125 results about "Traffic scene" patented technology

The application discloses a kind of hardware-in-the-loop test of automatic driving domain controller, comprising: real-time processor and graphics workstation;Graphics workstation is used to generate simulation traffic scene information, simulation sensor information is generated based on simulation traffic scene information, first simulation sensor information in simulation traffic scene information and simulation sensor information, simulation traffic scene information and simulation sensor information are transmitted to automatic driving domain controller, second simulation sensor information in simulation sensor information is transmitted to real-time processor;Real-time processor is used to transmit second simulation sensor information to automatic driving domain controller, receives the vehicle control information that automatic driving domain controller feedback.This application provides the hardware-in-the-loop test system, device, storage medium of automatic driving domain controller, can avoid the problem that scene information and sensor perception information do not match when hardware-in-the-loop test, improves the accuracy of test.

The utility model discloses a magnetic attraction piece and a magnetic attraction assembled three-dimensional track. The magnetic attraction assembled three-dimensional track comprises the magnetic attraction piece, a track piece and a toy car. The magnetic attraction piece is provided with a middle through opening, the middle through opening is provided with a rail butt joint groove, and inserting holes used for being connected with a vehicle rail piece are formed in the four corners of the magnetic attraction piece. The magnetic attraction splicing three-dimensional track further comprises a track piece and a toy car, inserting columns are arranged at the two ends of the track piece, the inserting columns are used for being connected with the inserting holes of the magnetic attraction pieces in an inserted mode, and guide rail parts corresponding to the track butt joint grooves of the magnetic attraction pieces are arranged at the track piece. By adopting the structural design of combining the magnetic attraction pieces and the vehicle rail pieces, not only can a richer, more real and layered urban traffic scene be built, but also the hand-eye coordination and logical thinking ability of children can be exercised in the splicing process, the imagination and creativity of the children can be fully expanded, the openness and playability of the toy can be improved, and the toy is worthy of popularization and application. The construction requirements of children of different age groups are met.

The application provides a traffic event video slice setting method, device, equipment and product, the method comprises the following steps: obtaining a target traffic scene image, and constructing a three-dimensional scene in the target traffic scene image to obtain a three-dimensional traffic scene image; setting a foreground material and a virtual material in the three-dimensional traffic scene image, and setting an initial motion state corresponding to the foreground material and the virtual material respectively to obtain an initial traffic scene image; the foreground material is a material obtained by cutting out from an original image; performing simulation and running based on the initial traffic scene image, and determining a traffic event video slice according to a traffic event detected in the simulation and running process. The method can efficiently and conveniently determine the initial traffic scene image, perform simulation and running based on the initial traffic scene image, and determine the traffic event video slice according to the traffic event detected in the simulation and running process, so that the efficiency of setting the traffic event video slice can be improved.

The application discloses a kind of attention mechanism trajectory generation and simulation methods for sparse data completion, belong to intelligent transportation and trajectory prediction field.The application fills in trajectory missing value using high-precision sensor and map information, combines graph attention network (GAT) and multi-modal fusion technology;Adopt the attention module based on distance (D-GAT) and based on view (V-GAT), capture the interaction between vehicles, improve the understanding of complex traffic scene;Through prediction supervision generator and multi-modal trajectory generator, combine LSTM and Gaussian mixture model (GMM) to generate multiple possible trajectories, and use Kalman filter for online adjustment, ensure the accuracy and real-time of trajectory.The application realizes the intelligent completion of sparse traffic data and the accurate generation of trajectory, provides reliable data support and decision basis for intelligent transportation system, helps the efficient operation and sustainable development of urban traffic planning and management.

Embodiments of this disclosure relate to methods, apparatus, devices, storage media, and program products for generating maps. The method includes: providing environmental information of a traffic scene to an element recognition model; determining a first set of points associated with target map elements in the traffic scene and evaluation information of the first set of points, the evaluation information indicating the quality of each point in the first set of points; filtering at least one point from the first set of points based on the evaluation information to determine a second set of points, wherein the quality of at least one point is below a threshold; and generating a map representation of the target map elements based on the second set of points, the map representation indicating at least the category and shape of the target map elements. In this manner, embodiments of this disclosure can improve the accuracy of the generated maps.

The present application relates to a kind of layered reinforcement learning driven car-road cooperation automatic driving longitudinal following control method and system, belong to the field of automatic driving vehicle control.For the problem that the generalization ability of existing following control method is insufficient, strategy update is easy to lose stability, easy to fall into local optimum, adopt layered architecture, upper layer adapts intersection, high-speed straight road and ramp multiple typical traffic scenes, based on real-time car-road cooperation communication information dynamic planning follow target, after normalizing processing to multi-source heterogeneous traffic characteristics, generate environment state vector;Lower layer uses proximal policy optimization algorithm, outputs continuous accelerator or brake command through actor-critic double network, and updates network parameters by combining strategy entropy with pruning loss with Jacobian correction.The method improves the cross-scene generalization ability and perception robustness, effectively guarantees the stability and monotonicity of strategy update, avoids instruction step mutation, breaks the exploration deadlock, and balances driving safety, traffic efficiency and driving comfort.

This invention provides a method and system for constructing an interactive predictive risk field based on dynamic environmental characteristics. The method includes: constructing an obstacle field based on the attributes and predicted behavior of surrounding vehicles; constructing a road field based on road conditions; configuring driving factors based on the driver's own characteristics; and integrating and interacting the obstacle field, road field, and driving factors to construct an interactive predictive risk field. This invention constructs corresponding risk field models for each potential factor in the proposed traffic scenario, including an obstacle field caused by traffic participants, a road field caused by lane lines and boundaries, and driver factors determined by the driver's own characteristics. By incorporating trajectory prediction and relative motion calculation of surrounding vehicles, the model can perceive impending risks earlier, demonstrating the potential of this method for subsequent application in decision-making and planning.

The application discloses a kind of behavior modeling method and system based on generation confrontation imitation and interactive representation, wherein the method comprises: collecting the multi-modal driving and perception data of each traffic subject, and processing the data to obtain state input data in the form of uniform input sequence;State input data is input into behavior prediction model to obtain predicted driver driving behavior.The application establishes driver behavior prediction model, outputs implicit scene coding that fuses the historical motion state of ego vehicle, candidate trajectory and neighbor vehicle interaction information, converts scene information into output vector suitable for imitation learning, can highly summarize current traffic scene from decision-related perspective, thereby significantly reducing state space dimension and improving modeling stability;And implicit scene coding is input into generation confrontation imitation structure, optimized by minimizing the maximum value of sample discrimination loss function of discriminator network, improve the accuracy and reliability of driver behavior prediction.

The application discloses an abnormal driving behavior detection method based on an improved residual double-layer graph attention network (Res-DBiGATv2), and belongs to the technical field of Internet of Vehicles. Firstly, the vehicle trajectory data is constructed into a space-time dynamic graph sequence according to time steps. Then, a ResBiGATv2 module is designed, and spatial feature aggregation is realized through double-layer graph attention, a multi-head mechanism and residual connection. Then, a ContraNorm contrast normalization layer is used to enhance feature uniformity and inhibit dimension collapse and oversmoothing. A graph external attention enhancement module GEA is introduced, and a learnable external memory unit is used to inject global information. Finally, a GRU is used to model time sequence features, and an abnormal node is identified through reconstruction error in an encoder-decoder framework. The application can accurately and timely detect various abnormal driving behaviors such as slow driving, overspeeding, following, and stagnation in complex traffic scenes, and significantly improves detection accuracy and robustness.

The application discloses a kind of dangerous event chain extraction method and system based on complex traffic scene, including obtaining test scene data;Based on the key factors of dynamic attribute and interaction between vehicle and environment determined by preset natural driving data and sensor performance information, and the uncertainty of key factors is quantified;Based on the risk estimation multidimensional feature set and uncertainty information preset constructs driving risk estimation model for test scene;The driving risk estimation model is optimized, and the optimized risk estimation model for test is obtained;Based on the time evolution characteristics of dangerous event, a dangerous time chain model is constructed according to the optimized risk estimation model;Dangerous time chain model is solved, and dangerous event chain is obtained.The application uses traffic accident data and traffic conflict data to reveal the space-time evolution law of complex traffic scene, and improves the effect of later test scene reproduction and reconstruction by solving the dangerous time chain model.

The application relates to the technical field of vehicles, in particular to a long-time-domain driving behavior decision method suitable for high-speed and loop traffic scenes, which comprises the following steps: acquiring a global planning path of a vehicle, a current motion state of the vehicle and current motion states of all surrounding vehicles in a region; generating an optimal driving behavior sequence of the vehicle according to the global planning path, the current motion state of the vehicle and the current motion states of all surrounding vehicles in the region; planning a driving track of the vehicle based on a first driving behavior of the optimal driving behavior sequence, and regenerating the optimal driving behavior sequence after the vehicle executes the first driving behavior based on the driving track, until the global planning path is completed. Therefore, the embodiment of the application can realize multi-step decision of driving behaviors such as lane-changing gap selection and lane-changing overtaking, has foresight, and simultaneously gives an optimal solution to long-time-domain driving behavior sequence planning based on feasibility discrimination, considers safety and high efficiency, and meets the needs of general driving.

The application discloses a kind of 4D imaging millimeter wave radar stereo traffic dataset construction method and related equipment, comprising: obtaining the multi-source original data collected by data acquisition platform, which is equipped with 4D imaging millimeter wave radar, laser radar, high-definition camera and positioning system etc., and the multi-source original data contains the above-mentioned each sensor data and is collected in the target scene of preset scene library, each sensor data is aligned under unified space-time reference after time synchronization and space calibration;Multi-source original data is analyzed and processed, point cloud is obtained and quality verification is carried out;4D imaging millimeter wave radar point cloud and laser radar point cloud are fused, visual reference is provided according to image data, motion compensation is provided by positioning data, and multi-modal sensing data is formed;Multi-modal sensing data is labeled, and dataset is obtained.The application constructs multi-modal high-quality dataset with 4D imaging millimeter wave radar as core, cooperates multiple sensors, covers multiple stereo traffic scenes, and provides high-precision labeled information with time sequence continuity.

The application relates to a heterogeneous agent trajectory prediction method, device and computer equipment under a complex mixed traffic scene. The method comprises the following steps: acquiring comprehensive trajectory information of a self-vehicle, historical trajectory information of a heterogeneous agent, historical trajectory information of a nearby vehicle and map information where the self-vehicle is located, and identifying a trajectory coding tensor and a time interaction tensor based on the comprehensive trajectory information of the self-vehicle, the historical trajectory information of the heterogeneous agent and the historical trajectory information of the nearby vehicle; coding the map information to obtain a map semantic coding tensor, and constructing a directed spatio-temporal graph containing each agent based on the trajectory coding tensor; constructing a context fusion tensor based on the directed spatio-temporal graph, the trajectory coding tensor, the time interaction tensor and the map semantic coding tensor, and decoding and analyzing the context fusion tensor to obtain a trajectory prediction result of the heterogeneous agent. The method can improve the accuracy of agent trajectory prediction under a complex mixed traffic scene.

This invention discloses an intelligent diagnostic system and method for intersection operation status based on feature matching. Specifically, it involves: collecting peak and off-peak video data from an unmanned aerial vehicle (UAV) above the intersection; detecting dynamic parameters such as traffic flow and speed of motor vehicles and non-motor vehicles at the intersection, as well as static parameters such as the coordinates of lane markings and guide arrows inside and at the intersection's entrances; constructing a multi-dimensional fusion indicator system; establishing a database of intersection influencing factors and a database of improvement measures; simulating and extrapolating data obtained from static modeling and dynamic perception using digital twin technology to acquire data features of different traffic scenarios from the intersection influencing factor database; and achieving a one-to-one strong rule association between the influencing factor database and the improvement measure database through multi-dimensional Apriori association rules. This invention effectively improves the low accuracy and efficiency of existing non-motor vehicle detection methods and provides adaptive diagnostic evaluation for intersections of different specifications.

This invention discloses a visibility estimation method and apparatus based on binocular vision. The method includes: simultaneously acquiring left and right views using a calibrated binocular camera; inputting the image pairs into a stereo matching deep neural network to obtain a disparity map and converting it into a depth map; inputting the left view and depth map into a dual-branch deep convolutional neural network to extract multi-scale features; fusing RGB features and depth features through a cross-modal feature fusion module; and finally outputting a visibility estimate through a feature aggregation and regression module. The apparatus includes a binocular image acquisition unit, a data processing and visibility estimation unit, and a result output unit. This invention solves the depth ambiguity problem of monocular vision by fusing binocular depth information and image appearance information, achieving high-precision, non-contact visibility surface measurement. It has the advantages of low cost and flexible deployment, and is suitable for visibility monitoring in traffic scenarios such as highways.

This invention discloses a method for drone swarm inspection under budget constraints in large-scale traffic scenarios, belonging to the field of intelligent transportation and unmanned system collaborative control technology. The method first models the traffic environment as a discrete latent information field using information field theory to complete the initial embedding of drones and inspection nodes. Then, it utilizes the multi-directional attention mechanism of graph neural networks to achieve environmental feature fusion and generate an initial node-drone allocation strategy. Subsequently, it satisfies the endurance budget constraint through dual-criteria heuristic pruning and recovers high-value unvisited nodes using field gradient re-insertion. Finally, it outputs an optimized inspection path and completes model reinforcement learning training and cross-scenario transfer. This invention also discloses the corresponding inspection system, storage medium, and computer equipment. This invention can generate highly collaborative and high-coverage inspection paths in seconds under strict endurance constraints, solving the problems of slow solution, low coverage, poor collaboration, and weak generalization ability of traditional algorithms.

The application discloses a novel traffic scene target detection evaluation method and system, and relates to the technical field of target detection model performance evaluation. The application obtains the prediction result of a target detection model, determines true positive TP or false positive FP based on the matching relationship between the prediction frame and the real label, and respectively calculates the position weight factor and the scale weight factor. Then, the application obtains PSTP / PSFP by weighting TP / FP, constructs a position-scale weighted PR curve, and integrates to obtain the position-scale weighted average precision PSAP. The application can significantly improve the evaluation sensitivity of edge regions and small-scale high-risk targets, support scene adaptation and security level linkage, and make the evaluation results more suitable for the actual safety needs of automatic driving.

The present application belongs to the technical field of automatic driving test, and in particular to a test scene credible acceleration generation method based on multi-simulator consistency verification. First, the traffic scene to be tested is parameterized, and a simulator with different dynamics, sensors and behavior models is selected. The same scene parameters are synchronously mapped into each simulator for execution, and the results such as the host vehicle response, collision state, minimum distance, etc. are collected. Then, the consistency deviation of the multi-simulator in risk scoring, failure judgment and trajectory evolution is calculated, a credible risk benefit index is constructed, and credible high-risk scenes are selected. According to the index distribution sampling weight, the acceleration generation distribution is updated, and the new scene is ensured to be compliant through constraint projection. Finally, the credible failure risk of the automatic driving system is estimated by using the reweighting relationship between the natural scene and the generated distribution and combining the consistency results. The present application can effectively eliminate the pseudo-high-risk scenes caused by the deviation of a single simulator, and improve the credibility and stability of the test scene generation.

The application discloses a kind of vehicle-mounted millimeter wave radar error correction methods for complex traffic scene, it is related to radar data processing technical field, and stable weak scattering point set Sca is extracted by candidate scattering point set Csp, and the weak point trace in the sheltered area is not ignored by the aid of boundary corresponding set Bdr, so that track keeps continuity and credibility on time series and spatial coordinates;Combined with the track compensation set Com generated by time synchronization error Ets and calibration residual Ecl and applied to track fitting set Fit, the corrected track set Cor obtained can effectively eliminate the drift and misplacement caused by radar sampling delay and installation angle deviation, to correct track set Cor as output result is used for scene reconstruction and target identification, so that vehicles can still obtain time alignment and spatially accurate error correction final track results in complex traffic environment such as night multi-lane highway, rain and fog weather intersection.

The application provides a traffic energy source and load prediction and collaborative optimization method and system based on data fusion, relates to the technical field of traffic energy management, and first accesses multiple-source traffic energy data to generate a traffic energy data correlation pool; generates a dynamic source and load correlation network based on the traffic energy data correlation pool; maps traffic scene features to the dynamic source and load correlation network to generate a multi-dimensional fusion deduction input set; performs source and load change trend deduction on the multi-dimensional fusion deduction input set through a multi-dimensional fusion deduction model adapted to the scene to obtain a traffic energy source and load prediction result; and forms a collaborative optimization scheme based on the traffic energy source and load prediction result and dispatching requirements and feeds back the collaborative optimization scheme to network optimization scene adaptation weights, thereby effectively improving the operation efficiency and reliability of the traffic energy system and reducing energy waste and operation costs.

Systems and methods for language-conditioned trajectory diffusion for understanding complex traffic scenes. Complex multi-modality scene context information that includes map information and agent information for agents in input videos can be captured with a language-conditioned trajectory diffusion simulation (LDTS) model. Spatiotemporal scene information can be extracted based on semantic information from text instructions with the LDTS model. The map information, agent information, and semantic information can be fused using a cross-attention fusion module of the LDTS model into text-conditioned encodings. Language-conditioned trajectories can be generated based on the text-conditioned encodings with the LDTS for performing downstream tasks.

The application discloses an automatic driving safety early warning dynamic twin closed loop parallel simulation deduction method and system, the method comprises the following steps: acquiring multi-source data in a real traffic scene, constructing a dynamic twin simulation environment evolving synchronously with the real scene; generating multiple future scene evolution branches based on the current scene state and trajectory prediction information, and performing super real-time parallel deduction to obtain scene deduction results at multiple future moments; performing consistency determination on the deduction results and real observation results, and dynamically cutting branches that do not meet the consistency condition; automatically cutting branches with low threat degree based on a threat degree index; the vehicle-mounted automatic driving control system interacts with the dynamic twin simulation environment to form a virtual-real closed loop feedback; and outputting safety early warning, control suggestions and safety decisions based on the cut deduction results. The application realizes super real-time deduction, dynamic correction and active early warning of potential risks in complex traffic scenes, and improves the real-time performance, accuracy and reliability of automatic driving safety early warning.

This invention belongs to the field of intelligent transportation and autonomous driving technology, and discloses a dynamic decision-making method and system for unprotected left turns of unmanned vehicles based on neuromorphic neural networks. The method constructs a traffic scene graph and introduces a neuromorphic neural network for prefrontal cortex-like collaborative spatial reasoning to generate scene-level semantic representations. Based on this, a structured left-turn action graph is constructed, and feasibility and risk assessments are performed on candidate driving behaviors. An action mask is generated to dynamically compress the decision space. Within the framework of a proximal policy optimization algorithm, a risk budget mechanism is combined to achieve an adaptive balance between safety and efficiency. Simultaneously, a scene graph memory is constructed, and historically successful strategies are reused through similarity retrieval to improve cross-scene generalization ability. This invention effectively solves the problems of excessively large decision space, unstable behavior, difficulty in balancing safety and efficiency, and insufficient generalization ability of existing unmanned vehicles in complex intersection scenarios, significantly improving the stability, safety, and adaptability of decision-making.

This application discloses a method and apparatus for reconstructing the process of a perceived traffic incident, addressing the problem of lacking technical means to obtain the on-site conditions at the time of the incident in traffic incident judgment. The method involves acquiring the perception information of traffic participants during the incident; generating twin data of the traffic scene at the incident location; the twin data including graphic objects of the traffic participants; determining the location information and corresponding time information of the vehicles involved in the incident based on the perception information; and displaying the driving process of the vehicles involved in the incident in the playback image of the twin data. This application's embodiments fully utilize existing equipment, avoiding redundant construction, and integrate static and dynamic data to generate more comprehensive and accurate digital information, contributing to improved accuracy in traffic management and monitoring. It provides traffic management departments with richer data support, helping them make more scientific and rational management decisions.

This invention provides an intelligent event flow control method and system applied to traffic scenarios, belonging to the field of traffic management technology. First, it receives comprehensive perception of traffic events and dynamic correlation information of the traffic network to generate traffic event flow baseline information. Based on the traffic event flow baseline information, it divides the flow into stages and reconstructs the traffic event flow links. According to the traffic event flow stage division and link reconstruction information, it constructs a distributed collaborative response rule set for control resources to generate resource collaborative response configuration information. Following the resource collaborative response configuration information, it drives a distributed control node group to execute collaborative control actions, generating collaborative execution information. Based on feedback traffic event evolution correction data and traffic network operation optimization data, iteratively updates the traffic event flow link reconstruction information and resource collaborative response configuration information to generate traffic event flow control completion information. This invention provides comprehensive perception and precise control, achieving efficient handling of traffic events and improving road network operation efficiency and safety.

The application discloses a kind of intelligent vehicle active obstacle avoidance control methods based on visual perception, to improve the driving safety of vehicle in complex traffic scene.The application relates to the field of intelligent driving.The application includes upper and lower two layers, the upper layer is environment perception module, the lower layer includes risk assessment module, path planning module and pure tracking algorithm.The upper layer environment perception module includes RRW-YOLOv11n, multi-scale features are extracted through RGCEBlock module, feature fusion is enhanced by Re-calibration FPN structure, the boundary box regression accuracy is optimized using WPIoU loss function, and finally the perception information is outputted;The lower layer risk assessment module carries out multi-source risk assessment to obtain multi-source risk probability, and the posterior risk probability is calculated after Bayes risk fusion, and the path is planned out by path planning module;Path tracking control is carried out by pure tracking algorithm, to realize the safe driving of vehicle in complex traffic scene.

The present application relates to a kind of automatic driving vehicle collision risk assessment method and system based on visual large language model, the method includes the following steps: step 1, utilize structure theme model to model collision accident narrative data set;Step 2, use XGBoost model and Shapley explanation to establish the relationship between theme and collision accident severity;Step 3, establish the visual big model training data set for collision risk assessment;Step 4, based on visual large language model, the fine-tuning training of visual large language model is carried out for collision risk assessment;Step 5, build joint simulation platform and carry out risk assessment simulation verification.The method improves the perception understanding ability of automatic driving vehicle to traffic scene, accurately analyzes the potential collision threat of current driving under complex traffic environment, realizes the assessment of collision risk in different scenarios, effectively improves the accuracy and practical value of collision risk assessment, and improves the safety of automatic driving car driving.

The application provides an automatic driving chassis control method based on a world model and a network slice closed loop, and belongs to the field of intelligent driving, and comprises the following steps: acquiring multi-modal data in a traffic scene through perception, encoding the multi-modal data to obtain a hidden state, performing future state prediction and control generation on the hidden state through a world model to obtain a future state and a control instruction, optimizing the world model according to the future state, constructing a digital twin scene, receiving and controlling the automatic driving chassis according to the control instruction in the digital twin scene, and acquiring multi-modal data through perception, transmitting corresponding data flow of the perceived, predicted future state and control instruction to a network slice resource, and dynamically adjusting and configuring the network slice resource, and online optimizing the world model and the network slice resource according to a feedback mechanism to realize multi-step high-precision environment prediction and chassis dynamic response while meeting end-to-end low latency.