Automatic driving track planning method and device considering perception uncertainty and medium

Abstract

The invention relates to an automatic driving track planning method and device considering perception uncertainty and a medium. The method comprises the following steps: constructing a probabilistic target detection model for environment perception, and constructing an obstacle buffer area; constructing a trajectory planning algorithm based on a soft constraint convex feasible set, and generating an optimal planning trajectory; according to the invention, modeling is carried out on obstacles in a driving environment by using multivariate Gaussian distribution, a probabilistic perception model is established, and high-precision perception and variance inference are carried out on the surrounding environment; according to the method, envelope frame attribute variance information is utilized to construct a sensing obstacle maximum jitter area and establish an obstacle buffer area, so that the sensing uncertainty can be quantified and transmitted, and more reliable sensing data support is provided for a downstream module; a relaxation factor is added into a trajectory planning objective function, so that the method adapts to a perceived uncertain environment in a planning process, and the smoothness of a planned trajectory and the adaptive capacity and driving comfort of a vehicle in the uncertain environment are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent connected vehicles, and specifically to an automatic driving trajectory planning method, device and medium taking into account perception uncertainty. Background Art

[0002] In recent years, autonomous driving technology has made significant progress and has demonstrated excellent performance in many typical driving tasks. However, when dealing with dynamic and complex traffic scenarios, the performance of autonomous driving systems still has a lot of room for improvement. Among them, factors such as bad weather, sensor noise, and the limitations of the algorithm itself often cause the perception module to produce jitters or even false detections in many scenarios, making the system's understanding of the surrounding environment uncertain and unstable. In actual engineering development, the decision-making and planning module often completely trusts the reasoning results of the perception module, and thus tends to adopt overly conservative strategies when dealing with perception uncertainties to reduce potential risks and ensure driving safety.

[0003] However, when the vehicle is driving in an area with dense obstacles, frequent conservative avoidance not only brings a significant computational burden to the decision-making and planning algorithm, increasing computing time and resource consumption, but may also lead to overly cautious driving decisions, significantly reducing the driving comfort of passengers. At the same time, this strategic conservatism may cause a contradiction between safety and efficiency in certain situations, especially in crowded urban streets or complex interactive scenarios, affecting the overall performance and practicality of the system.

[0004] Therefore, how to deal with perception uncertainty in complex environments, improve the flexibility of planning strategies while ensuring safety, and plan the smoothness of trajectories is one of the key issues that need to be urgently addressed in the current development of autonomous driving technology. In response to the above problems, the present invention proposes an autonomous driving trajectory planning method that considers perception uncertainty, effectively performs multi-objective optimization solutions between the safety and comfort of trajectory planning, handles the perception uncertainty problem of the autonomous driving system more intelligently, and improves its adaptability and passenger experience in actual scenarios. Summary of the invention

[0005] The technical problem to be solved by the present invention is to overcome the above-mentioned problems existing in the prior art, and provide a method, device and medium for autonomous driving trajectory planning considering perception uncertainty. The present invention proposes an autonomous driving trajectory planning method considering perception uncertainty, which is used to deal with the multi-objective solution problem between safety and comfort in trajectory planning under the condition of uncertain perception. Based on the multivariate Gaussian distribution, a probabilistic target detection model is constructed to perform high-precision perception and variance inference on obstacles in the surrounding environment of the autonomous driving vehicle. Based on the jitter range of the perception results in space, an obstacle buffer is constructed. Finally, a soft-constrained convex feasible set planning algorithm is used to solve the optimal trajectory, and the safety and smoothness of the planned trajectory are improved by introducing slack variables in the objective function.

[0006] In order to solve the above technical problems, the present invention is implemented by adopting the following technical solutions, which are described as follows in conjunction with the accompanying drawings:

[0007] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device.

[0008] A trajectory planning method for autonomous driving considering perception uncertainty, comprising:

[0009] Construct a probabilistic target detection model for environmental perception and build an obstacle buffer zone; construct a trajectory planning algorithm based on a soft-constrained convex feasible set to generate the optimal planning trajectory.

[0010] Furthermore, the probabilistic target detection model includes four parts: an environment encoder, a category decoder, an envelope decoder and a variance decoder. When the environment information X is input into the probabilistic target detection model, the environment encoder performs feature extraction to obtain high-order features of the environment information, and the category decoder, envelope decoder and variance decoder obtain obstacle categories, envelope attributes and variance information respectively. The data flow is expressed as:

[0011]

[0012] in, High-level semantic information representing environmental information, Indicates the predicted obstacle type, Represents the predicted obstacle envelope information, f enc(·; θ) is the encoding process under parameter θ, D cls (·; θ) is the category decoding process under parameter θ, D reg (·; θ) is the envelope decoding process under parameter θ, D var (·; θ) is the variance decoding process under parameter θ, and a softplus layer is added at the output of this process to ensure that the output result is always positive. is the covariance matrix of obstacles predicted by the model;

[0013] Furthermore, each attribute of the envelope is independent of each other, and the covariance matrix is a 7×7 diagonal matrix.

[0014] Furthermore, the probabilistic target detection model is constructed based on multivariate Gaussian distribution;

[0015] Assuming that the attribute information of the envelope of obstacles in the environment obeys the multivariate Gaussian distribution, the probabilistic target detection model outputs the parameters of the obstacle distribution, and the mean vector of the multivariate Gaussian distribution is used as the attribute information of the envelope. Variance information constitutes the covariance matrix of the distribution It is expressed as:

[0016]

[0017] Among them, θ represents the network parameters, B represents the true value of the obstacle attribute; |·| represents taking the determinant of the matrix, and ·T represents the transpose operation of the matrix;

[0018] The label information of the data set is regarded as a multivariate Gaussian distribution with a covariance matrix of all zero matrices, and as a Dirac distribution, which can be expressed as:

[0019]

[0020] Among them, P G represents the distribution of labels, and δ(·) represents the Dirac function.

[0021] By optimizing the network parameters θ, the difference between the distribution predicted by the model and the true distribution is minimized, so that the model has the ability of environmental perception and variance reasoning, which can be expressed as:

[0022]

[0023] in, represents the Kullback-Leibler (KL) divergence between two distributions;

[0024] Use KL divergence as an indicator to measure the difference between two distributions, use this difference as the learning optimization target of the neural network, and construct the loss function Optimize the regression task and variance inference task of each variance inference neural network; the loss function is expressed as:

[0025]

[0026] in, Indicates the number of obstacles detected by the model; B n represents the true value of the nth envelope box, and They represent the attribute information and covariance matrix of the nth envelope predicted by the model, respectively, and n represents the ordinal index of the obstacle.

[0027] Furthermore, the constructed probabilistic target detection model is trained using the dataset. The loss function consists of three parts: category loss, regression loss, and direction loss. Focal Loss is selected as the category loss:

[0028]

[0029] in, represents the classification loss function, p a is the softmax output of the anchor in the category decoder, α, γ are the empirical parameters in the loss function;

[0030] The loss function of the probabilistic target detection model is expressed as:

[0031]

[0032] in, Represents the total loss function of the model, β cls , β KL and β dir is the balance coefficient of the three loss categories, is a softmax classification loss.

[0033] After training the probabilistic target detection model for multiple rounds, a probabilistic target detection model with environmental perception and variance inference capabilities was constructed. When the environmental information X was input into the model, the probabilistic perception result was obtained:

[0034]

[0035] in, Represents the set of perception results predicted by the model;

[0036] The encoder and decoder of the model gradually have the ability to extract and predict point cloud information features.

[0037] Furthermore, an obstacle buffer zone is constructed, the specific contents are as follows:

[0038] Based on the distribution assumptions and output variance information of the probabilistic target detection model, an obstacle buffer is set;

[0039] Get probabilistic target detection results

[0040]

[0041] pass It can be seen that the standard deviation of each attribute of the nth obstacle envelope is:

[0042]

[0043] Where n represents the ID of the obstacle; σ x , σ y and σ z are the standard deviations of the obstacle center points on the x, y, and z axes respectively; l , σ w and σ h are the standard deviations of the obstacle envelope length, width and height respectively; σ θ is the standard deviation of obstacle heading angles. ;

[0044] The envelope attributes of all perceived obstacles should follow the 3σ principle, and the jitter range of each attribute information should be between its ±3σ value, that is:

[0045] x∈[x-3σ x , x+3σ x ] l∈[l-3σ l , l+3σ l ]

[0046] y∈[y-3σ y ,y+3σ y ] w∈[w-3σ w , w+3σ w ] θ∈[θ-3σ θ ,θ+3σ θ ]

[0047] z∈[z-3σ z , z+3σ z ] h∈[h-3σ h , h+3σ h ]

[0048] Select the envelope, translate it in the coordinate system under the BEV perspective, and select the union as the obstacle expansion envelope. The length and width of the expansion envelope are:

[0049]

[0050] Among them, l and w represent the length and width of the obstacle envelope predicted by the perception module, respectively. e and w e Respectively represent the length and width of the expanded obstacle envelope.

[0051] Furthermore, a trajectory planning algorithm based on soft-constrained convex feasible sets is constructed. The specific contents are as follows:

[0052] Define the planning trajectory point sequence as

[0053]

[0054] in, represents the coordinates of the vehicle's planned trajectory, H represents the planning step length of the preview, and p0 represents the current position of the vehicle;

[0055] Solve the optimal trajectory of the vehicle, expressed as:

[0056]

[0057] stΓ(S)=∩ n Γ n (S) = ∩ n {P:φ n (P)≥-S}

[0058] Among them, J(P) represents the objective function in the trajectory planning process, ω||S|| 2 is the penalty for intrusion into the soft constraint, P * , S * represents the optimal state in the state space, φ n (P) represents the distance to the nth obstacle, Γ n (·) represents the constraint condition for the nth obstacle, Γ(·) represents all The union of;

[0059] The state variables are expanded to

[0060]

[0061] The feasible set is expressed as:

[0062]

[0063] Arranged

[0064]

[0065] Among them, Z represents the intermediate variable in the reasoning process, represents the feasible set of the algorithm under the nth obstacle, represents the feasible set in the entire environment, k represents the kth iteration process, and L and S represent the coefficients in the derivation process;

[0066] In each round of iteration, the algorithm starts from the current solution and finds a new solution in the global convex feasible set so that the state variables move in the optimal direction, that is,

[0067]

[0068] The convex feasible set algorithm is used to solve the state variables multiple times. When the difference between the number of iterations is less than the threshold, the optimal state Z is obtained. * , thus obtaining the optimal output trajectory P * .

[0069] Furthermore, before building a probabilistic target detection model for environmental perception, the following steps need to be performed:

[0070] Step 1: The vehicle uses onboard sensors to collect environmental information during driving;

[0071] Step 2: Label the data collected in step 1 to form a labeled data set;

[0072] Step 3: Divide the data set in step 2 into training set, validation set and test set;

[0073] In step 1, step 2, and step 3, the vehicle uses onboard sensors to collect environmental information during the vehicle's driving process and mark and divide it. The specific contents are as follows:

[0074] Smart cars are equipped with various on-board sensors that can sense the surrounding environment and obtain environmental information, denoted as X. Each piece of environmental information X contains N obstacles, denoted as:

[0075]

[0076] in, Represents the obstacle list, o n represents the nth obstacle, and N represents the number of obstacles;

[0077] The dataset D includes the type information c of each obstacle n and 3-D envelope information B n ={x n ,y n , z n , w n , l n ,h n ,θ n}; where (x n ,y n , zn ) indicates an obstacle o n The location information (w n , l n ,h n ,θ n ) indicates an obstacle o n The properties and orientation of the 3-D envelope;

[0078] Definition In each X, the obstacle label is defined as:

[0079]

[0080] The above dataset is divided into training set, validation set and test set.

[0081] An apparatus comprising one or more processors;

[0082] A memory for storing one or more programs;

[0083] When the one or more programs are executed by the one or more processors, the one or more processors implement the method as described above.

[0084] A computer-readable storage medium stores a computer program, which implements the method described above when executed by a processor.

[0085] Compared with the prior art, the present invention has the following beneficial effects:

[0086] 1. This invention proposes a trajectory planning method that takes into account perception uncertainty. This method uses multivariate Gaussian distribution to model obstacles in the driving environment, establishes a probabilistic perception model, and performs high-precision perception and variance inference on the surrounding environment.

[0087] 2. Based on the 3σ principle, the present invention utilizes the variance information of envelope box attributes to construct the maximum jitter area of ​​perceived obstacles, thereby establishing an obstacle buffer zone, which can quantify and transmit perception uncertainty, provide more reliable perception data support for downstream modules, and help improve the safety and stability of vehicle decision-making in complex environments.

[0088] 3. The present invention utilizes a trajectory planning algorithm of a convex feasible set integrating soft constraints, and adds a relaxation factor to the trajectory planning objective function, so as to adapt to the perceived uncertain environment during the planning process, thereby improving the smoothness of the planned trajectory and the adaptability and driving comfort of the vehicle in an uncertain environment. BRIEF DESCRIPTION OF THE DRAWINGS

[0089] The present invention will be further described below in conjunction with the accompanying drawings:

[0090] Figure 1 The technical roadmap of the present invention

[0091] Figure 2 Schematic diagram of obstacle buffer zone DETAILED DESCRIPTION

[0092] In order to make the purpose, technical scheme and advantages of the implementation of the present invention clearer, the technical scheme in the embodiment of the present invention will be described in more detail below in conjunction with the drawings in the embodiment of the present invention. In the drawings, the same or similar reference numerals throughout represent the same or similar elements or elements with the same or similar functions. The described embodiments are part of the embodiments of the present invention, not all of the embodiments. The embodiments described below with reference to the drawings are exemplary and are intended to be used to explain the present invention, and should not be construed as limitations on the present invention. Based on the embodiments in the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention. The embodiments of the present invention are described in detail below in conjunction with the drawings.

[0093] In the description of the present invention, it should be understood that the terms "center", "longitudinal", "lateral", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside" and the like indicate orientations or positional relationships based on the orientations or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore should not be understood as limiting the scope of protection of the present invention.

[0094] The present invention is described in detail below in conjunction with the accompanying drawings:

[0095] like Figure 2 As shown, the present invention proposes an automatic driving trajectory planning method considering perception uncertainty, which is used to deal with the multi-objective solution problem between safety and comfort in trajectory planning under the condition of uncertain perception. First, based on the multivariate Gaussian distribution, a probabilistic target detection model is constructed to perform high-precision perception and variance inference on obstacles in the environment. Secondly, based on the obstacle perception results, an obstacle perception jitter range is constructed in space, and an obstacle buffer zone is constructed. Finally, a soft-constrained convex feasible set planning algorithm is used to consider the perception uncertainty of obstacles in space, introduce slack variables in the objective function, and solve the optimal trajectory to ensure the safety and smoothness of the planned trajectory. Specifically including:

[0096] Step 1: The vehicle uses onboard sensors to collect environmental information during driving;

[0097] Step 2: Label the data collected in step 1 to form a labeled data set;

[0098] Step 3: Divide the data set in step 2 into training set, validation set and test set;

[0099] In step 1, step 2, and step 3, the smart car is equipped with various on-board sensors, which can sense the surrounding environment and obtain environmental information, which is recorded as X. Each piece of environmental information X contains N obstacles, which are recorded as:

[0100]

[0101] in, Represents the obstacle list, o n represents the nth obstacle, and N represents the number of obstacles.

[0102] The dataset D includes the type information c of each obstacle n , such as pedestrians, bicycles, and cars, and 3-D envelope information B n ={x n ,y n , z n , w n , l n ,h n ,θ n}. Among them, (xn ， y n , z n ) indicates an obstacle o n The location information (w n , l n ,h n ,θ n ) indicates an obstacle o n The properties and orientation of the 3-D envelope.

[0103] Therefore, in each X, the obstacle label is defined as:

[0104]

[0105] The above data set is divided into training set, validation set and test set according to a certain ratio (for example, 8:1:1).

[0106] Step 4: Build a probabilistic target detection model based on multivariate Gaussian distribution:

[0107] Construct a probabilistic target detection model with variance inference capability. The model can be simplified into the following four parts: environment encoder, category decoder, envelope decoder and variance decoder. When the environment information X is input into the model, the environment encoder extracts features to obtain high-order features of the environment information. The decoder further obtains information such as obstacle category, envelope attributes and variance. The data flow can be expressed as:

[0108]

[0109] in, High-level semantic information representing environmental information, Indicates the predicted obstacle type, Represents the predicted obstacle envelope information, f enc (·; θ) is the encoding process under parameter θ, D cls (·; θ) is the category decoding process under parameter θ, D reg (·; θ) is the envelope decoding process under parameter θ, D var (·; θ) is the variance decoding process under parameter θ, and a softplus layer is added at the output of this process to ensure that the output result is always positive. is the covariance matrix of the obstacle predicted by the model. In the present invention, it is assumed that each attribute of the envelope is independent of each other, so the covariance matrix is a 7×7 diagonal matrix.

[0110] The probabilistic target detection model represents the target detection model being constructed in step 4. This model is different from the ordinary target detection model and can output the probability distribution of the detected target. This output is also the data basis for the subsequent steps.

[0111] Due to the lack of label information for uncertainty in the existing dataset, it is impossible to directly train the model parameters through some supervised learning algorithms. Therefore, it is assumed that the attribute information of the envelope of obstacles in the environment follows a multivariate Gaussian distribution, and the model outputs the parameters of the obstacle distribution, and the mean vector of the multivariate Gaussian distribution is used as the attribute information of the envelope. Variance information constitutes the covariance matrix of the distribution It can be expressed as:

[0112]

[0113] Among them, θ represents the network parameters, which is a set of learnable variables, and B represents the true value of the obstacle attribute. |·| represents taking the determinant of the matrix, · T Represents the transpose operation of a matrix.

[0114] Obviously, the label information of the data set can be regarded as a multivariate Gaussian distribution with a covariance matrix of all zero matrices, which can be regarded as a Dirac distribution and can be expressed as:

[0115]

[0116] Among them, P G represents the distribution of labels, and δ(·) represents the Dirac function.

[0117] By optimizing the network parameters θ, the difference between the distribution predicted by the model and the true distribution is minimized, so that the model has the ability of environmental perception and variance reasoning, which can be expressed as:

[0118]

[0119] in, Represents the Kullback-Leibler (KL) divergence between two distributions. The model in “minimizing the model prediction distribution” refers to the probabilistic object detection model being built.

[0120] Use KL divergence as an indicator to measure the difference between two distributions, use this difference as the learning optimization target of the neural network, and construct the loss function Optimize the regression task and variance inference task of each variance inference neural network. The loss function can be expressed as:

[0121]

[0122] in, Indicates the number of obstacles detected by the model. n represents the true value of the nth envelope box, and They represent the nth envelope attribute information and covariance matrix predicted by the model, and n represents the ordinal index of the obstacle. The "model" in "represents the number of obstacles detected by the model" specifically refers to the probabilistic target detection model being built.

[0123] Step 5: Use the dataset to train the target detection model:

[0124] Use the dataset to train the probabilistic target detection model built in step 4. The loss function consists of three parts: category loss, regression loss, and direction loss. Here, Focal Loss is selected as the category loss:

[0125]

[0126] in, represents the classification loss function, p a is the softmax output of the anchor in the category decoder, and α, γ are the empirical parameters in the loss function.

[0127] In summary, the loss function of the model can be expressed as:

[0128]

[0129] in, Represents the total loss function of the model, βcls , β KL and β dir is the balance coefficient of the three loss categories, is a softmax classification loss.

[0130] After training the model for multiple rounds, a probabilistic target detection model was successfully constructed, which has the ability of environmental perception and variance inference. When the environmental information X is input into the model, the probabilistic perception result can be obtained:

[0131]

[0132] in, Represents the set of perception results predicted by the model.

[0133] The encoder and decoder of the model gradually have the ability to extract and predict the features of point cloud information, and can achieve good perception effects.

[0134] Step 6: Use the probabilistic target detection model built in step 5 to perceive the environment and build an obstacle buffer:

[0135] In real traffic scenarios, there are many traffic participants, and they are dynamic and time-varying, with complex topological relationships between them. The perception model of smart cars often has perception uncertainties such as jitter or even jump in timing. To ensure the safety of vehicle driving, an obstacle buffer is set based on the distribution assumptions and output variance information of the probabilistic target detection model in step 5.

[0136] In step 5, the probabilistic target detection results are obtained

[0137]

[0138] pass It can be seen that the standard deviation of each attribute of the nth obstacle envelope is:

[0139]

[0140] Where n represents the ID of the obstacle; σ x , σ y and σ z are the standard deviations of the obstacle center points on the x, y, and z axes respectively; l , σ w and σ h are the standard deviations of the obstacle envelope length, width and height respectively; σ θ is the standard deviation of the obstacle heading angles.

[0141] The envelope attributes of all perceived obstacles should follow the 3σ principle, and the jitter range of each attribute information should be between its ±3σ value, that is:

[0142] x∈[x-3σ x , x+3σ x ] l∈[l-3σ l , l+3σ l ]

[0143] y∈[y-3σ y ,y+3σ y ] w∈[w-3σ w , w+3σ w ] θ∈[θ-3σ θ ,θ+3σ θ ]

[0144] z∈[z-3σ z , z+3σ z ] h∈[h-3σ h , h+3σ h ]

[0145] Therefore, the attached Figure 1 As shown in the figure, the largest envelope is selected (the dotted box in the figure), the coordinate system under the BEV perspective is translated, and the union is selected as the obstacle expansion envelope. For simplicity, a new rectangle B is constructed. e (The solid rectangle in the outer circle in the figure) is used as the new extended envelope. Based on the geometric relationship, we can infer that the length and width of the extended envelope are:

[0146]

[0147] Among them, l and w represent the length and width of the obstacle envelope predicted by the perception module, respectively. e and w e Respectively represent the length and width of the expanded obstacle envelope. e It is input into the planning algorithm of the downstream planning module.

[0148] Step 7: Construct a trajectory planning algorithm based on soft-constrained convex feasible sets:

[0149] Step 8: Generate optimal planning trajectory.

[0150] In traditional planning algorithms, obstacles are usually regarded as a hard constraint, and vehicles must avoid any spatial overlap with them. In the present invention, step six designs a dynamic buffer zone for perceived obstacles, and outputs the perception results to the planning module in a very conservative paradigm. If it is completely regarded as a hard constraint, it will greatly reduce the drivable range of the autonomous driving vehicle and the comfort of the occupants. Therefore, a trajectory planning algorithm based on a soft-constrained convex feasible set is adopted in step seven, and the obstacle buffer zone is regarded as a soft constraint, so that the vehicle can slightly invade the buffer zone when planning the trajectory without affecting safety, so as to more calmly adapt to the jitter changes of perceived obstacles and significantly improve the smoothness of the trajectory.

[0151] In the trajectory planning problem of autonomous driving vehicles, the planned trajectory point sequence is defined as

[0152]

[0153] in, represents the coordinates of the vehicle's planned trajectory, H represents the planning step of preview, and p0 represents the current position of the vehicle.

[0154] In order to solve the optimal trajectory of the vehicle, the problem is modeled as a convex feasible set optimization problem with slack variables added, which can be expressed as:

[0155]

[0156] stΓ(S)=∩ n Γ n (S) = ∩ n {P:φ n (P)≥-S}

[0157] Among them, J(P) represents the objective function in the trajectory planning process, ω||S|| 2 is the penalty for intrusion into the soft constraint, P * , S * represents the optimal state in the state space, φ n (P) represents the distance to the nth obstacle, Γ n (·) represents the constraint condition for the nth obstacle, Γ(·) represents all The union of .

[0158] Compared with the traditional convex optimization feasible set algorithm, in this problem, the state variables are expanded to

[0159]

[0160] Therefore, the feasible set is expressed as:

[0161]

[0162] Arranged

[0163]

[0164] Among them, Z represents the intermediate variable in the reasoning process, represents the feasible set of the algorithm under the nth obstacle, represents the feasible set in the entire environment, k represents the kth iteration process, and L and S represent the coefficients in the derivation process.

[0165] In each round of iteration, the algorithm starts from the current solution and finds a new solution in the global convex feasible set so that the state variables move in the optimal direction, that is,

[0166]

[0167] The convex feasible set algorithm is used to solve the state variables multiple times. When the difference between the number of iterations is less than the threshold, it is considered that the optimal state Z is obtained. * , thus obtaining the optimal output trajectory P * The solution is mainly to obtain the optimal z, that is, the optimal [P, S], through multiple iterations of the above formula. The solution process is based on the convex feasible set method, which is a more common solution method. Due to the addition of the penalty term, the buffer zone is used as a region that can be slightly invaded in extreme cases, which improves the smoothness of the planned trajectory and the comfort of the occupants.

[0168] Based on the above-mentioned automatic driving trajectory planning method considering perception uncertainty, the present invention provides another device. The device includes but is not limited to: one or more processors and memories.

[0169] The memory, as a computer-readable storage medium, can be used to store software programs, computer executable programs and modules, such as program instructions corresponding to the automatic driving trajectory planning method considering perception uncertainty in the embodiment of the present invention. The processor executes various functional applications and data processing of the vehicle by running the software programs, instructions and modules stored in the memory, that is, realizing the automatic driving trajectory planning method considering perception uncertainty.

[0170] The memory may mainly include a program storage area and a data storage area, wherein the program storage area may store an operating system and at least one application required for a function; the data storage area may store data created according to the use of the terminal, etc. In addition, the memory may include a high-speed random access memory, and may also include a non-volatile memory, such as at least one disk storage device, a flash memory device, or other non-volatile solid-state storage devices.

[0171] The present invention also provides a computer-readable storage medium having a computer program stored thereon. When the program is executed by a processor, an automatic driving trajectory planning method considering perception uncertainty is implemented. The automatic driving trajectory planning method considering perception uncertainty comprises the following steps:

[0172] Step 1: The vehicle uses onboard sensors to collect environmental information during driving;

[0173] Step 2: Label the data collected in step 1 to form a labeled data set;

[0174] Step 3: Divide the data set in step 2 into training set, validation set and test set;

[0175] Step 4: Build a probabilistic target detection model based on multivariate Gaussian distribution;

[0176] Step 5: Use the dataset to train the target detection model;

[0177] Step 6: Use the constructed probabilistic target detection model to perceive the environment and build an obstacle buffer zone;

[0178] Step 7: Construct a trajectory planning algorithm based on soft-constrained convex feasible sets;

[0179] Step 8: Generate optimal planning trajectory.

[0180] The present invention provides a computer-readable storage medium, whose computer executable instructions are not limited to the method operations described above, and can also execute related operations in the automatic driving trajectory planning method considering perception uncertainty provided by any embodiment of the present invention.

[0181] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed in this application can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0182] In the above embodiments, it can be implemented in whole or in part by software, hardware, firmware or any combination thereof. When implemented using software, it can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When loading and executing computer program instructions on a computer, the process or function described in the embodiment of the present invention is generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. The computer instructions can be transmitted from a website site, a computer, a server or a data center to another website site, a computer, a server or a data center by wired (for example, coaxial cable, optical fiber, digital subscriber line DSL) or wireless (for example, infrared, wireless, microwave, etc.) mode. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server, a data center, etc. that contains one or more available media integrated. The available medium can be a magnetic medium (for example, a floppy disk, a hard disk, a tape), an optical medium (for example, a DVD), or a semiconductor medium (for example, a solid-state hard disk Solid State Disk, SSD), etc.

[0183] The above is only a specific implementation of the present invention, but the protection scope of the present invention is not limited thereto. Any modification, equivalent replacement and improvement made by any person skilled in the art within the technical scope disclosed by the present invention and within the spirit and principle of the present invention shall be covered within the protection scope of the present invention. At the same time, the contents not described in detail in this specification belong to the prior art known to those skilled in the art.

Claims

1. An autonomous driving trajectory planning method considering perception uncertainty, characterized in that: include: Build a probabilistic target detection model for environmental perception and construct an obstacle buffer zone; Construct a trajectory planning algorithm based on soft-constrained convex feasible sets to generate the optimal planning trajectory.

2. The method for autonomous driving trajectory planning considering perception uncertainty according to claim 1, characterized in that: The probabilistic target detection model includes four parts: environment encoder, category decoder, envelope decoder and variance decoder; When the environmental information X is input into the probabilistic target detection model, the environmental encoder performs feature extraction to obtain high-order features of the environmental information. The category decoder, envelope decoder and variance decoder obtain obstacle categories, envelope attributes and variance information respectively. The data flow is represented as: in, High-level semantic information representing environmental information, Indicates the predicted obstacle type, Represents the predicted obstacle envelope information, f enc (·; θ) is the encoding process under parameter θ, D cls (·; θ) is the category decoding process under parameter θ, D reg (·; θ) is the envelope decoding process under parameter θ, D var (·; θ) is the variance decoding process under parameter θ, and a softplus layer is added at the output of this process to ensure that the output result is always positive. is the covariance matrix of obstacles predicted by the model.

3. The method for autonomous driving trajectory planning considering perception uncertainty according to claim 2, characterized in that: Each attribute of the envelope is independent of each other, and the covariance matrix is a 7×7 diagonal matrix.

4. The method for autonomous driving trajectory planning considering perception uncertainty according to claim 2, characterized in that: The probabilistic target detection model is constructed based on multivariate Gaussian distribution; Assuming that the attribute information of the envelope of obstacles in the environment obeys the multivariate Gaussian distribution, the probabilistic target detection model outputs the parameters of the obstacle distribution, and the mean vector of the multivariate Gaussian distribution is used as the attribute information of the envelope. Variance information constitutes the covariance matrix of the distribution It is expressed as: Among them, θ represents the network parameters, B represents the true value of the obstacle attribute; |·| represents taking the determinant of the matrix, · T Represents the transpose operation of a matrix; The label information of the data set is regarded as a multivariate Gaussian distribution with a covariance matrix of all zero matrices, and as a Dirac distribution, which can be expressed as: Among them, P G represents the distribution of labels, and δ(·) represents the Dirac function. By optimizing the network parameters θ, the difference between the distribution predicted by the model and the true distribution is minimized, so that the model has the ability of environmental perception and variance reasoning, which can be expressed as: in, represents the Kullback-Leibler (KL) divergence between two distributions; Use KL divergence as an indicator to measure the difference between two distributions, use this difference as the learning optimization target of the neural network, and construct the loss function Optimize the regression task and variance inference task of each variance inference neural network; the loss function is expressed as: in, Indicates the number of obstacles detected by the model; B n represents the true value of the nth envelope box, and They represent the attribute information and covariance matrix of the nth envelope predicted by the model, respectively, and n represents the ordinal index of the obstacle.

5. The method for autonomous driving trajectory planning considering perception uncertainty according to claim 4, characterized in that: The constructed probabilistic target detection model is trained using the dataset. The loss function consists of three parts: category loss, regression loss, and direction loss. Focal Loss is selected as the category loss: in, represents the classification loss function, p a is the softmax output of the anchor in the category decoder, α,γ are the empirical parameters in the loss function; The loss function of the probabilistic target detection model is expressed as: in, Represents the total loss function of the model, β cls , β KL and β dir is the balance coefficient of the three loss categories, is a softmax classification loss. After training the probabilistic target detection model for multiple rounds, a probabilistic target detection model with environmental perception and variance inference capabilities was constructed. When the environmental information X was input into the model, the probabilistic perception result was obtained: in, Represents the set of perception results predicted by the model; The encoder and decoder of the model gradually have the ability to extract and predict point cloud information features.

6. The method for autonomous driving trajectory planning considering perception uncertainty according to claim 5, characterized in that: Construct an obstacle buffer zone, the specific contents are as follows: Based on the distribution assumptions of the probabilistic target detection model and the output variance information, an obstacle buffer is set; Get probabilistic target detection results pass It can be seen that the standard deviation of each attribute of the nth obstacle envelope is: Where n represents the ID of the obstacle; σ x , σ y and σ z are the standard deviations of the obstacle center points on the x, y, and z axes respectively; l , σ w and σ h are the standard deviations of the obstacle envelope length, width and height respectively; σ θ is the standard deviation of obstacle heading angles. ; The envelope attributes of all perceived obstacles should follow the 3σ principle, and the jitter range of each attribute information should be between its ±3σ value, that is: x∈[x-3σ x ,x+3σ x ]l∈[l-3σ l ,l+3σ l ] y∈[y-3σ y ,y+3σ y ]w∈[w-3σ w ,w+3σ w ]θ∈[θ-3σ θ ,θ+3σ θ ] z∈[z-3σ z ,z+3σ z ]h∈[h-3σ h ,h+3σ h ] Select the envelope, translate it in the coordinate system under the BEV perspective, and select the union as the obstacle expansion envelope. The length and width of the expansion envelope are: Among them, l and w represent the length and width of the obstacle envelope predicted by the perception module, respectively. e and w e Respectively represent the length and width of the expanded obstacle envelope.

7. The method for autonomous driving trajectory planning considering perception uncertainty according to claim 6, characterized in that: Construct a trajectory planning algorithm based on a soft-constrained convex feasible set. The specific contents are as follows: Define the planning trajectory point sequence as in, represents the coordinates of the vehicle's planned trajectory, H represents the planning step length of the preview, and p0 represents the current position of the vehicle; Solve the optimal trajectory of the vehicle, expressed as: stΓ(S)=∩ n C n (S)=∩ n {P:φ n (P)≥-S} Among them, J(P) represents the objective function in the trajectory planning process, ω||S|| 2 is the penalty for intrusion into the soft constraint, P * , S * represents the optimal state in the state space, φ n (P) represents the distance to the nth obstacle, Γ n (·) represents the constraint condition for the nth obstacle, Γ(·) represents all The union of; The state variables are expanded to The feasible set is expressed as: Arranged Among them, Z represents the intermediate variable in the reasoning process, represents the feasible set of the algorithm under the nth obstacle, represents the feasible set in the entire environment, k represents the kth iteration process, and L and S represent the coefficients in the derivation process; In each round of iteration, the algorithm starts from the current solution and finds a new solution in the global convex feasible set so that the state variables move in the optimal direction, that is, The convex feasible set algorithm is used to solve the state variables multiple times. When the difference between the number of iterations is less than the threshold, the optimal state Z is obtained. * , thus obtaining the optimal output trajectory P * .

8. The method for autonomous driving trajectory planning considering perception uncertainty according to claim 1, characterized in that: Before building a probabilistic target detection model for environmental perception, the following steps are required: Step 1: The vehicle uses onboard sensors to collect environmental information during driving; Step 2: Label the data collected in step 1 to form a labeled data set; Step 3: Divide the data set in step 2 into training set, validation set and test set; In step 1, step 2, and step 3, the vehicle uses onboard sensors to collect environmental information during the vehicle's driving process and mark and divide it. The specific contents are as follows: Smart cars are equipped with various on-board sensors that can sense the surrounding environment and obtain environmental information, denoted as X. Each piece of environmental information X contains N obstacles, denoted as: in, Represents the obstacle list, o n represents the nth obstacle, and N represents the number of obstacles; The dataset D includes the type information c of each obstacle n and 3-D envelope information B n ={x n ,y n ,z n ,w n ,l n ,h n ,θ n }; where (x n ,y n ,z n ) indicates an obstacle o n The location information (w n ,l n ,h n ,θ n ) indicates an obstacle o n The properties and orientation of the 3-D envelope; Definition In each X, the obstacle label is defined as: The above dataset is divided into training set, validation set and test set.

9. A device, characterized in that: comprising one or more processors; A memory for storing one or more programs; When the one or more programs are executed by the one or more processors, the one or more processors implement the method according to any one of claims 1 to 8.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the method according to any one of claims 1 to 8 is implemented.

Images