Vehicle control method, trajectory planning model training method and device, and vehicle

Abstract

The present disclosure relates to a vehicle control method, a trajectory planning model training method, a device, a vehicle, an electronic device, a non-transitory computer-readable storage medium and a computer program product, and relates to the technical field of intelligent driving. The present disclosure collects the ego vehicle state data of the target vehicle at the current time and the historical data of the surrounding environment of the target vehicle. The historical data of the surrounding environment of the target vehicle at least includes the state data of the traffic participants within the first range from the target vehicle. Based on the historical data of the surrounding environment of the target vehicle, an environment prediction result of a plurality of time steps within a future target time length is generated through a prediction model. The ego vehicle state data, the historical data of the surrounding environment of the target vehicle and the environment prediction result are input into a trajectory planning model to obtain a target planning trajectory of the target vehicle within the target time length. Based on the target planning trajectory, the target vehicle is controlled to intelligently drive, thereby improving the accuracy of trajectory planning of the target vehicle and the driving safety of the vehicle.

Description

Technical Field

[0001] This disclosure relates to the field of intelligent driving technology, and more specifically, to a vehicle control method, a trajectory planning model training method, a device, a vehicle, an electronic device, a non-transitory computer-readable storage medium, and a computer program product. Background Technology

[0002] With the continuous development of intelligent driving technology, intelligent driving is gradually moving from closed parks to open roads, thus requiring continuous upgrades in environmental adaptability and driving safety.

[0003] In intelligent driving technology, trajectory planning is a core capability for ensuring driving safety. It requires accurately calculating the future trajectory of a target vehicle based on current and historical environmental and vehicle status information. However, related technologies suffer from inaccurate trajectory planning, making it crucial to improve the accuracy of predicted trajectories a key technical challenge that urgently needs to be addressed. Summary of the Invention

[0004] To overcome the problems existing in related technologies, this disclosure provides a vehicle control method, a trajectory planning model training method, an apparatus, a vehicle, an electronic device, a non-transitory computer-readable storage medium, and a computer program product.

[0005] According to a first aspect of the present disclosure, a vehicle control method is provided, the method comprising: Collect the target vehicle's current status data and historical data of the target vehicle's surrounding environment; the historical data of the target vehicle's surrounding environment includes at least the status data of traffic participants within a first range from the target vehicle; Based on historical data of the surrounding environment of the target vehicle, environmental prediction results for multiple time steps within the future target duration are generated through a prediction model. The vehicle status data, historical data of the target vehicle's surrounding environment, and the environmental prediction results are input into the trajectory planning model to obtain the target planned trajectory of the target vehicle within the target duration; wherein, the target planned trajectory is obtained by the trajectory planning model through recursive iteration based on the vehicle status data, historical data of the target vehicle's surrounding environment, and environmental prediction results at multiple time steps within the future target duration according to the temporal dimension; Based on the target planned trajectory, the target vehicle is controlled to drive intelligently.

[0006] In some exemplary embodiments of this disclosure, the step of inputting the vehicle state data, historical data of the target vehicle's surrounding environment, and the environmental prediction results into the trajectory planning model to obtain the target planned trajectory of the target vehicle within the target time period includes: Iterate through each time step sequentially within the target duration: For each current time step, the planned trajectory corresponding to the previous time step, the vehicle status data corresponding to the current time step, the historical data of the target vehicle's surrounding environment, and the environmental prediction results are input into the trajectory planning model to generate the planned trajectory for the current time step. The planned trajectory generated at the current time step is used as the input basis for the next time step. The planned trajectory for the next time step is iteratively generated by combining the vehicle status data corresponding to the next time step, the historical data of the target vehicle's surrounding environment, and the environmental prediction results. The process is repeated iteratively step by step until all time steps within the target duration are traversed, thus obtaining the target planned trajectory.

[0007] In some exemplary embodiments of this disclosure, the step of inputting the planned trajectory corresponding to the previous time step, the vehicle state data corresponding to the current time step, the historical data of the target vehicle's surrounding environment, and the environmental prediction results into the trajectory planning model for each current time step to generate the planned trajectory for the current time step includes: The vehicle status data is encoded to obtain first trajectory encoding information; and the planned trajectory and the status data of traffic participants within a first range from the target vehicle are encoded to obtain historical environment encoding information. The environmental prediction results are encoded to obtain future environmental coding information; The first trajectory coding information, historical environment coding information, and future environment coding information are processed for route planning to obtain the planned trajectory at the current time step.

[0008] In some exemplary embodiments of this disclosure, route planning processing is performed on the first trajectory encoding information, historical environment encoding information, and future environment encoding information to obtain the planned trajectory at the current time step, including: The first trajectory encoding information is subjected to time-series encoding processing to obtain trajectory time-series encoding information; Cross-attention processing is performed on the trajectory temporal coding information, the historical environment coding information, and the future environment coding information to obtain the second trajectory coding information; Based on the second trajectory encoding information, the planned trajectory for the current time step is obtained.

[0009] In some exemplary embodiments of this disclosure, before performing route planning processing on the first trajectory encoding information, historical environment encoding information, and future environment encoding information, the method further includes: The first trajectory coding information, the historical environment coding information, and the future environment coding information are time-aligned to obtain aligned first trajectory coding information, historical environment coding information, and future environment coding information, so as to perform route planning processing on the aligned first trajectory coding information, historical environment coding information, and future environment coding information.

[0010] In some exemplary embodiments of this disclosure, the historical data of the environment surrounding the target vehicle further includes a bird's-eye view of the region of interest within a first range from the target vehicle. The step of generating environmental prediction results for multiple time steps within a future target duration based on the historical data of the environment surrounding the target vehicle using a prediction model includes: Based on a bird's-eye view of the region of interest within a first range from the target vehicle and the status data of traffic participants within the first range from the target vehicle, an environmental prediction result for multiple time steps within a future target duration is generated through a prediction model.

[0011] In some exemplary embodiments of this disclosure, the environmental prediction results include one or more of the following: future trajectory data of the traffic participants, traffic light status change data, dynamic obstacle appearance / disappearance data, and road topology change data.

[0012] According to a second aspect of the present disclosure, a trajectory planning model training method is provided, the method comprising: Acquire training samples, which include at least: vehicle status data, future environmental data for multiple time steps within the planning duration, labeled planning trajectory, and historical data of the vehicle's surrounding environment; the historical data of the vehicle's surrounding environment includes at least the status data of traffic participants within a first range from the vehicle. The vehicle status data, historical data of the vehicle's surrounding environment, and future environment data are input into the trajectory planning model to be trained to obtain the predicted planned trajectory of the vehicle within the planning time period; wherein, the predicted planned trajectory is obtained by the trajectory planning model to be trained through recursive iteration based on the vehicle status data, historical data of the vehicle's surrounding environment, and future environment data in a time-series dimension. The trajectory planning model to be trained is trained based on the predicted planned trajectory and the labeled planned trajectory to obtain the target trajectory planning model.

[0013] In some exemplary embodiments of this disclosure, the step of inputting the vehicle state data, historical data of the vehicle's surrounding environment, and future environment data into the trajectory planning model to be trained to obtain the predicted planned trajectory of the vehicle within the planning time period includes: Within the planned duration, each time step is traversed sequentially: For each current time step, the planned trajectory corresponding to the previous time step, the vehicle status data corresponding to the current time step, the historical data of the vehicle's surrounding environment, and the future environment data are input into the trajectory planning model to be trained to generate the predicted planned trajectory for the current time step. The planned trajectory generated at the current time step is used as the input basis for the next time step. Combined with the vehicle status data corresponding to the next time step, the historical data and future environmental data of the vehicle's surrounding environment, the predicted planned trajectory for the next time step is generated iteratively. The process is iterated step by step according to time steps until all time steps within the planned duration are traversed to obtain the predicted planned trajectory.

[0014] In some exemplary embodiments of this disclosure, the historical data of the vehicle's surrounding environment further includes a bird's-eye view of the region of interest within a first range from the vehicle. For each current time step, the process of inputting the planned trajectory corresponding to the previous time step, the vehicle's state data corresponding to the current time step, the historical data of the vehicle's surrounding environment, and the future environmental data into the trajectory planning model to be trained to generate a predicted planned trajectory for the current time step includes: The vehicle status data is encoded to obtain first trajectory encoding information; and the planned trajectory, the status data of traffic participants within a first range from the target vehicle, and the bird's-eye view of the region of interest within a first range from the vehicle are encoded to obtain historical environment encoding information. The environmental prediction results are encoded to obtain future environmental coding information; The first trajectory coding information, historical environment coding information, and future environment coding information are processed for route planning to obtain the predicted planned trajectory at the current time step.

[0015] According to a third aspect of the present disclosure, a vehicle control device is provided, the device comprising: The data acquisition unit is used to collect the vehicle status data of the target vehicle at the current moment, as well as the historical data of the surrounding environment of the target vehicle; the historical data of the surrounding environment of the target vehicle includes at least the status data of traffic participants within a first range from the target vehicle. An environmental prediction unit is used to generate environmental prediction results for multiple time steps within a future target duration based on historical data of the environment surrounding the target vehicle and a prediction model. The trajectory planning unit is used to input the vehicle status data, the historical data of the target vehicle's surrounding environment, and the environmental prediction results into the trajectory planning model to obtain the target planned trajectory of the target vehicle within the target duration; wherein, the target planned trajectory is obtained by the trajectory planning model through recursive iteration based on the vehicle status data, the historical data of the target vehicle's surrounding environment, and the environmental prediction results at multiple time steps within the future target duration according to the temporal dimension; The vehicle control unit is used to control the intelligent driving of the target vehicle based on the target planned trajectory.

[0016] In some exemplary embodiments of this disclosure, the trajectory planning unit is used for: Iterate through each time step sequentially within the target duration: For each current time step, the planned trajectory corresponding to the previous time step, the vehicle status data corresponding to the current time step, the historical data of the target vehicle's surrounding environment, and the environmental prediction results are input into the trajectory planning model to generate the planned trajectory for the current time step. The planned trajectory generated at the current time step is used as the input basis for the next time step. The planned trajectory for the next time step is iteratively generated by combining the vehicle status data corresponding to the next time step, the historical data of the target vehicle's surrounding environment, and the environmental prediction results. The process is repeated iteratively step by step until all time steps within the target duration are traversed, thus obtaining the target planned trajectory.

[0017] In some exemplary embodiments of this disclosure, the trajectory planning unit is used for: The vehicle status data is encoded to obtain first trajectory encoding information; and the planned trajectory and the status data of traffic participants within a first range from the target vehicle are encoded to obtain historical environment encoding information. The environmental prediction results are encoded to obtain future environmental coding information; The first trajectory coding information, historical environment coding information, and future environment coding information are processed for route planning to obtain the planned trajectory at the current time step.

[0018] In some exemplary embodiments of this disclosure, the trajectory planning unit is further configured to: The first trajectory coding information, the historical environment coding information, and the future environment coding information are time-aligned to obtain aligned first trajectory coding information, historical environment coding information, and future environment coding information, so as to perform route planning processing on the aligned first trajectory coding information, historical environment coding information, and future environment coding information.

[0019] In some exemplary embodiments of this disclosure, the historical data of the surrounding environment of the target vehicle further includes a bird's-eye view of the region of interest within a first range from the target vehicle, and the environment prediction unit is further configured to: Based on a bird's-eye view of the region of interest within a first range from the target vehicle and the status data of traffic participants within the first range from the target vehicle, an environmental prediction result for multiple time steps within a future target duration is generated through a prediction model.

[0020] In some exemplary embodiments of this disclosure, the environmental prediction results include one or more of the following: future trajectory data of the traffic participants, traffic light status change data, dynamic obstacle appearance / disappearance data, and road topology change data.

[0021] According to a fourth aspect of the present disclosure, a trajectory planning model training apparatus is provided, the apparatus comprising: The sample acquisition unit is used to acquire training samples, which include at least: vehicle status data, future environmental data for multiple time steps within the planning time, labeled planning trajectory, and historical data of the vehicle's surrounding environment; the historical data of the vehicle's surrounding environment includes at least the status data of traffic participants within a first range from the vehicle. The trajectory acquisition unit is used to input the vehicle state data, the historical data of the vehicle's surrounding environment, and the future environment data into the trajectory planning model to be trained, and obtain the predicted planned trajectory of the vehicle within the planning time period; wherein, the predicted planned trajectory is obtained by the trajectory planning model to be trained through recursive iteration based on the vehicle state data, the historical data of the vehicle's surrounding environment, and the future environment data in a time-series dimension; The model training unit is used to train the trajectory planning model to be trained based on the predicted planned trajectory and the labeled planned trajectory to obtain the target trajectory planning model.

[0022] In some exemplary embodiments of this disclosure, the planned trajectory acquisition unit is used for: Within the planned duration, each time step is traversed sequentially: For each current time step, the planned trajectory corresponding to the previous time step, the vehicle status data corresponding to the current time step, the historical data of the vehicle's surrounding environment, and the future environment data are input into the trajectory planning model to be trained to generate the predicted planned trajectory for the current time step. The planned trajectory generated at the current time step is used as the input basis for the next time step. Combined with the vehicle status data corresponding to the next time step, the historical data and future environmental data of the vehicle's surrounding environment, the predicted planned trajectory for the next time step is generated iteratively. The process is iterated step by step according to time steps until all time steps within the planned duration are traversed to obtain the predicted planned trajectory.

[0023] In some exemplary embodiments of this disclosure, the historical data of the vehicle's surrounding environment further includes a bird's-eye view of a region of interest within a first range from the vehicle, and the planned trajectory acquisition unit is used for: The vehicle status data is encoded to obtain first trajectory encoding information; and the planned trajectory, the status data of traffic participants within a first range from the target vehicle, and the bird's-eye view of the region of interest within a first range from the vehicle are encoded to obtain historical environment encoding information. The environmental prediction results are encoded to obtain future environmental coding information; The first trajectory coding information, historical environment coding information, and future environment coding information are processed for route planning to obtain the predicted planned trajectory at the current time step.

[0024] According to a fifth aspect of the present disclosure, a vehicle is provided, comprising: a processor; a memory for storing processor-executable instructions; wherein the processor is configured to implement the vehicle control method described in any one of the first aspects.

[0025] According to a sixth aspect of the present disclosure, an electronic device is provided, comprising: a processor; and a memory for storing processor-executable instructions; wherein the processor is configured to: implement the vehicle control method as described in any one aspect of the first aspect; or implement the trajectory planning model training method as described in any one aspect of the second aspect.

[0026] According to a seventh aspect of the present disclosure, a non-transitory computer-readable storage medium is provided, storing a computer program that, when executed by a processor, implements the vehicle control method described in any one of the first aspects; or, executes the trajectory planning model training method described in any one of the second aspects.

[0027] According to an eighth aspect of the present disclosure, a computer program product is provided, comprising a computer program that, when executed by a processor, implements the vehicle control method as described in any one of the first aspects; or implements the trajectory planning model training method as described in any one of the second aspects.

[0028] The technical solutions provided by the embodiments of this disclosure may include the following beneficial effects: This disclosure provides a vehicle control method. This method collects the current vehicle status data and historical data of the surrounding environment of the target vehicle. The historical data of the surrounding environment includes at least the status data of traffic participants within a first range from the target vehicle. Then, based on the historical data of the surrounding environment, a prediction model generates environmental prediction results for multiple time steps within a future target duration. Further, the vehicle status data, the historical data of the surrounding environment, and the environmental prediction results can be input into a trajectory planning model to obtain the target planned trajectory of the target vehicle within the target duration. The target planned trajectory is obtained by the trajectory planning model iteratively applying the vehicle status data, the historical data of the surrounding environment, and the environmental prediction results for multiple time steps within the future target duration along a temporal dimension. Clearly, this high-quality target planned trajectory greatly improves the accuracy, stability, and rationality of the target vehicle's trajectory planning. Therefore, based on the target planned trajectory, intelligent driving of the target vehicle can be controlled, ensuring vehicle driving safety.

[0029] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and are not intended to limit this disclosure. Attached Figure Description

[0030] The accompanying drawings, which are incorporated in and form a part of this specification, illustrate embodiments consistent with this disclosure and, together with the description, serve to explain the principles of this disclosure.

[0031] Figure 1 This is a flowchart illustrating a vehicle control method according to an exemplary embodiment of the present disclosure.

[0032] Figure 2 This is a schematic diagram illustrating a trajectory planning model according to an exemplary embodiment of the present disclosure.

[0033] Figure 3 This is a process for obtaining a target planning trajectory according to an exemplary embodiment of the present disclosure. Figure 1 .

[0034] Figure 4 This is a process for obtaining a planned trajectory at the current time step, as illustrated in an exemplary embodiment of this disclosure. Figure 1 .

[0035] Figure 5 This is a process for obtaining a planned trajectory at the current time step, as illustrated in an exemplary embodiment of this disclosure. Figure 2 .

[0036] Figure 6This is a process for obtaining a planned trajectory at the current time step, as illustrated in an exemplary embodiment of this disclosure. Figure 3 .

[0037] Figure 7 This is a schematic diagram illustrating a process for obtaining a target planning trajectory according to an exemplary embodiment of the present disclosure.

[0038] Figure 8 This is a flowchart illustrating a trajectory planning model training method according to an exemplary embodiment of the present disclosure.

[0039] Figure 9 This is a process for obtaining a predicted planning trajectory according to an exemplary embodiment of the present disclosure. Figure 1 .

[0040] Figure 10 This is a process for obtaining a predicted planning trajectory according to an exemplary embodiment of the present disclosure. Figure 2 .

[0041] Figure 11 This is a block diagram illustrating a vehicle control device according to an exemplary embodiment of the present disclosure.

[0042] Figure 12 This is a block diagram illustrating a trajectory planning model training apparatus according to an exemplary embodiment of the present disclosure.

[0043] Figure 13 This is a functional block diagram of a vehicle according to an exemplary embodiment of the present disclosure.

[0044] Figure 14 This is a functional block diagram of an electronic device according to an exemplary embodiment of the present disclosure. Detailed Implementation

[0045] Exemplary embodiments of this disclosure will be described in detail herein, examples of which are illustrated in the accompanying drawings. When the following description refers to the drawings, the same numbers in different drawings denote the same or similar elements unless otherwise indicated. Various changes, modifications, and equivalents of the methods, apparatus, and / or systems described herein will become apparent upon understanding this disclosure. For example, the order of operations described herein is merely illustrative and is not limited to those orders set forth herein, but can be changed as will become apparent upon understanding this disclosure, except for operations that must be performed in a particular order. Furthermore, for clarity and brevity, descriptions of features known in the art may be omitted.

[0046] The embodiments described below, which are examples of some of the embodiments of this disclosure, do not represent all embodiments consistent with this disclosure. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this disclosure as detailed in the appended claims.

[0047] Currently, when performing trajectory prediction in related technologies, the problem of spatiotemporal inconsistency between input information (current data / historical data) and output true value (predicted trajectory affected by future events) is generally alleviated by introducing longer historical sequences or enhanced perception modules. However, this cannot fundamentally solve the learning bias caused by the "invisibility of future events", resulting in problems such as unreasonable, inaccurate and unstable trajectory planning when predicting the future trajectory of vehicles in related technologies.

[0048] In view of the above-mentioned problems, this disclosure provides a vehicle control method. This method collects the current vehicle state data and historical data of the surrounding environment of the target vehicle. The historical data of the surrounding environment includes at least the state data of traffic participants within a first range from the target vehicle. Then, based on the historical data of the surrounding environment, a prediction model generates environmental prediction results for multiple time steps within a future target duration. Further, the vehicle state data, the historical data of the surrounding environment, and the environmental prediction results can be input into a trajectory planning model to obtain the target planned trajectory of the target vehicle within the target duration. The target planned trajectory is obtained by the trajectory planning model iteratively applying the vehicle state data, the historical data of the surrounding environment, and the environmental prediction results for multiple time steps within the future target duration along a temporal dimension. Clearly, this high-quality target planned trajectory greatly improves the accuracy, stability, and rationality of the target vehicle's trajectory planning. Thus, intelligent driving of the target vehicle can be controlled based on its target planned trajectory, ensuring vehicle driving safety.

[0049] The steps of each method in the exemplary embodiments of this disclosure will now be described in more detail with reference to the accompanying drawings and examples.

[0050] Figure 1 This is a flowchart illustrating a vehicle control method according to an exemplary embodiment of the present disclosure. The method of this embodiment can be applied to a vehicle or to a cloud / server. This disclosure uses an in-vehicle application as an example for illustration.

[0051] like Figure 1 As shown, in some embodiments, the vehicle control method of this disclosure includes: In step S110, the vehicle status data of the target vehicle at the current moment and the historical data of the surrounding environment of the target vehicle are collected; the historical data of the surrounding environment of the target vehicle includes at least the status data of traffic participants within a first range from the target vehicle.

[0052] In this embodiment, various sensors and communication devices can be installed on the target vehicle. For example, cameras, lidar, millimeter-wave radar, and inertial measurement units can be installed on the target vehicle. These sensors and communication devices can sense and acquire relevant data about the target vehicle, including its own vehicle status data and historical data about the surrounding environment. The historical data about the surrounding environment includes at least the vehicle status data of traffic participants within a first range from the target vehicle.

[0053] In a possible embodiment, traffic participants are objects within a first range from the target vehicle. These traffic participants may include other vehicles, pedestrians, non-motorized vehicles, or other moving objects. For example, based on the target vehicle's own position, preset distances are defined forward, backward, left, and right, such as rectangular areas of 60 meters forward and backward and 40 meters left and right. This rectangular area constitutes the first range. Any vehicle, pedestrian, non-motorized vehicle, or other moving object that is within this first range in real time is considered a traffic participant. The preset distances can be adaptively adjusted according to actual conditions.

[0054] In possible embodiments, the vehicle status data mainly includes the vehicle's current position, speed, acceleration, steering angle, braking status, lane information, and vehicle attitude (pitch angle, roll angle), etc.; taking a traffic participant including a vehicle as an example, its status data includes at least: position (distance and azimuth angle) relative to the vehicle, speed, acceleration / deceleration, driving direction, lane information, etc.

[0055] In step S120, based on historical data of the environment surrounding the target vehicle, environmental prediction results for multiple time steps within the future target duration are generated through a prediction model.

[0056] In this embodiment of the disclosure, the prediction model takes historical data of the surrounding environment of the target vehicle as input, focuses on predicting changes in the surrounding environment of the target vehicle, and outputs environmental information within a specific future time window. That is, the prediction model generates environmental prediction results for multiple time steps within the future target duration.

[0057] In other words, at the current time step, the prediction model outputs a global environmental time series prediction result for multiple consecutive time steps over the next N seconds, which serves as the environmental prediction result. This result includes the complete traffic environment state at every moment within the next N seconds, representing a continuous prediction sequence covering a long time series. Here, N seconds is the aforementioned target duration, i.e., the duration of the trajectory planning for the target vehicle. This target duration can be a default setting for the vehicle or obtained based on user input. Of course, the target duration can also be adjusted, and this embodiment does not limit this. For example, the target duration could be 5 seconds, 8 seconds, 10 seconds, etc., and this embodiment does not limit this.

[0058] In this embodiment of the disclosure, the state data of traffic participants within the first range of the target vehicle can be input into the prediction model to obtain environmental prediction results for multiple time steps within a future target duration.

[0059] In exemplary embodiments, the prediction model can be built on different architectures; this disclosure uses a Transformer-based architecture. Temporal motion prediction model Let's take an example to illustrate. Leveraging the predictive model's efficient processing capabilities for time-series motion data, it accurately captures the motion patterns of targets around the vehicle (such as other vehicles and pedestrians). Specifically, this predictive model can incorporate a self-attention mechanism, directly modeling long-distance motion relationships between targets, avoiding the frame-by-frame dependence of traditional models on time-series data, and improving the coherence and accuracy of motion trajectory prediction.

[0060] Optionally, the prediction model can also be built based on multi-agent interaction modeling. For the complex interactive behaviors of multiple targets (vehicles, non-motorized vehicles, pedestrians, etc.) in traffic scenarios, by modeling the behavioral relationships and influences between different traffic participants, the behavioral intentions of different "agents" can be analyzed, effectively identifying their cooperative or evasive behaviors in the actual road environment. For example, it can predict the impact of a vehicle changing lanes on vehicles behind, or predict the possible interaction between a pedestrian crossing an intersection and a vehicle going straight, thereby improving the ability to perceive dynamic changes in the surrounding environment and making the prediction results more consistent with the behavioral logic of real traffic scenarios.

[0061] Optionally, the prediction model can integrate two modeling mechanisms. On the one hand, it uses a temporal motion prediction model to detect the motion characteristics of a single traffic participant in the time dimension. On the other hand, it uses multi-agent interaction modeling capabilities to analyze the behavioral intentions and mutual influences of different traffic participants. Finally, it outputs the prediction results of the environment (including dynamic targets) around the target vehicle in a specific time period in the future, providing a highly reliable environmental prediction input for autonomous vehicle trajectory planning.

[0062] In step S130, the vehicle status data, historical data of the target vehicle's surrounding environment, and environmental prediction results are input into the trajectory planning model to obtain the target planned trajectory of the target vehicle within the target duration. The target planned trajectory is obtained by the trajectory planning model through recursive iteration based on the vehicle status data, historical data of the target vehicle's surrounding environment, and environmental prediction results at multiple time steps within the future target duration, according to the temporal dimension.

[0063] In this embodiment of the disclosure, environmental prediction results, vehicle status data, and historical data of the surrounding environment of the target vehicle can be input into the trajectory planning model to obtain the target planned trajectory.

[0064] To plan the trajectory of a vehicle over T time steps, the vehicle control methods include: The vehicle's state data and historical data of the target vehicle's surrounding environment are input into the first encoder to generate historical trajectory encoding information and historical environment encoding information, which are used to characterize the vehicle's state and environmental context at the current moment and within historical time periods. The environmental prediction results generated by the prediction model are input into the second encoder to obtain future environmental context encoding information, which characterizes the environmental state at each future time step. During trajectory generation, an autoregressive decoding mechanism can be used to progressively generate the planned trajectory of the target vehicle for multiple future time steps. At each time step, the decoder receives the following inputs: historical trajectory encoding information, generated trajectory information, and environmental encoding information for the corresponding future time step obtained by masking the current time step based on the future environmental context encoding information.

[0065] Specifically, for trajectory generation at time step t+k, the decoder only accesses the future environment encoding information up to time step t+k, and ensures through a masking mechanism that the model does not depend on future information from t+k+1 onwards when generating the trajectory at time step t+k.

[0066] Furthermore, the historical trajectory encoding information is input into the temporal attention module in the decoder to generate trajectory temporal encoding information. Subsequently, through a cross-attention mechanism, the trajectory temporal encoding information is fused with the future environment encoding information of the current time step and the historical environment encoding information to generate trajectory encoding information for trajectory decision-making.

[0067] During the continuous time step processing, the vehicle status data of the current time step and the historical environment data are input into the first encoder to update the historical trajectory coding information and the historical environment coding information; at the same time, based on the environment coding information corresponding to the next time step in the environment prediction results, the data is input into the second encoder to obtain the future environment coding information for the next time step.

[0068] Repeating the encoding and decoding process described above, trajectory points for the next T time steps are generated step by step. Finally, the trajectory encoding information of the last layer can be mapped through a multilayer perceptron to output the planned trajectory of the target vehicle within a preset planning time period (e.g., T time steps). Then, the intelligent driving of the target vehicle is controlled according to the aforementioned planned trajectory.

[0069] In this embodiment of the disclosure, the trajectory planning model will be introduced first. The training of the trajectory planning model will be introduced later and will not be repeated here.

[0070] Figure 2 This is a schematic diagram illustrating a trajectory planning model according to an exemplary embodiment of the present disclosure.

[0071] like Figure 2 As shown, the trajectory planning model includes a first encoder, a second encoder, and a decoder. The first encoder encodes historical data of the target vehicle's surrounding environment and the vehicle's state data; the second encoder encodes the environmental prediction results; and the decoder decodes the encoder outputs from the first and second encoders to obtain the target planned trajectory.

[0072] In this example embodiment, Figure 2 The first encoder shown includes a first multilayer perceptron (MLP) and a first self attention mechanism module. Figure 2 The second encoder shown includes a second multilayer perceptron and a second self-attention mechanism module. The self-attention mechanism module can be implemented using a Transformer architecture, or it can be implemented using one or more attention structures such as local window self-attention, linear self-attention, sparse self-attention, axial self-attention, and piecewise self-attention; this disclosure does not limit the specific implementation.

[0073] as well as, Figure 2 The decoder shown includes a temporal attention module and a cross-attention mechanism module. The temporal attention module and the cross-attention mechanism module can be implemented using the Transformer architecture, or they can be implemented using one or more attention structures such as segmented temporal attention, local window attention, sparse temporal attention, gated cross-attention, axial temporal attention, and lightweight linear attention. This disclosure does not limit the specific implementation of these structures.

[0074] In this example embodiment, the first encoder can also be called a historical observation encoder, wherein the input of the first encoder is the vehicle state data and the historical data of the target vehicle's surrounding environment, and the output is the vehicle trajectory coding information and the vehicle historical environment coding information.

[0075] In this example embodiment, the second encoder can also be referred to as the future environment encoder. The first encoder takes as input environmental state data for a future time period, i.e., the aforementioned environmental prediction results, and outputs future environment encoded information.

[0076] In this example embodiment, the decoder can be a conditional autoregressive decoder, wherein the decoder inputs are vehicle trajectory coding information, vehicle historical environment coding information, and future environment coding information, and the output is the target planned trajectory.

[0077] In step S140, the target vehicle is controlled to drive intelligently based on the target planned trajectory.

[0078] In this embodiment, by processing environmental prediction results, vehicle status data, and historical data of the surrounding environment of the target vehicle based on a target trajectory planning model, high-quality target trajectory generation can be achieved, greatly improving the accuracy, stability, and rationality of target vehicle trajectory prediction. Furthermore, intelligent driving of the target vehicle can be controlled based on this target trajectory.

[0079] In some exemplary embodiments of this disclosure, historical data on the surrounding environment of the target vehicle also includes a bird's-eye view of a region of interest within a first range from the target vehicle. Figure 1 Based on the vehicle control method shown. Figure 1 Step S120 may include: generating environmental prediction results for multiple time steps within a future target duration through a prediction model based on a bird's-eye view of the region of interest within a first range from the target vehicle and the status data of traffic participants within the first range from the target vehicle.

[0080] In this example embodiment, the bird's-eye view of the region of interest around the vehicle is a global overhead image constructed by projecting a top-down perspective onto the road environment, traffic participants, and infrastructure within a preset range around the target vehicle, centered on the target vehicle. This bird's-eye view can fully present the road topology, lane distribution, road markings, intersection areas, isolation facilities, as well as the spatial position, size outline, and relative distribution of various traffic participants such as motor vehicles, non-motor vehicles, and pedestrians within a certain area around the target vehicle. It can intuitively reflect the overall environmental situation around the vehicle, eliminate the visual blind spots of the conventional eye-level perspective, and provide comprehensive and objective overhead environmental feature information for traffic scene perception, target interaction judgment, and driving situation prediction.

[0081] As can be seen, in this exemplary embodiment, the prediction model takes a bird's-eye view of the region of interest within a first range from the target vehicle and the state data of traffic participants within a first range from the target vehicle as input, predicts the trajectory of moving targets around the target vehicle and the overall changes in the surrounding environment, and outputs environmental information within a specific future time window. That is, the prediction model generates environmental prediction results for multiple time steps within the future target duration.

[0082] In this example embodiment, the environmental prediction results include one or more of the following: future trajectory data of traffic participants, traffic light status change data, dynamic obstacle movement trajectory data, and road topology change data.

[0083] In this exemplary embodiment, the scheme of obtaining environmental prediction results, i.e., future environmental data, based on the fusion of surrounding vehicle state data and bird's-eye view fully utilizes the advantages of bird's-eye view, such as global top-down view, unobstructed view, complete presentation of road topology, lane distribution, traffic markings, and global positional relationships of multiple vehicles. It makes up for the shortcomings of relying solely on historical vehicle state data, such as limited perspective and lack of environmental semantics. It can more accurately capture the interactive behavior of vehicles in complex scenarios such as congestion, lane changing, merging and exiting, and can also perceive road conditions and potential traffic conflicts at a distance in advance, improving the accuracy and completeness of environmental prediction results and providing more comprehensive and reliable environmental input support for autonomous vehicle trajectory planning.

[0084] Figure 3 This is a flowchart illustrating a method for obtaining a target planning trajectory according to an exemplary embodiment of the present disclosure.

[0085] like Figure 3 As shown, in Figure 1 Based on the vehicle control method shown. Figure 1 Step S130 shown may include the following steps.

[0086] In step S310, each time step is traversed sequentially within the target duration.

[0087] In step S320, for each current time step, the planned trajectory corresponding to the previous time step, the vehicle status data corresponding to the current time step, the historical data of the target vehicle's surrounding environment, and the environmental prediction results are input into the trajectory planning model to generate the planned trajectory for the current time step.

[0088] In this embodiment, the environmental prediction result for the current time step refers to the single-step future environmental data for that time step. In other words, the environmental prediction result for each time step refers to the single-step future environmental data for that time step. This not only avoids interference from long-term data but also reduces planning deviations.

[0089] For example, if the current time step is the 5th second, the 10-second long-term environmental prediction result is masked by a masking mechanism to remove other irrelevant time step information, and only the environmental prediction data of the 6th second single time step is retained and extracted as the single-step future environmental data corresponding to the current time step, thereby realizing the accurate use of the long-term prediction result at a single moment.

[0090] In step S330, the planned trajectory generated in the current time step is used as the input basis for the next time step. The planned trajectory for the next time step is generated iteratively by combining the vehicle status data, historical data of the target vehicle's surrounding environment, and environmental prediction results corresponding to the next time step.

[0091] In this embodiment, the planned trajectory generated at the current time step and the vehicle state data, historical data of the target vehicle's surrounding environment, and environmental prediction results at the next time step can be used as input information for generating the planned trajectory at the next time step. This input information is then fed into the trajectory planning model to obtain the planned trajectory for the next time step. In other words, the planned trajectory at each current time step serves as input information for the trajectory planning model at the next time step; that is, the planned trajectory for the adjacent next time step is obtained through iterative recursion using the planned trajectory at the current time step.

[0092] This enhances the coherence of the planned trajectory and, in response to complex road interaction scenarios such as sudden intrusion of lateral targets or unconventional weaving of adjacent vehicles during vehicle operation, the historical data of the newly determined vehicle target vehicle's surrounding environment, combined with the evolution trend of the future environment, can more realistically reflect scene changes, enhance anti-interference capabilities and adaptability to complex scenarios, and make the target planning trajectory predicted based on this more consistent with objective logic.

[0093] In step S340, the time steps are iterated sequentially until all time steps within the target duration are traversed to obtain the target planned trajectory.

[0094] In this embodiment of the disclosure, the time steps can be iterated sequentially until all time steps within the target duration are traversed, thereby obtaining the target planned trajectory.

[0095] Figure 4 This is a flowchart illustrating the process of obtaining the planned trajectory at the current time step according to an exemplary embodiment of the present disclosure.

[0096] like Figure 4 As shown, in Figure 3 Based on the obtained target trajectory shown, Figure 3 Step 320 shown may include the following steps.

[0097] In step S410, the vehicle status data is encoded to obtain first trajectory encoding information; and the status data of traffic participants within the first range of the planned trajectory and the distance to the target vehicle are encoded to obtain historical environment encoding information.

[0098] In this embodiment of the disclosure, the vehicle state data can be encoded by the first encoder in the trajectory planning model to obtain the first trajectory encoding information, and the historical data of the target vehicle's surrounding environment can be encoded by the first encoder in the trajectory planning model to obtain the historical environment encoding information.

[0099] In step S420, the environmental prediction results are encoded to obtain future environmental coding information.

[0100] In this embodiment of the disclosure, the environmental prediction results can be encoded by the second encoder in the trajectory planning model, thereby obtaining the future environmental coding information.

[0101] In step S430, route planning processing is performed on the first trajectory coding information, historical environment coding information, and future environment coding information to obtain the planned trajectory for the current time step.

[0102] In an exemplary embodiment, the first trajectory coding information, historical environment coding information, and future environment coding information can be time-aligned to obtain aligned first trajectory coding information, historical environment coding information, and future environment coding information.

[0103] In this exemplary embodiment, after obtaining the aligned first trajectory coding information, historical environment coding information, and future environment coding information, the aligned first trajectory coding information, vehicle historical environment coding information, and predicted future environment coding information can be input into the decoder in the trajectory planning model to obtain the planned trajectory at the current time step.

[0104] This approach not only avoids misalignment of data from different sources, ensuring consistency of multi-source data (i.e., trajectory data and environmental data) across time and enhancing the effectiveness of data fusion, but also allows trajectory data and environmental data to form a spatiotemporal linkage, reducing fusion errors. Furthermore, the aligned vehicle trajectory coding information, historical vehicle environmental coding information, and future environmental coding information enable the trajectory planning model to capture the patterns of trajectory changes with the environment and avoid trajectory abrupt changes caused by temporal disorder (such as predicting sudden lane changes or speed changes when misaligned). This ensures that the planned trajectory conforms to the physical inertia of motion, thereby improving the accuracy and logical coherence of the planned trajectory.

[0105] Figure 5 This is a flowchart illustrating the process of obtaining the planned trajectory at the current time step according to an exemplary embodiment of the present disclosure.

[0106] like Figure 5 As shown, in Figure 3 Based on the obtained target trajectory shown, Figure 3 Step S320 shown may include the following steps.

[0107] In step S510, the vehicle status data is encoded to obtain first trajectory encoding information; and the planned trajectory, the status data of traffic participants within a first range from the target vehicle, and the bird's-eye view of the region of interest within a first range from the target vehicle are encoded to obtain historical environment encoding information.

[0108] In this embodiment of the disclosure, the vehicle state data can be encoded by the first encoder in the trajectory planning model to obtain the first trajectory encoding information. In addition, the historical data of the target vehicle’s surrounding environment and the bird’s-eye view of the region of interest within a first range from the target vehicle can be encoded by the first encoder in the trajectory planning model to obtain the historical environment encoding information.

[0109] In step S520, the environmental prediction results are encoded to obtain future environmental coding information.

[0110] In this embodiment of the disclosure, the environmental prediction results can be encoded by the second encoder in the trajectory planning model, thereby obtaining the future environmental coding information.

[0111] In step S530, route planning processing is performed on the first trajectory coding information, historical environment coding information, and future environment coding information to obtain the planned trajectory for the current time step.

[0112] In an exemplary embodiment, the first trajectory coding information, historical environment coding information, and future environment coding information can be time-aligned to obtain aligned first trajectory coding information, historical environment coding information, and future environment coding information.

[0113] In this exemplary embodiment, after obtaining the aligned first trajectory coding information, historical environment coding information, and future environment coding information, the aligned first trajectory coding information, vehicle historical environment coding information, and predicted future environment coding information can be input into the decoder in the trajectory planning model to obtain the planned trajectory at the current time step.

[0114] This approach not only avoids misalignment of data from different sources, ensuring consistency of multi-source data (i.e., trajectory data and environmental data) across time and enhancing the effectiveness of data fusion, but also allows trajectory data and environmental data to form a spatiotemporal linkage, reducing fusion errors. Furthermore, the aligned vehicle trajectory coding information, historical vehicle environmental coding information, and future environmental coding information enable the trajectory planning model to capture the patterns of trajectory changes with the environment and avoid trajectory abrupt changes caused by temporal disorder (such as predicting sudden lane changes or speed changes when misaligned). This ensures that the planned trajectory conforms to the physical inertia of motion, thereby improving the accuracy and logical coherence of the planned trajectory.

[0115] Figure 6 This is a flowchart illustrating the planned trajectory at the current time step according to an exemplary embodiment of the present disclosure.

[0116] like Figure 6 As shown, in Figure 4 , Figure 5 Based on the planned trajectory obtained at the current time step, Figure 4 Step S430 shown Figure 5 Step 530 shown may include the following steps.

[0117] In step S610, the first trajectory encoding information is subjected to time-series encoding processing to obtain trajectory time-series encoding information.

[0118] In this embodiment of the disclosure, the decoder can perform time-series encoding processing on the first trajectory encoding information to obtain the first trajectory time-series encoding information.

[0119] This allows for in-depth analysis of the trajectory's dependencies over time (such as the influence of the previous moment's position on the next moment's position), avoiding temporal breaks caused by single encoding. Furthermore, based on the temporal encoding information of the first trajectory, the inertial patterns of the target's motion can be accurately captured (such as the gradual process of vehicle acceleration and deceleration), resulting in a smoother planned trajectory over time and reduced planning errors.

[0120] In step 620, cross-attention processing is performed on the trajectory temporal coding information, historical environment coding information, and future environment coding information to obtain the second trajectory coding information.

[0121] In this embodiment, the decoder can perform cross-attention processing on the first trajectory temporal encoding information, historical environment encoding information, and future environment encoding information to obtain the second trajectory encoding information. This not only solves the problem of heterogeneity in multi-source data but also allows for deep integration of trajectory information with environmental and temporal data, thereby improving the comprehensiveness of trajectory planning.

[0122] In this embodiment of the disclosure, the second vehicle trajectory coding information obtained through temporal coding processing and cross-attention processing can significantly improve the temporal coherence of trajectory planning, the utilization rate of multi-source information, and the adaptability to complex scenarios, thereby making the planning results more accurate and more in line with the actual motion logic.

[0123] In step 630, the planned trajectory for the current time step is obtained based on the second trajectory encoding information.

[0124] In this exemplary embodiment, after obtaining the second trajectory encoding information, the second trajectory encoding information can be converted (i.e., decoded) by a multilayer perceptron to obtain the planned trajectory for the current time step. In other words, the obtained second vehicle trajectory encoding information is a differential representation of the trajectory; therefore, the trajectory in floating-point form can be recovered based on the MLP decoder, thus obtaining the planned trajectory for the current time step. This provides continuously and precisely describing parameters such as position (e.g., x / y coordinates), speed, and heading angle, thereby avoiding the accuracy loss caused by discretization. Furthermore, the accurate numerical output can directly provide a clear reference for intelligent driving path planning, reducing errors in subsequent trajectory adjustments and improving the safety of obstacle avoidance, following, and other decision-making processes.

[0125] For example, such as Figure 7 As shown, Figure 7 This is a schematic diagram illustrating the process of obtaining a target planning trajectory according to an exemplary embodiment of the present disclosure.

[0126] In this embodiment, the system can collect the vehicle state data A of the target vehicle at the current time step, historical data A of the target vehicle's surrounding environment, and an environmental prediction result A determined based on a prediction model. The vehicle state data A and the historical data A of the target vehicle's surrounding environment are then input into a first encoder to obtain historical trajectory coding information A and historical environmental coding information A. Furthermore, the environmental prediction result A is input into a second encoder to obtain future environmental coding information A.

[0127] In this embodiment of the disclosure, the future environment coding information can be masked based on different times to obtain future environment coding information corresponding to different times. For example, if the current time step is 13:45:08 on April 15, 2026, then the future environment coding information A can be masked based on 13:45:09 on April 15, 2026. That is, the coding information of 13:45:09 on April 15, 2026 can be selected from the future environment coding information A to obtain the future environment coding information A1.

[0128] Please continue reading Figure 7Historical trajectory encoding information A can be input into the temporal attention module in the decoder to obtain trajectory temporal encoding information A. Then, cross-attention processing is performed on trajectory temporal encoding information A, future environment encoding information A1, and historical environment encoding information A to obtain trajectory encoding information B. Optionally, the planned trajectory for the current time step can be obtained based on trajectory encoding information B.

[0129] In this embodiment, the target vehicle's state data B, historical data B of the target vehicle's surrounding environment, and trajectory coding information B at the next time step of the current time step can be input into a first encoder to obtain historical trajectory coding information B and historical environment coding information B. Furthermore, future environment coding information A2 can be input into a second encoder to obtain future environment coding information B. The future environment coding information A2 is obtained by masking future environment coding information A based on 13:45:10 on April 15, 2026. The historical environment coding information B includes future environment coding information A1 and historical environment coding information A.

[0130] In this embodiment, historical trajectory encoding information B can be input into the temporal attention module in the decoder to obtain trajectory temporal encoding information B. Then, cross-attention processing is performed on the trajectory temporal encoding information B, future environment encoding information A2, and historical environment encoding information B to obtain trajectory encoding information C. Optionally, the planned trajectory for the next time step can be obtained based on the trajectory encoding information C.

[0131] In this embodiment, the target vehicle's state data C, historical data C of the target vehicle's surrounding environment, and trajectory coding information C at the next time step of the current time step can be input into a first encoder to obtain historical trajectory coding information C and historical environment coding information C. Furthermore, future environment coding information A3 can be input into a second encoder to obtain future environment coding information C. The future environment coding information A2 is obtained by masking future environment coding information A based on 13:45:11 on April 15, 2026. The historical environment coding information C includes future environment coding information A1, future environment coding information A2, and historical environment coding information A.

[0132] In this embodiment of the disclosure, historical trajectory encoding information C can be input into the temporal attention module in the decoder to obtain trajectory temporal encoding information C, and the trajectory temporal encoding information C, future environment encoding information A3 and historical environment encoding information C are subjected to cross-attention processing to obtain trajectory encoding information D.

[0133] In this embodiment of the disclosure, when the planning duration is determined to be 3 seconds, the trajectory coding information D can be converted into data through a multilayer perceptron to obtain the final output target planning trajectory.

[0134] As can be seen, in this embodiment, the decoder can autoregressively generate the trajectory points of the vehicle for the next 3 seconds. Furthermore, through a cross-attention mechanism model, the decoder can "sense" potential environmental changes at each step, thereby rationally planning the current action and ultimately outputting a more realistic future trajectory.

[0135] Figure 8 This is a flowchart illustrating a trajectory planning model training method according to an exemplary embodiment of the present disclosure.

[0136] In step S810, training samples are obtained. The training samples include at least: vehicle status data, future environmental data for multiple time steps within the planning time, labeled planning trajectory, and historical data of the vehicle's surrounding environment. The historical data of the vehicle's surrounding environment includes at least the status data of traffic participants within a first range from the vehicle.

[0137] In this embodiment of the disclosure, various sensors and communication devices can be installed on the vehicle. These sensors and communication devices can sense and acquire relevant vehicle data, thereby enabling the construction of training samples based on the acquired vehicle data.

[0138] In this exemplary embodiment, relevant data for multiple vehicles can be obtained from a database. These vehicles may have the same or different models; this exemplary embodiment does not impose any limitations. The database may include internal databases, external databases, etc. An external database can be understood as a database associated with a vehicle, while an internal database can be understood as data from a system corresponding to the vehicle.

[0139] In this exemplary embodiment, after obtaining relevant data for multiple vehicles, the following operations can be performed on the relevant data for each vehicle: The relevant data for each vehicle is divided according to a preset time period. This preset time period can be divided according to a certain number of preset time periods per day, weekdays and weekends, or weekdays and holidays; this exemplary embodiment does not limit this. Then, for the relevant data of vehicles within each time period, the division time for the data is determined. The relevant data of vehicles before the division time is taken as historical data, and the data after the division time is taken as future data. Next, the historical data is further divided into vehicle state data (i.e., the trajectory data of the vehicle before the division time) and historical data of the target vehicle's surrounding environment. And, the future data is divided into future trajectory data and future environment data. Further, the vehicle state data of one vehicle, the historical data of the target vehicle's surrounding environment, the future environment data, and the future trajectory data are used as a training sample.

[0140] For example, assuming the acquired data for vehicle A is from April 15, 2026 to April 15, 2026, the data for vehicle A can be divided into four time periods each day: 00:00-06:00; 06:00-12:00; 12:00-18:00; and 18:00-24:00. Then, for each of these four time periods, a dividing point can be determined. For example, for the 00:00-06:00 time period, 3:00 AM can be determined as the dividing point. Therefore, the data between 00:00 and 3:00 AM can be designated as historical data 1, and the data between 3:00 and 6:00 AM can be designated as future data 1. Furthermore, historical data 1 is further divided into vehicle status data 1 and historical data of the target vehicle's surrounding environment 1, and future data is divided into future trajectory data 1 and future environment data 1. Furthermore, the vehicle state data 1 of vehicle A, the historical data 1 of the target vehicle's surrounding environment, the future environment data 1, and the future trajectory data 1 are used as a training sample.

[0141] In an exemplary embodiment, the vehicle state data included in a training sample comprises vehicle state data, which includes data such as position, speed, and heading angle. Historical data of the target vehicle's surrounding environment includes vehicle state data of traffic participants within a first range from the vehicle and a bird's-eye view of the region of interest (ROI) around the vehicle. Future environment data includes one or more of the following: future trajectory data of surrounding traffic participants, traffic light status change data, dynamic obstacle appearance / disappearance data, and road topology change data.

[0142] In the embodiments of this disclosure, the aforementioned acquired training samples may be, for example, a training sample set, or a single training sample; this disclosure does not limit the type of training sample. The following description uses the example of training a trajectory planning model based on a single training sample for illustration.

[0143] In step S820, the vehicle state data, historical data of the vehicle's surrounding environment, and future environmental data are input into the trajectory planning model to be trained to obtain the predicted planned trajectory of the vehicle within the planning time. The predicted planned trajectory is obtained by the trajectory planning model to be trained through iterative calculation based on the vehicle state data, historical data of the vehicle's surrounding environment, and future environmental data in a time-series dimension.

[0144] In this embodiment of the disclosure, the training sample, including the vehicle state data, historical data of the target vehicle's surrounding environment, and future environmental data, can be input into the trajectory planning model to be trained, thereby obtaining the predicted planned trajectory of the training sample.

[0145] In step S830, the trajectory planning model to be trained is trained based on the predicted planned trajectory and the labeled planned trajectory to obtain the target trajectory planning model.

[0146] In this embodiment of the disclosure, after obtaining the predicted planned trajectory, the predicted planned trajectory can be compared with the labeled planned trajectories included in the training samples, and the trajectory planning model to be trained can be trained based on the comparison results to obtain the target trajectory planning model.

[0147] In this embodiment of the disclosure, during the model training phase of the trajectory planning model to be trained, future environmental data is explicitly introduced as a conditional input, thereby enabling the trajectory planning model to be trained to learn the correct spatiotemporal causal relationship between future events and the vehicle trajectory, thereby improving the stability and rationality of trajectory prediction during inference based on the trained trajectory planning model.

[0148] Figure 9 This is a flowchart illustrating a method for obtaining a predicted planning trajectory according to an exemplary embodiment of the present disclosure.

[0149] like Figure 9 As shown, in Figure 8 Based on the trajectory planning model training method shown, Figure 8 Step S820 shown may include the following steps.

[0150] In step 910, each time step is traversed sequentially within the planned duration.

[0151] In step 920, for each current time step, the planned trajectory corresponding to the previous time step, the vehicle state data corresponding to the current time step, the historical data of the vehicle's surrounding environment, and the future environment data are input into the trajectory planning model to be trained to generate the predicted planned trajectory for the current time step.

[0152] In step 930, the planned trajectory generated in the current time step is used as the input basis for the next time step. Combined with the vehicle status data, historical data of the vehicle's surrounding environment, and future environmental data corresponding to the next time step, the predicted planned trajectory for the next time step is generated iteratively.

[0153] In step 940, the time steps are iterated sequentially until all time steps within the planned duration are traversed to obtain the predicted planned trajectory.

[0154] In this embodiment of the disclosure, by using the planned trajectory generated at each time step as the input basis for the next time step, the coherence of the obtained predicted planned trajectory can be enhanced. In addition, for complex road interaction scenarios such as the sudden entry of lateral targets and the unconventional intermingling of adjacent vehicles during vehicle driving, the historical data of the newly determined vehicle target vehicle's surrounding environment, combined with the evolution trend of the future environment in the early stage, can more realistically reflect the scene changes, enhance the anti-interference ability and adaptability to complex scenarios, and make the predicted planned trajectory based on this prediction closer to actual needs.

[0155] Figure 10 This is a flowchart illustrating a method for obtaining a predicted planning trajectory according to an exemplary embodiment of the present disclosure.

[0156] like Figure 10 As shown, in Figure 9 Based on the predicted planned trajectory shown, Figure 9 Step S920 shown may include the following steps.

[0157] In step S1010, the vehicle status data is encoded to obtain first trajectory encoding information; and the planned trajectory, the status data of traffic participants within a first range from the target vehicle, and the bird's-eye view of the region of interest within a first range from the vehicle are encoded to obtain historical environment encoding information.

[0158] In step S1020, the environmental prediction results are encoded to obtain future environmental coding information.

[0159] In step S1030, route planning processing is performed on the first trajectory coding information, historical environment coding information, and future environment coding information to obtain the predicted planned trajectory for the current time step.

[0160] In this embodiment of the disclosure, by iterating through future environmental data step by step, the final output predicted trajectory is made closer to the actual trajectory, thereby improving the accuracy of the predicted trajectory.

[0161] The following are embodiments of the apparatus disclosed herein, which can be used to execute embodiments of the method disclosed herein. For details not disclosed in the apparatus embodiments of this disclosure, please refer to the embodiments of the method disclosed herein.

[0162] Figure 11 This is a block diagram illustrating a vehicle control device according to an exemplary embodiment of the present disclosure. The device of this embodiment can be applied to vehicles and / or electronic devices.

[0163] like Figure 11 As shown, the vehicle control device 1100 may include: a data acquisition unit 1101, an environmental prediction unit 1102, a trajectory planning unit 1103, and a vehicle control unit 1104.

[0164] The data acquisition unit 1101 is used to collect the vehicle status data of the target vehicle at the current moment, as well as the historical data of the surrounding environment of the target vehicle; the historical data of the surrounding environment of the target vehicle includes at least the status data of traffic participants within a first range from the target vehicle. The environmental prediction unit 1102 is used to generate environmental prediction results for multiple time steps within a future target duration based on historical data of the environment surrounding the target vehicle and through a prediction model. The trajectory planning unit 1103 is used to input the vehicle status data, the historical data of the target vehicle's surrounding environment, and the environmental prediction results into the trajectory planning model to obtain the target planned trajectory of the target vehicle within the target duration; wherein, the target planned trajectory is obtained by the trajectory planning model through recursive iteration based on the vehicle status data, the historical data of the target vehicle's surrounding environment, and the environmental prediction results at multiple time steps within the future target duration according to the temporal dimension; The vehicle control unit 1104 is used to control the intelligent driving of the target vehicle based on the target planned trajectory.

[0165] In some exemplary embodiments of this disclosure, the trajectory planning unit 1103 is used for: Iterate through each time step sequentially within the target duration: For each current time step, the planned trajectory corresponding to the previous time step, the vehicle status data corresponding to the current time step, the historical data of the target vehicle's surrounding environment, and the environmental prediction results are input into the trajectory planning model to generate the planned trajectory for the current time step. The planned trajectory generated at the current time step is used as the input basis for the next time step. The planned trajectory for the next time step is iteratively generated by combining the vehicle status data corresponding to the next time step, the historical data of the target vehicle's surrounding environment, and the environmental prediction results. The process is repeated iteratively step by step until all time steps within the target duration are traversed, thus obtaining the target planned trajectory.

[0166] In some exemplary embodiments of this disclosure, the trajectory planning unit 1103 is used for: The vehicle status data is encoded to obtain first trajectory encoding information; and the planned trajectory and the status data of traffic participants within a first range from the target vehicle are encoded to obtain historical environment encoding information. The environmental prediction results are encoded to obtain future environmental coding information; The first trajectory coding information, historical environment coding information, and future environment coding information are processed for route planning to obtain the planned trajectory at the current time step.

[0167] In some exemplary embodiments of this disclosure, the trajectory planning unit 1103 is used for: The first trajectory encoding information is subjected to time-series encoding processing to obtain trajectory time-series encoding information; Cross-attention processing is performed on the trajectory temporal coding information, the historical environment coding information, and the future environment coding information to obtain the second trajectory coding information; Based on the second trajectory encoding information, the planned trajectory for the current time step is obtained.

[0168] In some exemplary embodiments of this disclosure, the trajectory planning unit 1103 is further configured to: The first trajectory coding information, the historical environment coding information, and the future environment coding information are time-aligned to obtain aligned first trajectory coding information, historical environment coding information, and future environment coding information, so as to perform route planning processing on the aligned first trajectory coding information, historical environment coding information, and future environment coding information.

[0169] In some exemplary embodiments of this disclosure, the historical data of the surrounding environment of the target vehicle also includes a bird's-eye view of the region of interest within a first range from the target vehicle, and the environment prediction unit 1102 is further configured to: Based on a bird's-eye view of the region of interest within a first range from the target vehicle and the status data of traffic participants within the first range from the target vehicle, an environmental prediction result for multiple time steps within a future target duration is generated through a prediction model.

[0170] In some exemplary embodiments of this disclosure, the environmental prediction results include one or more of the following: future trajectory data of the traffic participants, traffic light status change data, dynamic obstacle appearance / disappearance data, and road topology change data.

[0171] As described above, the device in this embodiment processes environmental prediction results, vehicle status data, and historical data of the surrounding environment of the target vehicle based on a trajectory planning model. This enables the generation of high-quality target planning trajectories, improves the accuracy, stability, and rationality of target vehicle trajectory prediction, and thus enhances vehicle driving safety.

[0172] Regarding the apparatus in the above embodiments, the specific manner in which each unit performs its operation has been described in detail in the embodiments related to the method, and will not be elaborated upon here.

[0173] Figure 12 This is a block diagram illustrating a trajectory planning model training apparatus according to an exemplary embodiment of the present disclosure. The apparatus of this embodiment can be applied to electronic devices.

[0174] like Figure 12 As shown, the trajectory prediction model training device 1200 may include: a sample acquisition unit 1201, a planned trajectory acquisition unit 1202, and a model training unit 1203.

[0175] The sample acquisition unit 1201 is used to acquire training samples, which include at least: vehicle status data, future environmental data for multiple time steps within the planning time, labeled planning trajectory, and historical data of the vehicle's surrounding environment; the historical data of the vehicle's surrounding environment includes at least the status data of traffic participants within a first range from the vehicle. The trajectory planning unit 1202 is used to input the vehicle state data, the historical data of the vehicle's surrounding environment, and the future environment data into the trajectory planning model to be trained, and obtain the predicted planned trajectory of the vehicle within the planning time period; wherein, the predicted planned trajectory is obtained by the trajectory planning model to be trained through recursive iteration based on the vehicle state data, the historical data of the vehicle's surrounding environment, and the future environment data in a time-series dimension. The model training unit 1203 is used to train the trajectory planning model to be trained based on the predicted planned trajectory and the labeled planned trajectory to obtain the target trajectory planning model.

[0176] In some exemplary embodiments of this disclosure, the planned trajectory acquisition unit 1202 is used for: Within the planned duration, each time step is traversed sequentially: For each current time step, the planned trajectory corresponding to the previous time step, the vehicle status data corresponding to the current time step, the historical data of the vehicle's surrounding environment, and the future environment data are input into the trajectory planning model to be trained to generate the predicted planned trajectory for the current time step. The planned trajectory generated at the current time step is used as the input basis for the next time step. Combined with the vehicle status data corresponding to the next time step, the historical data and future environmental data of the vehicle's surrounding environment, the predicted planned trajectory for the next time step is generated iteratively. The process is iterated step by step according to time steps until all time steps within the planned duration are traversed to obtain the predicted planned trajectory.

[0177] In some exemplary embodiments of this disclosure, the historical data of the vehicle's surrounding environment also includes a bird's-eye view of the region of interest within a first range from the vehicle, and the trajectory acquisition unit 1202 is used for: The vehicle status data is encoded to obtain first trajectory encoding information; and the planned trajectory, the status data of traffic participants within a first range from the target vehicle, and the bird's-eye view of the region of interest within a first range from the vehicle are encoded to obtain historical environment encoding information. The environmental prediction results are encoded to obtain future environmental coding information; The first trajectory coding information, historical environment coding information, and future environment coding information are processed for route planning to obtain the predicted planned trajectory at the current time step.

[0178] As described above, in the model training phase of the trajectory planning model to be trained, the device of this embodiment explicitly introduces future environmental data as conditional input, thereby enabling the trajectory planning model to be trained to learn the correct spatiotemporal causal relationship between future events and the vehicle trajectory, and thus improving the stability and rationality of trajectory prediction during inference based on the trained trajectory planning model.

[0179] Regarding the apparatus in the above embodiments, the specific manner in which each unit performs its operation has been described in detail in the embodiments related to the method, and will not be elaborated upon here.

[0180] Figure 13 This is a functional block diagram of a vehicle according to an exemplary embodiment of the present disclosure. For example, vehicle 1300 may be a hybrid vehicle, a non-hybrid vehicle, an electric vehicle, a fuel cell vehicle, or other types of vehicles. Vehicle 1300 may be an intelligent driving vehicle, a semi-intelligent driving vehicle, or a non-intelligent driving vehicle.

[0181] Reference Figure 13 The vehicle 1300 may include various subsystems, such as an infotainment system 1310, a perception system 1320, a decision control system 1330, a drive system 1340, and a computing platform 1350. The vehicle 1300 may also include more or fewer subsystems, and each subsystem may include multiple components. Furthermore, each subsystem and component of the vehicle 1300 can be interconnected via wired or wireless means.

[0182] In some embodiments, the infotainment system 1310 may include a communication system, an entertainment system, and a navigation system, etc.

[0183] The perception system 1320 may include several sensors for sensing information about the environment surrounding the vehicle 1300. For example, the perception system 1320 may include a global positioning system (which may be a GPS system, a BeiDou system, or another positioning system), an inertial measurement unit (IMU), a lidar, a millimeter-wave radar, an ultrasonic radar, and a camera device.

[0184] The decision control system 1330 may include a computing system, a vehicle controller, a steering system, a throttle, and a braking system.

[0185] The drive system 1340 may include components that provide powered motion to the vehicle 1300. In one embodiment, the drive system 1340 may include an engine, an energy source, a transmission system, and wheels. The engine may be one or a combination of internal combustion engines, electric motors, and compressed air engines. The engine is capable of converting energy provided by the energy source into mechanical energy.

[0186] Some or all of the functions of the vehicle 1300 are controlled by a computing platform 1350. The computing platform 1350 may include at least one processor 1351 and a memory 1352, the processor 1351 being able to execute instructions 1353 stored in the memory 1352.

[0187] The processor 1351 can be any conventional processor, including, for example, a graphics processing unit (GPU), a field programmable gate array (FPGA), a system on chip (SOC), an application specific integrated circuit (ASIC), or a combination thereof.

[0188] The memory 1352 can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk or optical disk.

[0189] In addition to instruction 1353, memory 1352 can also store data, such as road maps, route information, vehicle position, direction, speed, and other data. The data stored in memory 1352 can be used by computing platform 1350.

[0190] In this embodiment of the disclosure, the processor 1351 may execute instructions 1353 to complete all or part of the steps of the above-described vehicle control method and trajectory planning model training method.

[0191] Reference Figure 14 The electronic device 1400 may include one or more of the following components: a processing component 1402, a memory 1404, a power supply component 1406, a multimedia component 1408, an audio component 1410, an input / output (I / O) interface 1412, a sensor component 1414, and a communication component 1416.

[0192] Processing component 1402 typically controls the overall operation of electronic device 1400, such as operations associated with display, telephone calls, data communication, camera operation, and recording. Processing component 1402 may include one or more processors 1420 to execute instructions to complete all or part of the steps of the vehicle control method and trajectory prediction model training method described above. Furthermore, processing component 1402 may include one or more modules to facilitate interaction between processing component 1402 and other components. For example, processing component 1402 may include a multimedia module to facilitate interaction between multimedia component 1408 and processing component 1402.

[0193] Memory 1404 is configured to store various types of data to support the operation of electronic device 1400. Memory 1404 can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0194] Power supply component 1406 provides power to various components of electronic device 1400. Power supply component 1406 may include a power management system, one or more power supplies, and other components associated with generating, managing, and distributing power to electronic device 1400.

[0195] Multimedia component 1408 includes a screen that provides an output interface between the electronic device 1400 and the user. In some embodiments, the screen may include a liquid crystal display (LCD) and a touch panel (TP). In some embodiments, multimedia component 1408 includes a front-facing camera and / or a rear-facing camera.

[0196] Audio component 1410 is configured to output and / or input audio signals. For example, audio component 1410 includes a microphone (MIC) configured to receive external audio signals when electronic device 1400 is in an operating mode, such as call mode, recording mode, and voice recognition mode. The received audio signals may be further stored in memory 1404 or transmitted via communication component 1416. In some embodiments, audio component 1410 also includes a speaker for outputting audio signals.

[0197] I / O interface 1412 provides an interface between processing component 1402 and peripheral interface modules, such as keyboards, click wheels, buttons, etc. These buttons may include, but are not limited to, home buttons, volume buttons, power buttons, and lock buttons.

[0198] Sensor assembly 1414 includes one or more sensors for providing state assessments of various aspects of electronic device 1400. For example, sensor assembly 1414 may detect the on / off state of electronic device 1400, the relative positioning of components such as the display and keypad of electronic device 1400, changes in position of electronic device 1400 or a component of electronic device 1400, the presence or absence of user contact with electronic device 1400, orientation or acceleration / deceleration of electronic device 1400, and temperature changes of electronic device 1400. Sensor assembly 1414 may include a proximity sensor configured to detect the presence of nearby objects without any physical contact. Sensor assembly 1414 may also include a light sensor for use in imaging applications. In some embodiments, sensor assembly 1414 may further include an accelerometer, a gyroscope, a magnetometer, a pressure sensor, or a temperature sensor.

[0199] The communication component 1416 is configured to facilitate wired or wireless communication between the electronic device 1400 and other devices.

[0200] In some embodiments of this disclosure, the electronic device 1400 may be implemented by one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field-programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors, or other electronic components to perform the methods described above.

[0201] In some embodiments of this disclosure, a non-transitory computer-readable storage medium stores a computer program, which, when executed by a processor, implements all or part of the steps of the above-described vehicle control method and trajectory planning model training method.

[0202] In some embodiments of this disclosure, a computer program is provided, which, when executed by a processor, implements all or part of the steps of the above-described vehicle control method and trajectory planning model training method.

[0203] Other embodiments of this disclosure will readily occur to those skilled in the art upon consideration of the specification and practice of the invention disclosed herein. This application is intended to cover any variations, uses, or adaptations of this disclosure that follow the general principles of this disclosure and include common knowledge or customary techniques in the art not disclosed herein. The specification and embodiments are to be considered exemplary only, and it should be understood that this disclosure is not limited to the precise structures described above and shown in the drawings, and various modifications and changes can be made without departing from its scope. The true scope and spirit of this disclosure are indicated by the appended claims.

Claims

1. A vehicle control method, characterized in that, The method includes: Collect the target vehicle's current status data and historical data of the target vehicle's surrounding environment; the historical data of the target vehicle's surrounding environment includes at least the status data of traffic participants within a first range from the target vehicle; Based on historical data of the surrounding environment of the target vehicle, environmental prediction results for multiple time steps within the future target duration are generated through a prediction model. The vehicle status data, historical data of the target vehicle's surrounding environment, and the environmental prediction results are input into the trajectory planning model to obtain the target planned trajectory of the target vehicle within the target duration; wherein, the target planned trajectory is obtained by the trajectory planning model through recursive iteration based on the vehicle status data, historical data of the target vehicle's surrounding environment, and environmental prediction results at multiple time steps within the future target duration according to the temporal dimension; Based on the target planned trajectory, the target vehicle is controlled to drive intelligently.

2. The method according to claim 1, characterized in that, The step of inputting the vehicle status data, historical data of the target vehicle's surrounding environment, and the environmental prediction results into the trajectory planning model to obtain the target vehicle's planned trajectory within the target time period includes: Iterate through each time step sequentially within the target duration: For each current time step, the planned trajectory corresponding to the previous time step, the vehicle status data corresponding to the current time step, the historical data of the target vehicle's surrounding environment, and the environmental prediction results are input into the trajectory planning model to generate the planned trajectory for the current time step. The planned trajectory generated at the current time step is used as the input basis for the next time step. The planned trajectory for the next time step is iteratively generated by combining the vehicle status data corresponding to the next time step, the historical data of the target vehicle's surrounding environment, and the environmental prediction results. The process is repeated iteratively step by step until all time steps within the target duration are traversed, thus obtaining the target planned trajectory.

3. The method according to claim 2, characterized in that, For each current time step, the planned trajectory corresponding to the previous time step, the vehicle state data corresponding to the current time step, the historical data of the target vehicle's surrounding environment, and the environmental prediction results are input into the trajectory planning model to generate the planned trajectory for the current time step, including: The vehicle status data is encoded to obtain first trajectory encoding information; and the planned trajectory and the status data of traffic participants within a first range from the target vehicle are encoded to obtain historical environment encoding information. The environmental prediction results are encoded to obtain future environmental coding information; The first trajectory coding information, historical environment coding information, and future environment coding information are processed for route planning to obtain the planned trajectory at the current time step.

4. The method according to claim 3, characterized in that, The first trajectory encoding information, historical environment encoding information, and future environment encoding information are processed for route planning to obtain the planned trajectory at the current time step, including: The first trajectory encoding information is subjected to time-series encoding processing to obtain trajectory time-series encoding information; Cross-attention processing is performed on the trajectory temporal coding information, the historical environment coding information, and the future environment coding information to obtain the second trajectory coding information; Based on the second trajectory encoding information, the planned trajectory for the current time step is obtained.

5. The method according to claim 3, characterized in that, Before performing route planning processing on the first trajectory coding information, historical environment coding information, and future environment coding information, the method further includes: The first trajectory coding information, the historical environment coding information, and the future environment coding information are time-aligned to obtain aligned first trajectory coding information, historical environment coding information, and future environment coding information, so as to perform route planning processing on the aligned first trajectory coding information, historical environment coding information, and future environment coding information.

6. The method according to claim 1, characterized in that, The historical data of the target vehicle's surrounding environment also includes a bird's-eye view of the region of interest within a first range from the target vehicle. The generation of environmental prediction results at multiple time steps within the future target duration based on the historical data of the target vehicle's surrounding environment using a prediction model includes: Based on a bird's-eye view of the region of interest within a first range from the target vehicle and the status data of traffic participants within the first range from the target vehicle, an environmental prediction result for multiple time steps within a future target duration is generated through a prediction model.

7. The method according to any one of claims 1-6, characterized in that, The environmental prediction results include one or more of the following: the future trajectory data of the traffic participants, traffic light status change data, dynamic obstacle appearance / disappearance data, and road topology change data.

8. A method for training a trajectory planning model, characterized in that, The method includes: Acquire training samples, which include at least: vehicle status data, future environmental data for multiple time steps within the planning duration, labeled planning trajectory, and historical data of the vehicle's surrounding environment; the historical data of the vehicle's surrounding environment includes at least the status data of traffic participants within a first range from the vehicle. The vehicle status data, historical data of the vehicle's surrounding environment, and future environment data are input into the trajectory planning model to be trained to obtain the predicted planned trajectory of the vehicle within the planning time period; wherein, the predicted planned trajectory is obtained by the trajectory planning model to be trained through recursive iteration based on the vehicle status data, historical data of the vehicle's surrounding environment, and future environment data in a time-series dimension. The trajectory planning model to be trained is trained based on the predicted planned trajectory and the labeled planned trajectory to obtain the target trajectory planning model.

9. The method according to claim 8, characterized in that, The step of inputting the vehicle state data, historical data of the vehicle's surrounding environment, and future environment data into the trajectory planning model to be trained, and obtaining the predicted planned trajectory of the vehicle within the planning time period, includes: Within the planned duration, each time step is traversed sequentially: For each current time step, the planned trajectory corresponding to the previous time step, the vehicle status data corresponding to the current time step, the historical data of the vehicle's surrounding environment, and the future environment data are input into the trajectory planning model to be trained to generate the predicted planned trajectory for the current time step. The planned trajectory generated at the current time step is used as the input basis for the next time step. Combined with the vehicle status data corresponding to the next time step, the historical data and future environmental data of the vehicle's surrounding environment, the predicted planned trajectory for the next time step is generated iteratively. The process is iterated step by step according to time steps until all time steps within the planned duration are traversed to obtain the predicted planned trajectory.

10. The method according to claim 9, characterized in that, The historical data of the vehicle's surrounding environment also includes a bird's-eye view of the region of interest within a first range from the vehicle. For each current time step, the planned trajectory from the previous time step, the vehicle's state data for the current time step, the historical data of the vehicle's surrounding environment, and the future environmental data are input into the trajectory planning model to be trained to generate a predicted planned trajectory for the current time step, including: The vehicle status data is encoded to obtain first trajectory encoding information; and the planned trajectory, the status data of traffic participants within a first range from the target vehicle, and the bird's-eye view of the region of interest within a first range from the vehicle are encoded to obtain historical environment encoding information. The environmental prediction results are encoded to obtain future environmental coding information; The first trajectory coding information, historical environment coding information, and future environment coding information are processed for route planning to obtain the predicted planned trajectory at the current time step.

11. A vehicle control device, characterized in that, The device includes: The data acquisition unit is used to collect the vehicle status data of the target vehicle at the current moment, as well as the historical data of the surrounding environment of the target vehicle; the historical data of the surrounding environment of the target vehicle includes at least the status data of traffic participants within a first range from the target vehicle. An environmental prediction unit is used to generate environmental prediction results for multiple time steps within a future target duration based on historical data of the environment surrounding the target vehicle and a prediction model. The trajectory planning unit is used to input the vehicle status data, the historical data of the target vehicle's surrounding environment, and the environmental prediction results into the trajectory planning model to obtain the target planned trajectory of the target vehicle within the target duration; wherein, the target planned trajectory is obtained by the trajectory planning model through recursive iteration based on the vehicle status data, the historical data of the target vehicle's surrounding environment, and the environmental prediction results at multiple time steps within the future target duration according to the temporal dimension; The vehicle control unit is used to control the intelligent driving of the target vehicle based on the target planned trajectory.

12. The apparatus according to claim 11, characterized in that, The trajectory planning unit is used for: Iterate through each time step sequentially within the target duration: For each current time step, the planned trajectory corresponding to the previous time step, the vehicle status data corresponding to the current time step, the historical data of the target vehicle's surrounding environment, and the environmental prediction results are input into the trajectory planning model to generate the planned trajectory for the current time step. The planned trajectory generated at the current time step is used as the input basis for the next time step. The planned trajectory for the next time step is iteratively generated by combining the vehicle status data corresponding to the next time step, the historical data of the target vehicle's surrounding environment, and the environmental prediction results. The process is repeated iteratively step by step until all time steps within the target duration are traversed, thus obtaining the target planned trajectory.

13. The apparatus according to claim 12, characterized in that, The trajectory planning unit is used for: The vehicle status data is encoded to obtain first trajectory encoding information; and the planned trajectory and the status data of traffic participants within a first range from the target vehicle are encoded to obtain historical environment encoding information. The environmental prediction results are encoded to obtain future environmental coding information; The first trajectory coding information, historical environment coding information, and future environment coding information are processed for route planning to obtain the planned trajectory at the current time step.

14. The apparatus according to claim 11, characterized in that, The historical data of the surrounding environment of the target vehicle also includes a bird's-eye view of the region of interest within a first range from the target vehicle. The environment prediction unit is further used for: Based on a bird's-eye view of the region of interest within a first range from the target vehicle and the status data of traffic participants within the first range from the target vehicle, an environmental prediction result for multiple time steps within a future target duration is generated through a prediction model.

15. A trajectory planning model training device, characterized in that, The device includes: The sample acquisition unit is used to acquire training samples, which include at least: vehicle status data, future environmental data for multiple time steps within the planning time, labeled planning trajectory, and historical data of the vehicle's surrounding environment; the historical data of the vehicle's surrounding environment includes at least the status data of traffic participants within a first range from the vehicle. The trajectory acquisition unit is used to input the vehicle state data, the historical data of the vehicle's surrounding environment, and the future environment data into the trajectory planning model to be trained, and obtain the predicted planned trajectory of the vehicle within the planning time period; wherein, the predicted planned trajectory is obtained by the trajectory planning model to be trained through recursive iteration based on the vehicle state data, the historical data of the vehicle's surrounding environment, and the future environment data in a time-series dimension; The model training unit is used to train the trajectory planning model to be trained based on the predicted planned trajectory and the labeled planned trajectory to obtain the target trajectory planning model.

16. The apparatus according to claim 15, characterized in that, The planned trajectory acquisition unit is used for: Within the planned duration, each time step is traversed sequentially: For each current time step, the planned trajectory corresponding to the previous time step, the vehicle status data corresponding to the current time step, the historical data of the vehicle's surrounding environment, and the future environment data are input into the trajectory planning model to be trained to generate the predicted planned trajectory for the current time step. The planned trajectory generated at the current time step is used as the input basis for the next time step. Combined with the vehicle status data corresponding to the next time step, the historical data and future environmental data of the vehicle's surrounding environment, the predicted planned trajectory for the next time step is generated iteratively. The process is iterated step by step according to time steps until all time steps within the planned duration are traversed to obtain the predicted planned trajectory.

17. The apparatus according to claim 16, characterized in that, The historical data of the vehicle's surrounding environment also includes a bird's-eye view of the area of ​​interest within a first range from the vehicle. The planned trajectory acquisition unit is used for: The vehicle status data is encoded to obtain first trajectory encoding information; and the planned trajectory, the status data of traffic participants within a first range from the target vehicle, and the bird's-eye view of the region of interest within a first range from the vehicle are encoded to obtain historical environment encoding information. The environmental prediction results are encoded to obtain future environmental coding information; The first trajectory coding information, historical environment coding information, and future environment coding information are processed for route planning to obtain the predicted planned trajectory at the current time step.

18. A vehicle, characterized in that, include: processor; Memory used to store processor-executable instructions; The processor is configured to implement the vehicle control method according to any one of claims 1 to 7.

19. An electronic device, characterized in that, include: processor; Memory used to store processor-executable instructions; The processor is configured to: implement the vehicle control method according to any one of claims 1 to 7; or implement the trajectory planning model training method according to claims 8 to 10.

20. A non-transitory computer-readable storage medium, characterized in that, The device stores a computer program, which, when executed by a processor, implements the vehicle control method as described in any one of claims 1 to 7; or implements the trajectory planning model training method as described in claims 8 to 10.

21. A computer program product, characterized in that, The method includes a computer program that, when executed by a processor, implements the vehicle control method as described in any one of claims 1 to 7; or, implements the trajectory planning model training method as described in claims 8 to 10.

Images