3120 results about "Autopilot" patented technology

The invention relates to the technical field of automatic driving, and discloses an automatic driving lane changing trajectory planning method based on deep learning, and the method comprises the steps: carrying out the data collection and preprocessing of a multi-modal sensor; performing spatial feature extraction and time sequence modeling on the preprocessed multi-modal data by adopting a CNN-LSTM hybrid architecture, performing feature fusion through an attention mechanism, and outputting a first feature extraction vector; taking the detected vehicles as graph nodes to construct a traffic graph, learning an interaction relationship between the vehicles through a graph attention network and a message passing mechanism, and calculating a scene urgency score and a safety score; generating a lane changing decision based on the deep Q network and the strategy gradient; and generating a trajectory based on the generative adversarial network. The technical problems that an existing lane changing track planning method cannot adapt to the dynamic traffic environment, lacks the ability of understanding complex multi-vehicle interaction and is difficult to balance safety and urgent conflict requirements are solved, and intelligent, safe and efficient automatic driving lane changing track planning is achieved.

The invention provides an end-to-end automatic driving decision control method and system, and belongs to the technical field of automatic driving. The invention relates to an end-to-end automatic driving decision control method based on multi-modal perception and hierarchical trajectory optimization, and the method comprises the steps: constructing an end-to-end decision closed loop through combining the zero sample migration capability of a vision-language-action (VLA) model with a hierarchical optimization architecture: analyzing multi-modal input (vision, language and point cloud) by using a pre-trained VLA model to generate path points; the vehicle pose is dynamically adjusted through upper-layer optimization to expand a feasible solution space, a smooth track meeting dynamics and collision avoidance constraints is solved in real time through lower-layer optimization, and finally a vehicle control instruction is output. According to the method, a multi-modal sensing and hierarchical trajectory optimization mechanism is fused, and the sensing generalization ability, the path planning feasibility and the control execution robustness of the system in a complex traffic environment are effectively improved.

The present application relates to the technical field of autonomous driving functions, and in particular to a control method and device integrating predictive cruise and lane changing decision-making. The method comprises: receiving the accelerations, speeds and positions of a controlled autonomous driving vehicle and surrounding traffic vehicles at a current moment so as to construct an IDM microscopic following model to predict surrounding traffic vehicle states in a preset future long time domain; on the basis of the surrounding traffic vehicle states in the preset future long time domain, constructing an optimization cost function, so as to plan a lane keeping strategy or lane changing strategy; on the basis of the lane keeping strategy or lane changing strategy, using a quintic polynomial to solve for a reference path; on the basis of the reference path, establishing an optimal control problem cost function for continuous intersection predictive cruise control, so as to solve for the optimal cruise speed within a traveling lane; and sending the reference path and the optimal cruise speed within the traveling lane to the controlled autonomous driving vehicle for control. Thus, the problems that existing lane changing decision-making only considers longitudinal motion and fails to pay attention to a development trend of the surrounding environment, possible dangers and the like are solved.

The invention discloses a unified framework for solving automatic driving track prediction and planning consistency based on a world model. According to the method, through cooperative work of the automatic driving domain controller and the vehicle-mounted sensing system, end-to-end joint optimization of track prediction and planning in a complex traffic scene is realized, time sequence dependence and interaction dynamics among intelligent agents are accurately captured, and the prediction capability and robustness of a model are remarkably improved. The method comprises the following specific steps: firstly, constructing a generative world model, and generating potential future state representation by utilizing a behavior conditional and backtracking expansion technology; secondly, in combination with global modeling and a local convolutional network, multi-scale features are extracted, adaptive fusion is carried out, and a multi-modal prediction trajectory is generated; then, a multi-target planning model is adopted to integrate various driving indexes, and a track with the minimum loss function is generated; finally, path planning parameters are dynamically optimized through real-time environment perception and decision feedback, and the problems of prediction uncertainty and planning consistency of the automatic driving track are effectively solved.

The invention relates to an automatic driving risk quantification method based on a conflict risk field. Comprising the steps of 1, constructing a basic risk field; step 2, under a basic risk field framework, constructing a conflict risk field by taking an ADV as a center; step 3, regarding the conflict risk field as a repulsive force acting on the ADV, and using the repulsive force to quantify the dynamic influence of the environmental elements and various TP states on the driving safety of the ADV; the method comprises the following steps: firstly, constructing a basic risk field on the basis of traffic vehicle distribution and motion characteristics, further forming a vehicle forward, lateral, backward and traffic regulation constraint multi-dimensional conflict risk field on the basis, and converting an abstract traffic conflict relationship into a computable repulsive force model; dynamic quantitative characterization of driving risks in a state dimension, a space dimension and a time dimension is realized, and continuous optimization and evolution of an end-to-end automatic driving algorithm are supported by means of a repulsive force model constructed based on a conflict risk field.

The invention discloses an Internet of Vehicles network selection and switching decision system supporting access of multiple operators. According to the method and the device, the timeliness and the continuity of network switching are remarkably improved through prediction and planning in advance. The module depends on a trajectory-network matching prediction sub-module, combines a vehicle navigation trajectory and a global digital twinborn model, identifies a base station coverage area in a future driving path and the network service quality of each operator in advance, and avoids the lag problem that switching is triggered only after signals are weakened in traditional passive switching. The pre-switching resource reservation sub-module further sends a resource request to a target operator base station and confirms reservation before the vehicle enters a new road section, ensures that resources such as bandwidth and time slot required during switching are in place, reduces switching failure or interruption caused by resource competition, avoids common communication interruption after signal sudden drop in traditional switching, and improves the switching efficiency. And continuous transmission of key businesses such as automatic driving control instructions and real-time road conditions is ensured.

The invention discloses an intelligent automobile cooperative cruise safety control method under Dos attacks and physical faults, and relates to automobile intelligent safety and automatic driving. A hierarchical control framework is adopted and comprises an observer layer and a tracking layer. The method comprises the following steps: establishing a vehicle longitudinal dynamics model and a DoS attack model for DoS attack in a V2X communication network and the problems of sensor and actuator faults and parameter isomerism of a vehicle; a completely distributed self-adaptive preset time observer based on event triggering is designed for each following vehicle in the observer layer, and rapid and accurate estimation of the state of the pilot vehicle is achieved; the method comprises the following steps: constructing an augmentation system at a tracking layer, designing a distributed intermediate observer, carrying out online estimation and compensation on faults of a sensor and an actuator, designing a distributed active fault-tolerant controller based on a fault estimation value and a pilot vehicle state estimation value, and calculating a wheel driving torque to realize safe cruise control. Multiple threats of coexistence of network attacks and physical faults are effectively handled, and the stability and safety of the cooperative cruise system are ensured.

A method performed by an apparatus for controlling autonomous driving of a vehicle is introduced. The method may comprise generating, based on a segmentation model processing point cloud data, a first semantic grid map, generating, based on an object detection model, a second semantic grid map, adjusting a probability regarding whether occupancy exists for an element included in each grid of the first semantic grid map and the second semantic grid map, and generating a fused grid map by determining, as a representative label, at least one label corresponding to a highest value among final probabilities of the at least one label, wherein the final probabilities are determined based on whether the at least one label matches the element, outputting, based on the fused grid map, a signal, and controlling, based on the signal, autonomous driving of the vehicle.

The invention is oriented to the field of autonomous vehicle perception planning, and discloses an autonomous vehicle trajectory prediction method fusing a Mama backbone network and a graph attention mechanism. According to the method, vehicle track, state, road and environment information is acquired, a Mamba backbone network is utilized to extract time sequence characteristics of tracks of an automatic driving vehicle and surrounding vehicles, and meanwhile, a spatial interaction relationship between the surrounding vehicles and other objects in a traffic scene is modeled through a graph attention network. In the feature fusion layer, the time sequence and space features are pooled to form high-dimensional space-time coding features. And the fusion features are input into a decoder, the decoder uses a Mamba module as a core to decode and output the information of the model coding, and the predicted trajectory of the vehicles around the autonomous vehicle is obtained. The provided model has the advantages that the mapping relation between the historical track and the predicted track is accurately modeled through deep learning, the mode of combining a Mama backbone network and a graph attention mechanism is utilized, the model is helped to understand the influence of scene features on the future driving track of the vehicle, and accurate, efficient and scene-robust prediction is achieved.

The invention relates to the field of automatic driving vehicle control, in particular to a sightseeing vehicle anti-collision control method and system based on a multi-mode radar. Comprising the following steps: acquiring a multi-modal radar original signal, and performing timestamp alignment and noise reduction processing to generate standardized fusion data; performing multi-target classification based on a convolutional neural network, and outputting a target recognition result in combination with Doppler frequency shift compensation and distance calculation; generating a hierarchical response strategy through dual-threshold comparison and environment interference dynamic weight adjustment; the strategy is converted into a PWM control signal, and a control instruction is generated through relay driving adaptation and hardware compatibility detection; a relay is driven in real time, motor response is collected, and execution effect evaluation and time sequence marking are completed; and brake release and system parameter synchronization are realized through PID closed-loop control. The problems of insufficient multi-target identification precision, poor environmental interference adaptability and response delay are solved, and the active safety performance of the sightseeing vehicle is remarkably improved.

The invention provides an anthropomorphic automatic driving simulation test scene construction method. The method comprises the steps that all vehicles in a traffic scene are divided into two types of intelligent agents including a test vehicle and an environment vehicle, the traffic scene is modeled into a Markov decision process, the test vehicle is a vehicle controlled by an automatic driving algorithm, and the environment vehicle is a vehicle controlled by a GAIL-GRU driving strategy model; a Markov decision process is utilized to extract a driving track from the human driving data set, and an expert track data set is generated; and training a GAIL-GRU driving strategy model by using the expert track data set, realizing interaction between an environment vehicle and a test vehicle by using the trained GAIL-GRU driving strategy model, and constructing a simulation test scene. The automatic driving simulation test scene constructed by the invention can effectively expose decision defects of an automatic driving algorithm in a complex interaction situation, has good anthropomorphism and relatively high risk scene coverage, and provides support for a decision control simulation test of a high-level automatic driving vehicle.

The invention discloses an automatic driving track generation method and device, equipment and a medium. The method comprises the steps that target state information of a vehicle is obtained, and the target state information comprises vehicle body state information, obstacle information, road structure information and traffic state information; the target state information is input into a trajectory planning model for trajectory generation, a plurality of candidate driving trajectories and trajectory evaluation results corresponding to the candidate driving trajectories are determined, and the trajectory planning model is constructed based on a deep neural network; and determining a target driving trajectory based on the plurality of candidate driving trajectories and the trajectory evaluation result corresponding to each candidate driving trajectory. According to the method, the automatic driving track of the vehicle can be automatically and accurately generated, emergencies in dynamic traffic are covered, the generalization ability during cross-scene migration is improved, the track generation flexibility is improved, and therefore the track precision and safety in complex scenes are guaranteed.

A vehicle safety system of an autonomous vehicle may determine predicted velocity vectors for a potential collision, and use the velocity vectors to determine a trajectory for the autonomous vehicle to traverse the environment. The vehicle safety system may analyze sensor data to determine a likelihood of a potential collision with a dynamic object in the environment. Predicted velocity vectors may be determined for the autonomous vehicle and the dynamic object at a time and / or location associated with the potential collision. The predicted velocity vectors may be used to determine a point of impact, relative angle, and / or relative velocity between the autonomous vehicle and dynamic object at the potential collision. The likelihood of the potential collision and the predicted velocity vectors may be used to determine a trajectory for the autonomous vehicle, which may include following a current trajectory or transitioning to one or more contingent trajectories.

The embodiment of the invention provides an automatic driving method, device and system, a model fine adjustment method, device and system and a vehicle, and relates to the field of automatic driving, the method comprises the steps that system prompt information and multi-modal data of an automatic driving task are acquired, and the system prompt information indicates that a reasoning mode is selected according to scene information; processing the multi-modal data and the system prompt information by using a VLA model to obtain current scene information of the automatic driving task and a current reasoning mode corresponding to the current scene information; in a current reasoning mode of the VLA model, processing the multi-modal data to obtain a current processing result; and executing an automatic driving task according to the current processing result. According to the scheme, the problems of insufficient cross-scene reasoning flexibility and low reasoning efficiency of the VLA model can be solved.

The invention discloses an automatic driving path planning method based on multi-sensor fusion, and relates to the technical field of automatic driving, and the method comprises the steps: collecting environment, vehicle and positioning data through multi-source sensors, such as a laser radar and a camera; mapping a vehicle body coordinate system through coordinate conversion, and carrying out downsampling, correction enhancement and denoising preprocessing on data; extracting multi-sensor features, and distributing weights by means of a cross-modal attention mechanism to complete multi-modal fusion; recognizing obstacles and passable areas, and predicting a 5s track of a dynamic obstacle; 5-8 initial paths are generated based on the coordinate system and dynamic constraints; constructing a multi-objective function to screen an optimal path; and finally, a steering angle and an acceleration instruction are calculated through model prediction control, and accurate path tracking is realized. The sensing robustness is improved through multi-sensor fusion, the dynamic weight adjustment adapts to the complex environment, the emergency obstacle avoidance mechanism is quick in response, and the automatic driving reliability and the scene adaptability are integrally enhanced.

The invention provides an off-line test method and an off-line test device for a vehicle steering system, and the method comprises the steps: controlling a vehicle to run according to a preset test track through activating an automatic driving mode of the vehicle; in the running process of the vehicle, running state data of the vehicle are collected in real time; and comparing the operation state data with a preset quality standard, and generating an offline test result of the vehicle steering system based on a comparison result. In the mode, the vehicle is controlled to run along the preset test track by adopting automatic driving, so that the test control precision of the vehicle steering system can be improved, and the accuracy and consistency of off-line test are further improved.

The invention discloses a vehicle driving right transfer method based on abnormal driving behaviors and electronic equipment, and the method comprises the steps: collecting driver state data, vehicle driving state data and road environment data, and carrying out the multi-source data fusion; abnormal driving behavior detection is performed according to the multi-source data, and the driving risk is evaluated; a rule engine decision-making model based on the situation is used for deciding countermeasures to be taken, and the countermeasures comprise early warning type actions, auxiliary control type actions, forced take-over type actions and mutual switching among the actions based on an advanced auxiliary driving system or an automatic driving system; after the vehicle is forced to take over and runs for n minutes, whether the current vehicle is suitable for the driver to take over or not is judged according to driver state data collected in real time and the current comprehensive driving risk score, and whether driving right transfer is conducted or not is determined according to the judgment result. The accuracy of abnormal driving behavior detection is improved, the vehicle driving right is intelligently transferred, and the driving safety is improved.

The invention discloses an automatic driving safety operation system integrating environmental perception and decision reasoning. The system generates and dynamically updates a security risk map covering an operation area by fusing real-time environment perception, historical operation data and traffic management information. On the basis, the collaborative safety decision-making module further carries out behavior modeling and intention prediction on other traffic participants, and in combination with map risks and traffic instructions, a driving strategy is actively adjusted under a dynamic game framework. According to the invention, the safety control is improved from passive response to an active mode of behavior pre-judgment and game dominance, and the safety and reliability of the operating vehicle in a complex environment and the cooperative capability of the operating vehicle and traffic management are obviously enhanced.

The invention provides an automatic driving hazard scene detection method and system, electronic equipment and a vehicle, and relates to the technical field of automatic driving. According to the method, a multi-dimensional hazard scene detection rule system is established, so that accurate identification and response to key security events are realized. In the automatic driving operation process, the system continuously collects and compares a vehicle state signal with a preset triggering condition in real time, once any hazard scene condition is met, vehicle burying point data in a preset time window are collected and uploaded to a cloud end, and the vehicle burying point data are stored in the cloud end. Accident scene restoration, system risk assessment and fault root cause analysis are carried out based on the collected data, then a targeted safety repair scheme is generated, vehicle software is remotely upgraded in an OTA mode, closed-loop processing of safety problems and continuous optimization of the system are achieved, the safety, traceability and remote operation and maintenance capacity of the automatic driving system are effectively improved, and the safety and reliability of the automatic driving system are improved. And the running safety of the vehicle is ensured.

The invention relates to the technical field of automatic driving and artificial intelligence, in particular to a driving track planning method based on a visual language model and an intelligent driving system. The trajectory planning method comprises the steps of multi-modal data acquisition and preprocessing, visual language model reasoning, execution control and the like. The intelligent driving system comprises a multi-modal data acquisition and preprocessing module, a visual language model reasoning module and an execution control module. The visual language model reasoning module is integrated with a Patch selection module, a token compression module and a speculation decoding module; according to the invention, an end-to-end driving planning reasoning framework is constructed based on the visual language model, the vehicle trajectory planning is directly generated from the original multi-view perception data and the high-level driving intention by a single model, the architecture of an automatic driving planning system is simplified, and the response speed and robustness are improved. The method can effectively meet the requirement of the automatic driving vehicle for generating the safe path in real time in a vehicle-mounted embedded environment, and has important practical value.

The invention relates to the technical field of automatic driving vehicle control, in particular to a drive-by-wire chassis whole vehicle test device and method, which considers the delay compensation of a vehicle steering actuator and comprises an upper computer, a real-time controller, a drive-by-wire chassis vehicle, a combined navigation system and a USB-CAN (Universal Serial Bus-Controller Area Network) acquisition module. The method comprises the following steps that a vehicle in-loop system is built on a rack, and the delay characteristic of a steering actuator is identified through step and sine mixed excitation signals; establishing a vehicle trajectory tracking error model containing delay compensation based on an identification result; deploying a delay compensation model in a real vehicle environment, and verifying trajectory tracking performance; comparing the transverse error and the course angle error before and after compensation, and evaluating the compensation effect. The invention provides a drive-by-wire chassis whole vehicle test device and method integrating delay identification, model injection and real vehicle verification, and an engineering scheme is provided for dynamic rapid calibration and algorithm migration of an automatic driving drive-by-wire chassis actuator.

Embodiments relate to hazard detection in autonomous and semi-autonomous systems and applications. A transformer may use sampled image and LiDAR features to extract and decode a representation of one or more features of each point (e.g., refined height, range, driving condition, etc.) on a sampled surface (e.g., the road). These detections may be provided to one or more control components of an autonomous vehicle, which may use the detections to navigate, plan, or otherwise perform one or more operations. Some embodiments employ an automated approach to derive ground truth data from sensor data collected by data collection vehicle(s), such as data representing detected ground surface models, detected surface features, detected weather and / or surface condition labels, and / or detected per-point artifact labels. Accordingly, surface features such as ground surface heights along a predicted trajectory may be detected and ground truth data may be generated for a variety of sensing tasks.

The invention provides a heterogeneous vehicle cooperative control method based on the combination of a semantic potential field and an artificial potential field, and belongs to the technical field of automatic driving collaboration.The method comprises the steps that environment data are collected through a vehicle-mounted sensor of an automatic driving heterogeneous vehicle, and the environment data comprise the position and speed of the surrounding automatic driving heterogeneous vehicle and distribution of static obstacles; setting a vehicle type, a task type, a task priority and a real-time state of the vehicle, generating a corresponding semantic tag, and constructing a semantic potential field; constructing an artificial potential field, and calculating static obstacle repulsive force and gravitational force of a target position; generating a comprehensive potential field, and planning a driving path of the automatic driving heterogeneous vehicle; transmitting a key semantic tag by adopting an event triggering communication rule, and transmitting the semantic tag through a local cache and a priority queue when the communication is limited; through a distributed game theory, path conflicts among automatic driving heterogeneous vehicles are eliminated. According to the invention, the cooperative efficiency of the automatic driving heterogeneous vehicles is improved, and the path conflict of the vehicles is effectively eliminated.

The invention provides a cooperative control method and system based on a port automatic driving mixed driving scene, and relates to the technical field of port traffic management, and the method comprises the steps: firstly obtaining a dynamic interaction data set containing information of multiple aspects such as an automatic driving vehicle and manual driving equipment in a port operation region; and then carrying out association feature extraction on the dynamic interaction data set, generating dynamic association representation data containing traffic participant association relationships and other features, executing collaborative decision analysis based on the dynamic association representation data, and determining a right-of-way priority sequence and a collaborative path planning scheme. And a cooperative control instruction set including speed cooperative parameters and the like is generated according to a cooperative decision analysis result, and finally the cooperative control instruction set is distributed to a corresponding control system, so that dynamic cooperative driving control in a port mixed driving scene is realized, and the port operation efficiency and safety are improved.

The invention relates to the technical field of automatic driving and intelligent control, in particular to a multi-transportation-equipment cooperative motion control method based on multi-stage mixed learning and a storage medium, and the method comprises the steps: generating expert demonstration data through employing a single-vehicle motion control algorithm based on an expert rule, performing initialization training on the strategy network through online iterative supervised learning to obtain a pre-training strategy model; the multi-vehicle cooperative motion control method comprises the following steps: selecting a multi-vehicle cooperative motion model, loading parameters of the model into an Actor strategy network of multi-agent reinforcement learning, performing interactive training on a plurality of agents in a simulation environment by adopting a centralized training and decentralized execution normal form, and performing online iterative optimization on the strategy based on a composite reward function and generalized advantage estimation to obtain a multi-vehicle cooperative motion control strategy. According to the method, the complementary advantages of imitation learning and reinforcement learning are exerted, the training efficiency, the strategy performance and the collaborative operation capability and robustness of the system in a complex scene are improved, and the method can be directly applied to collaborative scheduling and control of transportation equipment groups in scenes such as surface mines and ports.

The invention provides a method and a device for controlling and testing formation driving of autonomous vehicles. The method comprises the following steps: acquiring current scene features based on a sensing device configured on a road side unit and / or an autonomous vehicle; the current scene feature is matched with a preset scene library, target formation control information is obtained, the preset scene library stores a mapping relation between multiple groups of scene features and formation control information, and the target formation control information comprises pilot vehicle control information and control information corresponding to at least one following vehicle; and based on the target formation control information, controlling the leading vehicle and the following vehicle to perform formation driving. According to the method and the device, the formation scene of the automatic driving vehicles in multiple scenes such as expressways is introduced, the problems of scattered test organization and limited scene coverage in the traditional road test are solved, multi-source heterogeneous sensing fusion dynamic correction can be constructed, and the formation cooperative control of the automatic driving vehicles in the high-speed environment is realized; the problem of perception degradation caused by failure of a single sensor is solved.

According to the automatic driving data transmission scheduling method and device, the equipment and the readable storage medium, dynamic priority weights are generated in a mode of combining reinforcement learning and a rule engine, and bandwidth demand changes, historical transmission performance and queue occupation states of all sensors can be adapted in real time; the problem that a low-priority sensor cannot acquire the bandwidth at a high load is effectively avoided, and meanwhile, efficient utilization of bandwidth resources can be realized through dynamic weight adjustment when a high-priority sensor is idle, so that the bandwidth utilization rate in a vehicle-mounted multi-sensor data transmission process is remarkably improved; the real-time performance and reliability of data transmission required by automatic driving decision making are guaranteed, and then the overall performance of an automatic driving system is improved.

The invention is suitable for the technical field of automatic driving, and provides an automatic driving vehicle blind area track prediction and planning method based on Transform, and the method comprises the steps: firstly constructing a key entrance point, calculating a shielding region, and then generating and maintaining a belief state of a ghost target in combination with a risk score; then, real and ghost targets are coded into Token in a unified mode, the Token is input into Transform for interactive modeling, and multi-modal track distribution is output; and finally, a variable anisotropic repulsive force potential field is constructed based on the predicted risk, and a safety control instruction is obtained through optimization solution and hard safety filtering. According to the method, continuous probabilistic estimation and interactive reasoning of the potential risk of the blind area are realized, a flexible defensive track can be generated according to the risk confidence and the dynamic state of the vehicle, and the traffic efficiency is guaranteed while the safety is improved.

The invention belongs to the technical field of automatic driving test, and provides an automatic driving confrontation test scene generation method and system based on strategy switching, and the method comprises the steps: obtaining an automatic driving simulation scene; setting positions and initial speeds of a target vehicle and an intelligent body vehicle in the acquired simulation scene; in combination with the set position and initial speed of the target vehicle, the running state of the target vehicle is optimized by adopting a multi-agent near-end strategy, and the target vehicle is controlled; determining a current traffic state according to the control driving state of the target vehicle; in combination with the dynamic quantitative index of the artificial risk potential field, sensing the traffic flow risk of the intelligent body vehicle in the determined current traffic state in real time; and according to the obtained traffic flow risk level, automatically switching a defense strategy and an attack strategy of the agent vehicle, and driving the multi-agent vehicle to cooperatively generate an automatic driving confrontation test scene.

The invention relates to an end-to-end automatic driving system based on multi-modal fusion, and the system comprises an environment sensing module which comprises a plurality of types of sensors and is used for collecting environment information of different modals; the feature extraction module is used for carrying out feature extraction on the environment information of different modes and converting the environment information into aerial view space to obtain multi-mode BEV features; the feature fusion module is used for fusing the multi-modal BEV features to obtain fused features; the behavior planning module is used for outputting a future waypoint of the target vehicle and a prediction result of a control instruction according to the fusion features by adopting a trained behavior planning network; and the output fusion module is used for carrying out track prediction and multi-step control prediction and selecting an optimal driving behavior based on a situation strategy. Compared with the prior art, the method has the advantages that the environment sensing precision and decision generalization ability in a complex scene are effectively improved, error accumulation between modules is reduced, and a more efficient solution is provided for end-to-end automatic driving.