurrence of different interaction scenarios is calculated based on the intention and predicted trajectory of the at least one vehicle of interest, and the probability of generation of all interaction scenarios is calculated. The planned trajectories of the vehicle in multiple scenarios are calculated in parallel to evaluate the quality scores of the planned trajectories in different scenarios. The probability of occurrence of each scenario and the trajectory quality score are calculated as a comprehensive evaluation index for decision planning in the corresponding scenario. The largest comprehensive evaluation index among the multiple scenarios is obtained based on the comprehensive evaluation index, and the planning trajectory corresponding to the largest comprehensive evaluation index is taken as the final optimal trajectory.

2. The method according to claim 1, characterized in that, The formula for calculating the centerline of each lane edge is as follows: in, The x-coordinate, y-coordinate, angle with the x-axis, curvature, and cumulative distance of the i-th point on the lane centerline are respectively given. The x-coordinate, y-coordinate, angle with the x-axis, curvature, and cumulative distance of the i-th point on the left side of the lane are respectively given. The x-coordinate, y-coordinate, angle with the x-axis, curvature, and cumulative distance of the i-th point on the left side of the lane are respectively given.

3. The method according to claim 1, characterized in that, The formula for predicting the acceleration of obstacles around each vehicle is: Among them, a max v is the estimated maximum acceleration of the obstacle. i Let v be the velocity of the i-th obstacle, v0 be the velocity of the obstacle in front of the obstacle in the lane where the obstacle is located, and d be the velocity of the obstacle. * (v i ,Δv) is the desired distance between the obstacle and the vehicle in front, Δv is the speed difference between the obstacle and the vehicle in front, and d is the distance between the obstacle and the obstacle in front in the lane. g I represents the current distance between the obstacle and the traffic light. A (d g ,v max ,t g ) is an indicator function, v max For lane speed limits, t g θ1 represents the remaining green light duration, θ2 represents the following vehicle weight, and θ2 represents the traffic light weight.

4. The method according to claim 1, characterized in that, The identification of lane-changing intentions involving obstacles around each vehicle includes: Traverse the obstacles around the vehicle in each lane and obtain the coordinates of the obstacles around the vehicle; Based on the coordinates of obstacles around the vehicle, the lane-changing intention of the obstacles is predicted by a preset priority. The lane-changing benefit value of the corresponding lane is obtained according to the number of obstacles around each vehicle and the corresponding courtesy factor to determine the lane-changing intention.

5. The method according to claim 4, characterized in that, The formula for calculating the profit from lane changing is as follows: u ke =a th , in, The original calculated value for profit from left lane changing. a is the expected acceleration value of the obstacle after changing lanes to the left. i The expected acceleration value of the obstacle in the current lane. Let a be the expected acceleration value of the vehicle following in the left lane after a left lane change. lfv This represents the expected acceleration value of the vehicle following in the left lane before the left lane change. a is the expected acceleration value of the vehicle following in the current lane after a left lane change. fv Let p be the expected acceleration of the vehicle behind in the current lane before the left lane change, and p be the yield factor. The original calculated value for profit from right lane changing. a is the expected acceleration value of the obstacle after changing lanes to the right. i The expected acceleration value of the obstacle in the current lane. This represents the expected acceleration value of the vehicle following in the right lane after a right lane change. Let a be the expected acceleration value of the vehicle following in the current lane after a right lane change, ReLU be the activation function, and a be the expected acceleration value of the vehicle following in the current lane. th To determine the profit threshold for lane changing, u le The final value of the profit from a left lane change, u re The final value of the profit from right lane change, u ke Maintain profitability for the lane.

6. The method according to claim 1, characterized in that, The formula for calculating the comprehensive evaluation index is as follows: U i =p i *F i , Where, p i F represents the probability of behavior for each obstacle. i Let be the quality score of the i-th trajectory.

7. An autonomous driving prediction and decision-making device for vehicles, characterized in that, include: The first acquisition module is used to acquire the status information of obstacles around the vehicle at the current moment; The second acquisition module is used to acquire lane information of the current lane of the vehicle and surrounding lane information, so as to obtain the center line of each lane edge line; The traversal module is used to traverse each obstacle around the vehicle using the center line as a reference line, and find the point closest to the center line of the corresponding lane based on the state information of each obstacle around the vehicle. The third acquisition module is used to acquire the location information of the traffic light and the remaining green light duration, and to obtain the coordinate information of the traffic light based on the nearest point on the center line; The prediction module is used to predict the acceleration of obstacles around each vehicle based on the remaining green light duration and the coordinate information, and to identify the lane-changing intention of obstacles around each vehicle, so as to generate a predicted trajectory of obstacles around each vehicle based on the acceleration and the lane-changing intention. The first calculation module is used to obtain at least one vehicle of interest based on the actual location of the vehicle, calculate the probability of occurrence of different interaction scenarios based on the intention and predicted trajectory of the at least one vehicle of interest, and calculate the probability of generation of all interaction scenarios. The second calculation module is used to calculate the planned trajectory of the vehicle in multiple scenarios in parallel, to evaluate the quality score of the planned trajectory in different scenarios, and to calculate the scenario occurrence probability and trajectory quality score in each scenario as a comprehensive evaluation index for decision planning in the corresponding scenario. The generation module is used to obtain the largest comprehensive evaluation index among the multiple scenarios based on the comprehensive evaluation index, so as to take the planning trajectory corresponding to the largest comprehensive evaluation index as the final optimal trajectory.

8. The apparatus according to claim 7, characterized in that, The formula for calculating the centerline of each lane edge is as follows: in, The x-coordinate, y-coordinate, angle with the x-axis, curvature, and cumulative distance of the i-th point on the lane centerline are respectively given. The x-coordinate, y-coordinate, angle with the x-axis, curvature, and cumulative distance of the i-th point on the left side of the lane are respectively given. The x-coordinate, y-coordinate, angle with the x-axis, curvature, and cumulative distance of the i-th point on the left side of the lane are respectively given.

9. The apparatus according to claim 7, characterized in that, The formula for predicting the acceleration of obstacles around each vehicle is: Among them, a max v is the estimated maximum acceleration of the obstacle. i Let v be the velocity of the i-th obstacle, v0 be the velocity of the obstacle in front of the obstacle in the lane where the obstacle is located, and d be the velocity of the obstacle. * (v i ,Δv) is the desired distance between the obstacle and the vehicle in front, Δv is the speed difference between the obstacle and the vehicle in front, and d is the distance between the obstacle and the obstacle in front in the lane. g I represents the current distance between the obstacle and the traffic light. A (d g ,v max ,t g ) is an indicator function, v max For lane speed limits, t g θ1 represents the remaining green light duration, θ2 represents the following vehicle weight, and θ2 represents the traffic light weight.

10. The apparatus according to claim 7, characterized in that, The prediction module includes: The acquisition unit is used to traverse the obstacles around the vehicle in each lane and obtain the coordinates of the obstacles around the vehicle. The prediction unit is used to predict the lane-changing intention of obstacles based on the coordinate positions of obstacles around the vehicle and a preset priority. The lane-changing benefit value of the corresponding lane is obtained according to the number of obstacles around each vehicle and the corresponding courtesy factor to determine the lane-changing intention.

11. The apparatus according to claim 10, characterized in that, The formula for calculating the profit from lane changing is as follows: u ke =a th , in, The original calculated value for profit from left lane changing. a is the expected acceleration value of the obstacle after changing lanes to the left. i The expected acceleration value of the obstacle in the current lane. Let a be the expected acceleration value of the vehicle following in the left lane after a left lane change. lfv This represents the expected acceleration value of the vehicle following in the left lane before the left lane change. a is the expected acceleration value of the vehicle following in the current lane after a left lane change. fv Let p be the expected acceleration of the vehicle behind in the current lane before the left lane change, and p be the yield factor. The original calculated value for profit from right lane changing. a is the expected acceleration value of the obstacle after changing lanes to the right. i The expected acceleration value of the obstacle in the current lane. This represents the expected acceleration value of the vehicle following in the right lane after a right lane change. Let a be the expected acceleration value of the vehicle following in the current lane after a right lane change, ReLU be the activation function, and a be the expected acceleration value of the vehicle following in the current lane. th To determine the profit threshold for lane changing, u le The final value of the profit from a left lane change, u re The final value of the profit from right lane change, u ke Maintain profitability for the lane.

12. The apparatus according to claim 7, characterized in that, The formula for calculating the comprehensive evaluation index is as follows: U i =p i *F i , Where, p i F represents the probability of behavior for each obstacle. i Let be the quality score of the i-th trajectory.

13. An electronic device, characterized in that, include: A memory, a processor, and a computer program stored in the memory and executable on the processor, the processor executing the program to implement the autonomous driving predictive decision method for a vehicle as described in any one of claims 1-6.

14. A computer-readable storage medium having a computer program stored thereon, characterized in that, The program is executed by the processor to implement the autonomous driving predictive decision method for vehicles as described in any one of claims 1-6.

15. A computer program product, comprising a computer program, characterized in that, When the computer program is executed, it is used to implement the autonomous driving prediction and decision-making method for vehicles as described in any one of claims 1-6.

Images