A driving obstacle avoidance method, device and system based on trajectory prediction

Abstract

The present invention is applicable to the field of autonomous driving technology, and provides a driving obstacle avoidance method, device and system based on trajectory prediction. The method includes: obtaining a road traffic model to obtain a first type of predicted trajectory of each other vehicle, and obtaining a first type of restricted area; based on the abnormal motion vector of the problem vehicle, obtaining a second type of predicted trajectory of the problem vehicle to obtain a second type of restricted area; setting the other vehicles interfered by the second type of predicted trajectory as associated vehicles, and obtaining the motion vectors of the associated vehicles; based on the motion vectors of the associated vehicles, obtaining a third type of predicted trajectory of the associated vehicles; obtaining a third type of restricted area; after removing the restricted areas from the area within the road boundary coordinates, re-planning the driving trajectory of the host vehicle. This application is not limited to predicting the obstacle itself, but also takes into account the possible avoidance actions of other vehicles affected by its association, improving the safety and reliability of route planning.

Description

Technical Field

[0001] The present invention belongs to the technical field of autonomous driving, and particularly relates to a driving obstacle avoidance method, device and system based on trajectory prediction. Background Art

[0002] The automatic obstacle avoidance technology is a core part of the vehicle autonomous driving system, which enables the vehicle to autonomously identify and avoid obstacles in a complex road environment. The realization of this technology depends on the comprehensive operation of a variety of sensors, algorithms and a large number of decision-making systems, with a high degree of complexity.

[0003] The functional reliability of automatic obstacle avoidance not only depends on the perception system, but also requires a powerful path planning and decision-making system. Through real-time calculation, the system will select the best obstacle avoidance path to ensure driving safety. There are various existing obstacle avoidance algorithms, for example, the rapidly-exploring random tree algorithm, the probabilistic roadmap method, the model predictive control method, etc. The principle of such algorithms is that when the system detects that an obstacle suddenly appears in front of the vehicle, the system will continuously monitor and obtain the real-time dynamic position of the obstacle, predict the path of the obstacle, and then re-plan the driving trajectory of the vehicle itself according to the predicted path of the obstacle, so that the vehicle selects strategies such as braking, accelerating, detouring or changing lanes to achieve obstacle avoidance.

[0004] However, existing solutions often only focus on avoiding the obstacles that affect traffic themselves, but ignore that other vehicles may also be affected by the obstacles and make suddenly steering, lane-changing and other abnormal driving-related dangerous behaviors, and these related abnormal driving behaviors may also endanger their own driving safety. Therefore, it is necessary to further improve the existing path prediction and planning methods. Summary of the Invention

[0005] The purpose of the embodiments of the present application is to provide a driving obstacle avoidance method based on trajectory prediction, aiming to solve the problem that existing obstacle avoidance solutions often only focus on avoiding the obstacles that affect traffic themselves, but ignore that other vehicles may also be affected by the obstacles and make suddenly steering, lane-changing and other abnormal driving-related dangerous behaviors, and these related abnormal driving behaviors may also endanger their own driving safety.

[0006] The embodiments of the present application are implemented as follows. A driving obstacle avoidance method based on trajectory prediction is provided, and the method includes:

[0007] Obtain a road traffic model constructed with the vehicle itself as the coordinate origin, where the road traffic model includes the road boundary coordinates of the road and the coordinates of several other vehicles in the road;

[0008] Obtain the respective motion vectors of each other vehicle in the road traffic model, and based on the motion vectors, obtain the first type of predicted trajectories of each other vehicle; based on the first type of predicted trajectories, obtain the first type of no-go areas;

[0009] Monitor whether there are vehicles with abnormal offsets among several other vehicles, and set the vehicles with abnormal offsets as problem vehicles; the abnormal offset is an abnormal driving behavior in which the deviation between the actual trajectory and the historical trajectory of the vehicle is higher than the abnormal threshold;

[0010] Based on the abnormal motion vector of the problem vehicle, obtain the second type of predicted trajectory of the problem vehicle, and based on the second type of predicted trajectory, obtain the second type of no-go area;

[0011] Set the other vehicles interfered by the second type of predicted trajectory as associated vehicles, and obtain the motion vectors of the associated vehicles; based on the motion vectors of the associated vehicles, obtain the third type of predicted trajectories of the associated vehicles; based on the third type of predicted trajectories, obtain the third type of no-go areas;

[0012] After removing the first type of no-go area, the second type of no-go area, and the third type of no-go area from the area within the road boundary coordinates, re-plan the driving trajectory of the host vehicle.

[0013] Preferably, the method for obtaining the road traffic model with the host vehicle as the coordinate origin, where the road traffic model includes the road boundary coordinates of the road and the other vehicle coordinates of several other vehicles in the road is as follows:

[0014] Define the host vehicle coordinates as the coordinate origin, use the road extension direction as the y-axis, and use the driving direction of the host vehicle as the positive direction of the y-axis to establish a plane rectangular coordinate system;

[0015] Obtain the set M of other vehicle coordinates of other vehicles:

[0016] ;

[0017] where m is the vehicle number of the other vehicle, represents the position coordinates of the m-th vehicle;

[0018] Divide the road boundary into several differential segments, and define the road boundary B based on the differential segments:

[0019] ;

[0020] where, represents the b-th segment of the boundary, and the total number of boundaries is n segments.

[0021] Preferably, the method for obtaining the motion vectors of each other vehicle in the road traffic model and obtaining the first type of predicted trajectory of each other vehicle based on the motion vectors is as follows:

[0022] Obtain the historical coordinates of the vehicle to be predicted relative to the host vehicle and the set I of time stamps t corresponding to the coordinates:

[0023]

[0024] wherein, represents the position of the vehicle to be predicted at time;

[0025] Input the set I into a path prediction network, which is built based on a long short-term memory network and includes an input gate, a forget gate, and an output gate. The path prediction network is used to predict and output the position coordinates of the vehicle to be predicted at a future time moment .

[0026] Preferably, the loss function of the path prediction network is set as:

[0027]

[0028] where i represents a total of i historical data, k represents the kth, and represent the x-axis and y-axis coordinates of the kth fitted coordinate, and represent the x-axis and y-axis coordinates of the k historical coordinates.

[0029] Preferably, the method for monitoring whether there are vehicles with abnormal offsets among several other vehicles and setting the vehicles with abnormal offsets as problem vehicles is as follows:

[0030] Based on the real-time driving trajectory of the vehicle, obtain the absolute value of the displacement change per unit time of the vehicle at the current moment and the absolute value of the direction angle change per unit time ;

[0031] Based on the historical driving trajectory of the vehicle, obtain the absolute value of the displacement change per unit time of the vehicle in the historical unit time and the absolute value of the direction angle change per unit time ;

[0032] Set the vehicle that satisfies the following conditions as having an abnormal offset:

[0033] and / or

[0034] wherein, is the displacement anomaly threshold coefficient, is the direction angle anomaly threshold coefficient.

[0035] Preferably, the method for obtaining the associated vehicle is as follows:

[0036] Obtain the forward direction of other vehicles to be judged except the problem vehicle, and set it as the forward interference area;

[0037] Judge whether there is mutual interference overlap or intersection between the second type of predicted trajectory and the forward interference area of the vehicle to be judged, and obtain the first judgment result;

[0038] Judge whether there is mutual interference overlap or intersection between the second type of predicted trajectory and the first type of predicted trajectory of the vehicle to be judged, and obtain the second judgment result;

[0039] When one or both of the first judgment result and the second judgment result are satisfied, set the vehicle to be judged that meets the judgment condition as the associated vehicle.

[0040] Preferably, the trajectory prediction method for the associated vehicle is as follows:

[0041] Construct an emergency avoidance path dataset for the vehicle in an emergency avoidance state. The emergency avoidance path dataset includes several typical emergency avoidance paths that occur when the vehicle is disturbed, and each typical emergency avoidance path is marked with its corresponding typical path feature;

[0042] The typical path feature is used to characterize the data feature of the moving path of the typical emergency avoidance path at the initial stage of the path;

[0043] Based on the motion vector of the associated vehicle, obtain the real path of the associated vehicle in the emergency avoidance state, and obtain the real data feature of the real path;

[0044] Compare the real data feature with all the typical path features in the emergency avoidance path dataset to obtain several similarity comparison results;

[0045] Set the typical emergency avoidance paths corresponding to all the typical path features with similarity comparison results higher than the similarity threshold as the third type of predicted trajectory.

[0046] Another object of the embodiments of the present application is to provide a driving obstacle avoidance device based on trajectory prediction, and the device includes:

[0047] A road traffic model acquisition module, configured to acquire a road traffic model constructed with the host vehicle as the coordinate origin, where the road traffic model includes the road boundary coordinates of the road and the coordinates of several other vehicles in the road;

[0048] A first type of restricted area acquisition module, configured to acquire the respective motion vectors of each other vehicle in the road traffic model, and based on the motion vectors, obtain the first type of predicted trajectories of each other vehicle; based on the first type of predicted trajectories, obtain the first type of restricted areas;

[0049] A problem vehicle acquisition module, configured to monitor whether there is a vehicle with abnormal deviation among several other vehicles, and set the vehicle with abnormal deviation as a problem vehicle; the abnormal deviation is an abnormal driving behavior in which the deviation between the actual trajectory and the historical trajectory of the vehicle is higher than the abnormal threshold;

[0050] A second type of restricted area acquisition module, configured to acquire the second type of predicted trajectory of the problem vehicle based on the abnormal motion vector of the problem vehicle, and obtain the second type of restricted area based on the second type of predicted trajectory;

[0051] A third type of restricted area acquisition module, configured to set the other vehicles interfered by the second type of predicted trajectory as associated vehicles, and acquire the motion vectors of the associated vehicles; based on the motion vectors of the associated vehicles, obtain the third type of predicted trajectories of the associated vehicles; based on the third type of predicted trajectories, obtain the third type of restricted areas;

[0052] A path planning module, configured to re-plan the driving trajectory of the host vehicle after removing the first type of restricted area, the second type of restricted area, and the third type of restricted area from the area within the road boundary coordinates.

[0053] Another object of the embodiments of the present application is to provide a driving obstacle avoidance system based on trajectory prediction, including a memory and a processor, where a computer program is stored in the memory, and when the computer program is executed by the processor, the processor executes the steps of a driving obstacle avoidance method based on trajectory prediction as described above.

[0054] A driving obstacle avoidance method based on trajectory prediction provided by an embodiment of the present application has prominent advantages. That is, the present application predicts the driving trajectory in real time according to the motion states of the host vehicle and other vehicles, and is not limited to predicting the obstacle itself. It also takes into account the possible avoidance actions of other vehicles affected by its association, enabling the obstacle avoidance plan to adjust to the above potential dangers according to the real-time traffic state, rather than being based only on a static map or a pre-set trajectory, thereby improving the safety and reliability of route planning. By monitoring the abnormal deviation of the vehicle, potential dangerous driving behaviors or emergencies can be detected in a timely manner, improving road safety. By obtaining and analyzing different types of dangerous restricted areas, the method has higher accuracy and forward-looking predictability when dealing with complex traffic environments, improving the comprehensive safety level of the planned path. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] Figure 1 FIG. is an application environment diagram of a driving obstacle avoidance method based on trajectory prediction provided by an embodiment of the present application;

[0056] Figure 2 FIG. is a flowchart of a driving obstacle avoidance method based on trajectory prediction provided by an embodiment of the present application;

[0057] Figure 3 FIG. is a road state simulation diagram provided by an embodiment of the present application;

[0058] Figure 4 FIG. is another road state simulation diagram provided by an embodiment of the present application;

[0059] Figure 5 FIG. is a structural block diagram of a driving obstacle avoidance device based on trajectory prediction provided by an embodiment of the present application;

[0060] Figure 6 FIG. is an internal structural block diagram of a computer device in an embodiment. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0061] In order to make the objectives, technical solutions and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0062] It can be understood that the terms "first", "second", etc. used in the present application may be used herein to describe various elements, but unless otherwise specified, these elements are not limited by these terms. These terms are only used to distinguish one element from another. For example, without departing from the scope of the present application, the first script may be referred to as the second script, and similarly, the second script may be referred to as the first script.

[0063] Figure 1This is an application environment diagram of the driving obstacle avoidance method based on trajectory prediction provided by the embodiments of the present application. As Figure 1 shown, in this application environment, it includes a sensor 110 and a computer device 120.

[0064] The computer device 120 can be an independent physical server or terminal, or a server cluster composed of multiple physical servers. It can be a cloud server providing basic cloud computing services such as cloud servers, or a local in-vehicle computer, etc., but is not limited thereto.

[0065] The sensor 110 can be a lidar, a millimeter-wave radar, a visible camera device, etc., and can be used to detect the surrounding conditions of the vehicle. The sensor 110 and the computer device 120 can be connected through a wired or wireless network, and the present application does not limit this here.

[0066] As Figure 2 shown, in one embodiment, a driving obstacle avoidance method based on trajectory prediction is proposed. In this embodiment, this method is mainly illustrated by applying it to the Figure 1 computer device 120 in the above. A driving obstacle avoidance method based on trajectory prediction may specifically include the following steps:

[0067] Step S10: Obtain a road traffic model constructed with the vehicle itself as the coordinate origin. The road traffic model includes the road boundary coordinates of the road and the coordinates of several other vehicles in the road;

[0068] In this embodiment, the system first constructs or obtains a road traffic model with the vehicle itself as the coordinate origin, including the road boundary and the positions of other vehicles. Constructing the road traffic model with the vehicle itself as the coordinate origin enables the model to only consider the relative motion relationship between vehicles, simplifying the data calculation amount of the motion prediction algorithm.

[0069] Step S20: Obtain the respective motion vectors of each other vehicle in the road traffic model. Based on the motion vectors, obtain the first type of predicted trajectories of each other vehicle; based on the first type of predicted trajectories, obtain the first type of restricted areas;

[0070] In this embodiment, there are various methods for predicting the trajectories of vehicles based on the motion vectors. For example, it is possible to first obtain the relative speed, acceleration, azimuth angle of the vehicle to be predicted compared to the vehicle itself, and the historical driving data or relative position data of the vehicle, and then use polynomial regression to fit the trajectory, and then estimate the coefficients of the relative speed, acceleration, and azimuth angle in the future by least squares fitting of the historical trajectory data, thereby obtaining the predicted future position. Based on the footprint of the vehicle and the first type of predicted trajectories of the vehicle, the first type of restricted areas can be obtained.

[0071] In one embodiment, asFigure 3 As shown, the coordinate O is the position coordinate of the host vehicle, and vehicles A, B, and C are other vehicles. Among them, vehicle C accelerates relative to the host vehicle, then Figure 3 the dotted area on the top of vehicle C in [Figure] is the first type of restricted area.

[0072] Step S30: Monitor whether there are vehicles with abnormal offsets among several other vehicles, and set the vehicle with an abnormal offset as a problem vehicle; the abnormal offset is an abnormal driving behavior where the deviation between the actual trajectory and the historical trajectory of the vehicle is higher than the abnormal threshold.

[0073] In this embodiment, an abnormal vehicle may refer to a vehicle that exhibits sudden acceleration, deceleration, lane change, etc. For example, Figure 3 for vehicle A in [Figure], the system sensor detects that it does not give a lane change turn signal and predicts that it is driving straight forward. At this time, the vehicle suddenly changes lanes to the left lane, then it can be set as an abnormal vehicle.

[0074] Step S40: Based on the abnormal motion vector of the problem vehicle, obtain the second type of predicted trajectory of the problem vehicle, and obtain the second type of restricted area based on the second type of predicted trajectory.

[0075] In this embodiment, for the abnormal behavior of the vehicle marked as a problem vehicle, its future trajectory can be predicted through its abnormal motion vector, and the second type of restricted area is formed. As Figure 3 shown in [Figure], the dotted area on the left side of vehicle A is the second type of restricted area. The method for predicting the second type of predicted trajectory of the abnormal problem vehicle can be performed based on the current trajectory prediction method, which will not be elaborated here.

[0076] Step S50: Set the other vehicles affected by the second type of predicted trajectory as associated vehicles, and obtain the motion vectors of the associated vehicles; based on the motion vectors of the associated vehicles, obtain the third type of predicted trajectory of the associated vehicles; obtain the third type of restricted area based on the third type of predicted trajectory.

[0077] In this embodiment, if the second type of restricted area affects the driving trajectories of other vehicles, the other vehicles will be marked as associated vehicles. Based on the motion information of these associated vehicles, their trajectories are further predicted and the third type of restricted area is formed. As Figure 3 shown in [Figure], when vehicle A suddenly changes lanes during the normal straight driving of vehicle B, vehicle B may be affected abnormally and exhibit various relative motions relative to vehicle O. The third type of restricted area is Figure 3 the dotted area on the right side or the rear side of vehicle B in [Figure].

[0078] Step S60: After removing the first type of restricted area, the second type of restricted area, and the third type of restricted area from the area within the road boundary coordinates, re-plan the driving trajectory of the host vehicle.

[0079] In the embodiment of the present application, the vehicle will re-plan the route. After determining all the above-mentioned no-go areas, by removing these no-go areas, the driving trajectory of the host vehicle is re-planned to avoid entering these dangerous areas.

[0080] In the embodiment of the present application, the present application predicts the driving trajectory in real time according to the motion states of the host vehicle and other vehicles, and is not limited to predicting the obstacles themselves. It also takes into account the potential avoidance actions of other vehicles affected by their associations, so that the obstacle avoidance plan can adjust to the above-mentioned potential dangers according to the real-time traffic state, rather than being based only on a static map or a pre-set trajectory, thereby improving the safety and reliability of the route planning. By monitoring the abnormal deviation of the vehicle, potential dangerous driving behaviors or emergencies can be detected in time, improving road safety. By obtaining and analyzing different types of dangerous no-go areas, this method has higher accuracy and predictive foresight when dealing with complex traffic environments, improving the comprehensive safety level of the planned route.

[0081] In a preferred embodiment, the method for obtaining a road traffic model constructed with the host vehicle as the coordinate origin, where the road traffic model includes the road boundary coordinates of the road and the coordinates of several other vehicles in the road is as follows:

[0082] Define the host vehicle coordinates as the coordinate origin, use the road extension direction as the y-axis, and use the driving direction of the host vehicle as the positive direction of the y-axis to establish a plane rectangular coordinate system;

[0083] Obtain the set M of the coordinates of other vehicles:

[0084] ;

[0085] where m is the vehicle number of other vehicles, represents the position coordinates of the m-th vehicle;

[0086] Divide the road boundary into several differential segments, and define the road boundary B based on the differential segments:

[0087] ;

[0088] where, represents the b-th boundary, and the total number of boundaries is n segments.

[0089] In the embodiment of the present application, the road model can be obtained and established based on the above method, and the respective boundary coefficients and the coordinate positions of the vehicles can be obtained. In this coordinate system, only the coefficients such as the relative position change, speed change, and acceleration change between vehicles need to be obtained, which can reduce the computational amount of the system.

[0090] In a preferred embodiment, the method for obtaining the motion vectors of each other vehicle in the road traffic model and obtaining the first type of predicted trajectory of each other vehicle based on the motion vectors is as follows:

[0091] Obtain the historical coordinates of the vehicle to be predicted relative to the host vehicle and the set I of time stamps t corresponding to the coordinates:

[0092]

[0093] wherein, represents the position of the vehicle to be predicted at time;

[0094] Input the set I into the path prediction network. The path prediction network is built based on the long short-term memory network and includes three parts: an input gate, a forget gate, and an output gate. The path prediction network is used to predict and output the position coordinates of the vehicle to be predicted at the future time moment based on the data dependency in the set I .

[0095] In the embodiments of the present application, although methods such as polynomial regression combined with the least squares method can be used to achieve path prediction, since the above methods are not good at dealing with complex path problems such as non-linearity, in order to further improve the prediction accuracy of the path in this solution, a deep learning model is used to build the path prediction network. Therefore, preferably, the long short-term memory (LSTM) network model is used in the present application to process path prediction. The basic framework of the model consists of an input gate, a forget gate, and an output gate. It captures the data dependency in the input time series through a recurrent neural network, enabling the system to better handle the long-term dependency problem in time series data and improving the prediction accuracy of the path.

[0096] In a preferred embodiment, the loss function of the path prediction network is set as:

[0097]

[0098] where i represents a total of i historical data, k represents the kth, and represent the x-axis and y-axis coordinates of the kth fitted coordinate, and represent the x-axis and y-axis coordinates of the k historical coordinates.

[0099] In the embodiments of the present application, the loss function of the model is the key to controlling the deviation between the prediction and the true trajectory. Compared with other control methods, since multiple vehicles need to be path-predicted simultaneously in the present application, through comprehensive comparison, the following loss function is adopted to improve the prediction accuracy, and its comprehensive measurement accuracy is relatively high:

[0100]

[0101] In a preferred embodiment, the method for monitoring whether there are vehicles with abnormal offsets among several other vehicles and setting the vehicles with abnormal offsets as problem vehicles is as follows:

[0102] Based on the real-time driving trajectory of the vehicle, obtain the absolute value of the displacement change per unit time of the vehicle at the current moment and the absolute value of the direction angle change per unit time ;

[0103] Based on the historical driving trajectory of the vehicle, obtain the absolute value of the displacement change per unit time in the historical unit time and the absolute value of the direction angle change per unit time ;

[0104] Set the vehicle that meets the following conditions as a vehicle with an abnormal offset:

[0105] and / or

[0106] wherein, is the displacement abnormal threshold coefficient, is the direction angle abnormal threshold coefficient.

[0107] In the embodiments of the present application, considering that problem vehicles are often caused by sudden increases or decreases in the throttle, resulting in instantaneous speed changes, which are manifested as sudden changes in the displacement change per unit time of the vehicle detected by the sensor compared with the historical value, or sudden changes in the driving angle per unit time of the vehicle due to sharp turns of the vehicle direction, where is the displacement abnormal threshold coefficient, the direction angle abnormal threshold coefficient can both refer to the change rate threshold of the mutation, and the specific coefficient can be set based on the actual situation.

[0108] In a preferred embodiment, the method for obtaining the associated vehicle is as follows:

[0109] Obtain the forward direction of other vehicles to be judged except the problem vehicle and set it as the forward interference area;

[0110] Judge whether there is mutual interference overlap or intersection between the second type of predicted trajectory and the forward interference area of the vehicle to be judged, and obtain the first judgment result;

[0111] Determine whether there is mutual interference overlap or intersection between the second type of predicted trajectory and the first type of predicted trajectory of the vehicle to be judged, and obtain a second judgment result;

[0112] When one or both of the first judgment result and the second judgment result are satisfied, set the vehicle to be judged that meets the judgment condition as an associated vehicle.

[0113] In the embodiment of the present application, considering that since vehicles all drive forward, during driving, even if the trajectory of an abnormal vehicle does not intersect with some vehicles, some vehicles may still be frightened because the abnormal vehicle drives past in front of them, resulting in associated abnormalities, causing the driver to be affected and exhibit abnormal driving behaviors, such as emergency lane changes and other situations.

[0114] Therefore, this embodiment further expands the scope of vehicles that may exhibit abnormal driving behaviors, enables the path planning prediction system to simultaneously predict possible abnormal path changes of such vehicles, and does not need to predict the trajectories of normal vehicles that are not affected by them, thereby reducing the computational requirement and being able to further improve the safety factor of path planning.

[0115] In a preferred embodiment, the trajectory prediction method for the associated vehicle is as follows:

[0116] Construct an emergency avoidance path dataset in the emergency avoidance state of the vehicle. The emergency avoidance path dataset includes a number of typical emergency avoidance paths that the vehicle exhibits under interference, and each typical emergency avoidance path is marked with its corresponding typical path feature;

[0117] The typical path feature is used to characterize the data feature of the moving path of the typical emergency avoidance path at the initial stage of the path;

[0118] Based on the motion vector of the associated vehicle, obtain the real path of the associated vehicle in the emergency avoidance state, and obtain the real data feature of the real path;

[0119] Compare the real data feature with all the typical path features in the emergency avoidance path dataset to obtain a number of similarity comparison results;

[0120] Set the typical emergency avoidance paths corresponding to all the typical path features with similarity comparison results higher than the similarity threshold as the third type of predicted trajectory.

[0121] In the embodiment of the present application, as Figure 4As shown, in this schematic diagram, vehicle B is a vehicle driving normally, and its preset direction is the forward direction. At this time, vehicle A is a vehicle that suddenly changes lanes. Then, vehicle B may be affected and have relative movement paths formed by the solid-line direction arrow and the dashed-line direction arrow in the figure. These paths are called typical avoidance paths. Among them, the solid-line direction arrow is denoted as the movement path of the typical avoidance path in the initial stage of the path, and the characteristic performance of this route in terms of data is obtained and denoted as the typical path characteristic. Different paths have different typical path characteristics. By determining the characteristics in the initial stage, the complete predicted path can be approximately obtained. The data characteristics shown in the initial stage of the actual driving data of the vehicle are compared with all the typical path characteristics in the avoidance path dataset, and all the paths with a relatively high degree of similarity comparison results obtained are set as the third type of predicted trajectory, which can further improve the safety of prediction.

[0122] As Figure 5 shown, in one embodiment, a driving obstacle avoidance device based on trajectory prediction is provided. The driving obstacle avoidance device based on trajectory prediction can be integrated into the above-mentioned computer device 120, and specifically may include a road traffic model acquisition module 510, a first type of restricted area acquisition module 520, a problem vehicle acquisition module 530, a second type of restricted area acquisition module 540, a third type of restricted area acquisition module 550, and a path planning module 560.

[0123] Among them, the road traffic model acquisition module 510 is used to acquire a road traffic model constructed with the vehicle itself as the coordinate origin. The road traffic model includes the road boundary coordinates of the road and the other vehicle coordinates of several other vehicles on the road.

[0124] The first type of restricted area acquisition module 520 is used to acquire the motion vectors of each other vehicle in the road traffic model, and based on the motion vectors, obtain the first type of predicted trajectory of each other vehicle; based on the first type of predicted trajectory, obtain the first type of restricted area.

[0125] The problem vehicle acquisition module 530 is used to monitor whether there are vehicles with abnormal offsets among several other vehicles, and set the vehicles with abnormal offsets as problem vehicles; the abnormal offset is an abnormal driving behavior in which the deviation between the actual trajectory and the historical trajectory of the vehicle is higher than the abnormal threshold.

[0126] The second type of restricted area acquisition module 540 is used to obtain the second type of predicted trajectory of the problem vehicle based on the abnormal motion vector of the problem vehicle, and obtain the second type of restricted area based on the second type of predicted trajectory.

[0127] The third type of restricted area acquisition module 550 is configured to set other vehicles interfered by the predicted trajectory of the second type as associated vehicles, and acquire the motion vectors of the associated vehicles; based on the motion vectors of the associated vehicles, obtain the predicted trajectory of the associated vehicles of the third type; based on the predicted trajectory of the third type, obtain the third type of restricted area.

[0128] The path planning module 560 is configured to re-plan the driving trajectory of the vehicle itself after removing the first type of restricted area, the second type of restricted area, and the third type of restricted area from the area within the road boundary coordinates.

[0129] In the embodiments of the present application, the explanations and descriptions of the above driving obstacle avoidance device based on trajectory prediction can refer to the explanations and descriptions of the corresponding methods above. For the description of the above driving obstacle avoidance method based on trajectory prediction, please refer to the above, and details are not described herein again.

[0130] In the embodiments of the present application, the present application predicts the driving trajectory in real time according to the motion states of the vehicle itself and other vehicles, and is not limited to predicting the obstacle itself. It also takes into account the possible avoidance actions of other vehicles affected by its association, so that the obstacle avoidance plan can adjust to the above potential dangers according to the real-time traffic state, rather than being based only on a static map or a pre-set trajectory, thereby improving the safety and reliability of the route planning. By monitoring the abnormal deviation of the vehicle, potential dangerous driving behaviors or emergencies can be detected in time, improving road safety. By acquiring and analyzing different types of dangerous restricted areas, the method has higher accuracy and forward-looking prediction when dealing with complex traffic environments, and improves the comprehensive safety level of the planned path.

[0131] Figure 6 The internal structure diagram of a computer device in an embodiment is shown. The computer device may specifically be Figure 1 the computer device 120 in Figure 6 As shown, the computer device includes a processor, a memory, a network interface, an input device, and a display screen connected through a system bus. Among them, the memory includes a non-volatile storage medium and an internal memory. The non-volatile storage medium of the computer device stores an operating system and may also store a computer program. When the computer program is executed by the processor, the processor can implement the driving obstacle avoidance method based on trajectory prediction. The internal memory may also store a computer program. When the computer program is executed by the processor, the processor can execute the driving obstacle avoidance method based on trajectory prediction. The display screen of the computer device may be a liquid crystal display screen or an electronic ink display screen. The input device of the computer device may be a touch layer covered on the display screen, or a button, a trackball, or a touchpad provided on the housing of the computer device, or an external keyboard, a touchpad, or a mouse, etc.

[0132] Those skilled in the art can understand that Figure 6 The structure shown in Figure 6 is only a block diagram of some structures related to the solution of this application, and does not constitute a limitation on the computer device to which the solution of this application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine some components, or have different component arrangements.

[0133] In one embodiment, the driving obstacle avoidance device based on trajectory prediction provided by this application can be implemented in the form of a computer program, and the computer program can run on a computer device as shown in Figure 6 The memory of the computer device may store each program module that makes up the driving obstacle avoidance device based on trajectory prediction. For example, Figure 5 The road traffic model acquisition module 510, the first type of restricted area acquisition module 520, the problem vehicle acquisition module 530, the second type of restricted area acquisition module 540, the third type of restricted area acquisition module 550, and the path planning module 560 shown in Figure 5 . The computer program composed of each program module enables the processor to execute the steps in the driving obstacle avoidance method based on trajectory prediction in each embodiment of this application described in this specification.

[0134] For example, Figure 6 The computer device shown in Figure 6 can execute step S10 through the road traffic model acquisition module 510 in the driving obstacle avoidance device based on trajectory prediction as shown in Figure 5 Figure 5 . The computer device can execute step S20 through the first type of restricted area acquisition module 520. And so on.

[0135] In one embodiment, a driving obstacle avoidance system based on trajectory prediction is proposed, including a memory and a processor. A computer program is stored in the memory. When the computer program is executed by the processor, the processor is enabled to execute the steps of a driving obstacle avoidance method as described above.

[0136] In the embodiments of this application, for the description of the above driving obstacle avoidance method based on trajectory prediction, please refer to the above, and details are not described herein again.

[0137] In the embodiments of the present application, the present application predicts the driving trajectory in real time according to the motion states of the host vehicle and other vehicles, and is not limited to predicting the obstacles themselves. It also takes into account the evasive actions that other vehicles affected by their association may take, enabling the obstacle avoidance plan to adjust to the above-mentioned potential dangers according to the real-time traffic state, rather than being based only on a static map or a pre-set trajectory, thereby improving the safety and reliability of the route planning. By monitoring the abnormal deviation of the vehicle, potential dangerous driving behaviors or emergencies can be detected in a timely manner, improving road safety. By obtaining and analyzing different types of dangerous no-go zones, the method has higher accuracy and predictive foresight when dealing with complex traffic environments, improving the comprehensive safety level of the planned route.

[0138] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the processor is caused to execute the steps of the driving obstacle avoidance method based on trajectory prediction as described above.

[0139] In the embodiments of the present application, for the description of the above-mentioned driving obstacle avoidance method based on trajectory prediction, please refer to the above text and will not be elaborated here.

[0140] In the embodiments of the present application, the program running based on the method stored in the storage medium of the embodiments of the present application enables the present application to predict the driving trajectory in real time according to the motion states of the host vehicle and other vehicles, and is not limited to predicting the obstacles themselves. It also takes into account the evasive actions that other vehicles affected by their association may take, enabling the obstacle avoidance plan to adjust to the above-mentioned potential dangers according to the real-time traffic state, rather than being based only on a static map or a pre-set trajectory, thereby improving the safety and reliability of the route planning. By monitoring the abnormal deviation of the vehicle, potential dangerous driving behaviors or emergencies can be detected in a timely manner, improving road safety. By obtaining and analyzing different types of dangerous no-go zones, the method has higher accuracy and predictive foresight when dealing with complex traffic environments, improving the comprehensive safety level of the planned route.

[0141] It should be understood that although the steps in the flowcharts of the embodiments of the present application are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless there is a clear indication in this article, the execution of these steps has no strict order limit, and these steps can be executed in other orders. Moreover, at least some of the steps in each embodiment may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily executed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be executed alternately or in turn with at least a part of other steps or sub-steps or stages of other steps.

[0142] Those of ordinary skill in the art can understand that all or part of the processes in the methods of the above embodiments can be completed by instructing relevant hardware through a computer program. The program can be stored in a non-volatile computer-readable storage medium. When the program is executed, it can include the processes of the embodiments of the above methods. Among them, any reference to a memory, storage, database, or other medium used in the embodiments provided in the present application can include non-volatile and / or volatile memories. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or an external cache memory. By way of illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDR SDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), Rambus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and Rambus dynamic RAM (RDRAM), etc.

[0143] The technical features of the above embodiments can be combined arbitrarily. For the sake of brevity of description, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, it should be considered as the scope recorded in this specification.

[0144] The above embodiments merely represent several implementation manners of the present application. The description is relatively specific and detailed, but it should not be construed as a limitation on the patent scope of the present application. It should be noted that for those of ordinary skill in the art, without departing from the concept of the present application, several modifications and improvements can still be made, and these all belong to the protection scope of the present application. Therefore, the protection scope of the patent of the present application should be subject to the appended claims.

Claims

1. A driving obstacle avoidance method based on trajectory prediction, characterized in that, The method includes: Obtain a road traffic model constructed with the host vehicle as the coordinate origin, where the road traffic model includes the road boundary coordinates of the road and the coordinates of several other vehicles in the road; Obtain the respective motion vectors of each of the other vehicles in the road traffic model. Based on the motion vectors, obtain the first type of predicted trajectories of each of the other vehicles; based on the first type of predicted trajectories, obtain the first type of restricted areas; Monitor whether there are vehicles with abnormal offsets among several other vehicles, and set the vehicles with abnormal offsets as problem vehicles; the abnormal offset is an abnormal driving behavior where the deviation between the actual trajectory and the historical trajectory of the vehicle is higher than the abnormal threshold; Based on the abnormal motion vector of the problem vehicle, obtain the second type of predicted trajectory of the problem vehicle, and obtain the second type of restricted area based on the second type of predicted trajectory; Set the other vehicles interfered by the second type of predicted trajectory as associated vehicles, and obtain the motion vectors of the associated vehicles; based on the motion vectors of the associated vehicles, obtain the third type of predicted trajectories of the associated vehicles; based on the third type of predicted trajectories, obtain the third type of restricted areas; After removing the first type of restricted area, the second type of restricted area, and the third type of restricted area from the area within the road boundary coordinates, re-plan the driving trajectory of the host vehicle.

2. The vehicle driving obstacle avoidance method based on trajectory prediction according to claim 1, wherein The method for obtaining a road traffic model constructed with the host vehicle as the coordinate origin, where the road traffic model includes the road boundary coordinates of the road and the coordinates of several other vehicles in the road is as follows: Define the host vehicle coordinates as the coordinate origin, use the road extension direction as the y-axis, and use the driving direction of the host vehicle as the positive direction of the y-axis to establish a plane rectangular coordinate system; Obtain the set M of the coordinates of other vehicles; ； where m is the vehicle number of other vehicles, representing the position coordinates of the m-th vehicle; Divide the road boundary into several differential segments, and define the road boundary B based on the differential segments; ； Among them, represents the boundary of the b-th segment, and the total number of boundaries is n segments.

3. A driving obstacle avoidance method based on trajectory prediction according to claim 1, characterized in that, The method for obtaining the respective motion vectors of each of the other vehicles in the road traffic model and obtaining the first type of predicted trajectories of each of the other vehicles based on the motion vectors is as follows: Obtain the historical coordinates of the vehicle to be predicted relative to the host vehicle and the set I of the time stamps t corresponding to the coordinates: Among them, represents the position of the vehicle to be predicted at time; Input the collection I into the path prediction network, which is built based on the long short-term memory network and includes three parts: an input gate, a forget gate, and an output gate. The path prediction network is used to predict and output the position coordinates of the vehicle to be predicted at a future time moment .

4. A driving obstacle avoidance method based on trajectory prediction according to claim 3, characterized in that, The loss function of the path prediction network is set as: Among them, i represents that there are i historical data in total, and k represents the kth one. and represent the x-axis and y-axis coordinates of the kth fitted coordinate. and represent the x-axis and y-axis coordinates of the k historical coordinates.

5. A driving obstacle avoidance method based on trajectory prediction according to claim 1, characterized in that The method for monitoring whether there are vehicles with abnormal offsets among several other vehicles and setting the vehicles with abnormal offsets as problem vehicles is as follows: Based on the real-time driving trajectory of the vehicle, obtain the absolute value of the displacement change per unit time of the vehicle at the current moment and the absolute value of the change in the direction angle per unit time ; Based on the historical driving trajectory of the vehicle, obtain the absolute value of the displacement change of the vehicle in a historical unit time and the absolute value of the direction angle change in a unit time ; Set the vehicle as having an abnormal offset when it meets the following conditions: and / or Among them, is the displacement anomaly threshold coefficient, is the direction angle anomaly threshold coefficient.

6. The driving obstacle avoidance method based on trajectory prediction according to claim 1, wherein The method for obtaining the associated vehicles is as follows: Obtain the forward direction of other vehicles to be judged except the problem vehicle, and set it as the forward interference area; Judge whether there is mutual interference overlap or intersection between the second type of predicted trajectory and the forward interference area of the vehicle to be judged, and obtain the first judgment result; Judge whether there is mutual interference overlap or intersection between the second type of predicted trajectory and the first type of predicted trajectory of the vehicle to be judged, and obtain the second judgment result; When one or both of the first judgment result and the second judgment result are satisfied, set the vehicle to be judged that meets the judgment conditions as an associated vehicle.

7. A driving obstacle avoidance method based on trajectory prediction according to claim 1, characterized in that, The trajectory prediction method of the associated vehicle is as follows: Construct a risk avoidance path data set in the vehicle's emergency risk avoidance state. The risk avoidance path data set contains several typical risk avoidance paths that occur when the vehicle is disturbed, and each of the typical risk avoidance paths is labeled with its corresponding typical path features; The typical path feature is used to characterize the data feature of the movement path of the typical emergency avoidance path in the initial stage of the path; Based on the motion vector of the associated vehicle, obtain the real path of the associated vehicle in the emergency avoidance state, and obtain the real data feature of the real path; Compare the real data feature with all the typical path features in the emergency avoidance path dataset to obtain several similarity comparison results; Set the typical emergency avoidance paths corresponding to all the typical path features with similarity comparison results higher than the similarity threshold as the third type of predicted trajectory.

8. A driving obstacle avoidance device based on trajectory prediction, characterized in that The device includes: A road traffic model acquisition module, configured to acquire a road traffic model constructed with the host vehicle as the coordinate origin, where the road traffic model includes the road boundary coordinates of the road and the coordinates of several other vehicles in the road; A first type of restricted area acquisition module, configured to acquire the motion vector of each other vehicle in the road traffic model, and based on the motion vector, obtain the first type of predicted trajectory of each other vehicle; based on the first type of predicted trajectory, obtain the first type of restricted area; A problem vehicle acquisition module, configured to monitor whether there is an abnormally deviated vehicle among several other vehicles, and set the vehicle with abnormal deviation as the problem vehicle; the abnormal deviation is an abnormal driving behavior in which the deviation between the actual trajectory and the historical trajectory of the vehicle is higher than the abnormal threshold; A second type of restricted area acquisition module, configured to obtain the second type of predicted trajectory of the problem vehicle based on the abnormal motion vector of the problem vehicle, and obtain the second type of restricted area based on the second type of predicted trajectory; A third type of restricted area acquisition module, configured to set the other vehicles interfered by the second type of predicted trajectory as associated vehicles, and obtain the motion vector of the associated vehicles; based on the motion vector of the associated vehicles, obtain the third type of predicted trajectory of the associated vehicles; based on the third type of predicted trajectory, obtain the third type of restricted area; A path planning module, configured to re-plan the driving trajectory of the host vehicle after removing the first type of restricted area, the second type of restricted area, and the third type of restricted area from the area within the road boundary coordinates.

9. A driving obstacle avoidance system based on trajectory prediction, characterized in that, It includes a memory and a processor. When the computer program stored in the memory is executed by the processor, the processor executes the steps of the method for driving obstacle avoidance based on trajectory prediction according to any one of claims 1 to 7.

Images