Vehicle track prediction and planning error control and collision avoidance method and device

Abstract

The invention belongs to the technical field of intelligent network connection automobile safety control, and relates to a vehicle track prediction and planning error control and collision avoidance method. Collecting a true value track (position and course) of the vehicle; a space deformation model is used for setting distance and course deviation; a time evolution model is used to describe the variation of the offset at any time; 4, offsetting offset statistics by using a probability distribution model; generating a customized trajectory through the error mixing model, and injecting the customized trajectory into an open source simulator SUMO to form an error-controllable real-time trajectory; a plurality of collision early warning strategies are adopted to evaluate collision detection performance indexes under different prediction or planning precisions; and continuously outputting an evaluation report for iterative optimization and security upgrade of the algorithm. According to the invention, the error form and amplitude can be accurately controlled, the multi-scene risk can be reproduced, the collision avoidance capability of the automatic driving system can be objectively quantified, and the accuracy, repeatability and completeness of safety assessment can be improved.

Description

Technical Field

[0001] This invention relates to the field of autonomous driving and cooperative collision avoidance technology, and in particular to a vehicle trajectory prediction and planning error control and collision avoidance method and device. Background Technology

[0002] Safety and comfort are the core cornerstones of a sustainable transportation society. However, 75% to 90% of traffic accidents involve human error, making autonomous vehicles (AVs) a key solution to this problem. But AVs still face issues such as physical obstruction and sensor malfunctions, leading to frequent accidents. Related data shows that the accident rates of Level 3-5 Automated Driving Systems (ADS) and Level 2 Advanced Driver Assistance Systems (ADAS) remain high, and their safety is far from ideal. To reduce collision risk, AVs need to accurately predict the trajectories of surrounding road users to improve situational awareness over time. While significant progress has been made in vehicle trajectory prediction (VTP) with the help of deep learning, state-of-the-art methods offer limited improvement in accuracy, and the models are becoming increasingly large and complex, compromising real-time performance and generalization capabilities. Some methods also perform poorly in scenarios outside of benchmark datasets. More importantly, existing research focuses primarily on improving VTP accuracy while neglecting the impact of prediction accuracy on the effectiveness of downstream tasks (such as planning and collision avoidance). Human drivers, despite their inaccurate predictions, can effectively control their vehicles in scenarios with varying traffic densities, prompting reflection on whether high-precision trajectory prediction can truly improve the effectiveness of downstream tasks. Furthermore, accidents at and near intersections account for over 40% of all accidents. Traffic interactions are complex and visibility is easily obstructed in these areas. With the integration of vehicle-to-everything (V2X) communication technology into connected vehicles (CAVs), cooperative collision avoidance systems (CCAS) have become increasingly important. However, their performance depends on trajectory prediction and risk estimation. The impact of VTP accuracy upper bound on CCAS performance is still unclear. The main reason for this is the lack of trajectory prediction models that can simulate custom error distributions in current research (or products). This highlights the necessity of studying the relationship between VTP accuracy and key performance indicators of CCAS.

[0003] The closest existing technology can be found in "A Novel Traffic Simulation Framework for Testing Autonomous Vehicles Using SUMO and CARLA". This framework combines SUMO and CARLA to comprehensively test the perception, planning, and control of autonomous vehicles under real road networks and naturalized traffic flows. By generating microscopic traffic scenarios in SUMO and then providing high-fidelity perception data in CARLA, it realizes an end-to-end closed-loop evaluation platform from traffic flow modeling to sensor simulation.

[0004] However, this framework focuses on providing verification in complex environments and through multi-source sensor fusion, but it does not address the controllable injection of "trajectory prediction / planning errors": firstly, it lacks the ability to directionally shift the trajectory based on a spatial deformation model; secondly, it does not provide a mechanism to control the magnitude of errors over time and through probability distribution; and thirdly, it does not inject errors into SUMO in real time to evaluate the collision warning effect under different error levels. Therefore, it is difficult to meet the needs for "customized trajectories with fully controllable errors" and quantitative evaluation of the performance of multi-strategy collision warning systems. Summary of the Invention

[0005] To address the aforementioned problems in related technologies, embodiments of the present invention provide a method and device for vehicle trajectory prediction, planning error control, and collision avoidance.

[0006] In a first aspect, embodiments of the present invention provide a vehicle trajectory prediction and planning error control and collision avoidance method, comprising: acquiring true (or reference true, such as lane centerline) trajectory data of a vehicle, wherein the true trajectory data includes position coordinates and heading angles at various times; determining the spatial offset of the generated trajectory relative to the true trajectory data according to a spatial deformation model (SDM), wherein the spatial offset includes distance offset and heading offset, specifically including: front left offset, rear left offset, front right offset, and rear right offset; determining the changes of distance offset and heading offset over time according to a time evolution model (TEM); determining the probability distribution of the final distance offset and the probability distribution of the final heading offset according to a probability distribution model (PDM); fusing the spatial offset model, the time evolution model, and the probability distribution model through an error hybridization model (EMM) to generate a trajectory that conforms to customized error characteristics; integrating the trajectory into the open-source traffic simulator SUMO to construct a real-time error injection framework and generate a customized trajectory with fully controllable error; and employing different collision warning strategies to verify and quantify the key performance indicators of the collision detection system under different prediction / planning accuracies.

[0007] Based on the above method embodiments, the vehicle trajectory prediction and planning error control and collision avoidance method provided in this embodiment of the invention, wherein determining the spatial offset of the true trajectory according to the Spatial Deformation Model (SDM) includes:

[0008] Front left offset:

[0009] Left offset:

[0010] Front right offset:

[0011] Back to right offset:

[0012] in, For t i Left offset x-coordinate before time step; For t i The true value of the x-axis at time; For t i Time position offset vector; For t i Heading deviation at any moment; For t i True value of the course at any given moment; For t i Left offset of the ordinate before the time step; For t i True value of the ordinate at time; For t i Left offset x-coordinate after time step; For t i The ordinate shifts to the left after time step; For t i The x-coordinate shifted to the right before the given time; For t i The rightward offset of the ordinate before the time step; For t i The x-coordinate shifts to the right after time step; For t i The ordinate is shifted to the right after time step; sin represents the sine function; cos represents the cosine function.

[0013] Based on the above method embodiments, the vehicle trajectory prediction and planning error control and collision avoidance method provided in this embodiment of the invention, wherein determining the changes in distance offset and heading offset over time according to the time evolution model TEM includes:

[0014]

[0015] Where E is the distance offset error explosion parameter; For the final distance offset; t f is the duration of vehicle trajectory prediction / planning; log is the sign of the logarithmic function; G is the sublinear growth parameter of the heading deviation error; This is the final heading offset.

[0016] Based on the above method embodiments, the vehicle trajectory prediction and planning error control and collision avoidance method provided in this embodiment of the invention, wherein determining the probability distribution of the final distance offset and the probability distribution of the final heading offset according to the probability distribution model (PDM) includes:

[0017]

[0018] in, This represents the probability distribution of the final course deviation; π is the mathematical constant pi. σt represents the variance of the final heading deviation error; exp is an exponential function with base to the natural constant; μ α This represents the average of the final heading deviation error; The variance of the distance offset error; μ X The mean of the final distance offset that conforms to a normal distribution; This represents the final distance offset probability distribution.

[0019] Based on the above method embodiments, the vehicle trajectory prediction and planning error control and collision avoidance method provided in this embodiment of the invention, wherein the spatial offset model, temporal evolution model, and probability distribution model are fused through an error hybrid model (EMM) to generate a trajectory that conforms to customized error characteristics, includes: based on the expected final distance offset mean μ D Given parameter k, calculate the mean μ of the final distance offset that corresponds to a normal distribution. X =μ D / M(k) and standard deviation σ X =kμ X Where M(k) is the ratio of the mean of the folded normal distribution to the mean of the corresponding normal distribution; from the final distance offset normal distribution Mid-sampling final distance offset Normal distribution of final course deviation Mid-sample final heading offset Using the aforementioned time evolution model TEM, the time range [0, t] is calculated. f The distance offset sequence and heading offset sequence within the range are used; based on the distance offset sequence and heading offset sequence, the spatial deformation model (SDM) is used to generate a trajectory that conforms to the customized error characteristics.

[0020] Based on the above method embodiments, the vehicle trajectory prediction and planning error control and collision avoidance method provided in this embodiment integrates the trajectory into the open-source traffic simulator SUMO, constructs a real-time error injection framework, and generates a customized trajectory with fully controllable error. This includes: generating a mixed traffic scenario and recording all floating car data (FCD) to obtain the real trajectory; reproducing the mixed traffic scenario and, based on the true trajectory, loading the trajectory into the real-time error injection framework containing the open-source traffic simulator SUMO; wherein, the speed of the autonomous vehicle (CAV) is randomly sampled from a uniform distribution U (10 km / h, 30 km / h), and the human-driven vehicle (HDV) is controlled by the extended intelligent driver model (EIDM).

[0021] Based on the above method embodiments, the vehicle trajectory prediction and planning error control and collision avoidance method provided in this invention adopts different collision warning strategies and verifies and quantifies the key performance indicators of the collision detection system under different VTP accuracies, including:

[0022] The number of warnings N within the observation window is calculated based on the following general collision warning strategy GCWS. w :

[0023]

[0024] Density-based collision warning strategy (DCWS):

[0025] w o +w r =t f

[0026] in, For t i The x-coordinate of vehicle A at that time; For t i The ordinate of vehicle A at that time; For t i The x-coordinate of vehicle B at that time; For t i The ordinate of vehicle B at time d; thd This is the spatial distance threshold; ∈ is the existence symbol; ∈ is the membership symbol; TRA A For the trajectory of vehicle A; TRA B For vehicle B's trajectory; N w The number of warnings observed in the window; Δt is the observation interval; w o For the observation time window; ρ thd The warning density threshold; w r Reserve a time window; || ||2 is the 2-norm symbol.

[0027] Secondly, embodiments of the present invention provide a vehicle trajectory prediction and planning error control and collision avoidance device, comprising: a first main module for acquiring true trajectory data of a vehicle, the true trajectory data including position coordinates and heading angles at various times; a second main module for determining the spatial offset of the true trajectory data according to a spatial deformation model (SDM), the spatial offset including distance offset and heading offset, specifically including: forward left offset, rear left offset, forward right offset, and rear right offset; a third main module for determining the changes in distance offset and heading offset over time according to a time evolution model (TEM); The fourth main module is used to determine the probability distribution of the final distance offset and the probability distribution of the final heading offset based on the probability distribution model (PDM). The fifth main module is used to generate a trajectory that conforms to customized error characteristics by fusing the spatial offset model, temporal evolution model, and probability distribution model through the error hybridization model (EMM). The sixth main module is used to integrate the trajectory into the open-source traffic simulator SUMO, build a real-time error injection framework, and generate a customized trajectory with fully controllable error. The seventh main module is used to verify and quantify the key performance indicators of the collision detection system under different VTP accuracies by adopting different collision warning strategies.

[0028] Thirdly, embodiments of the present invention provide an electronic device, comprising:

[0029] At least one processor, at least one memory, and a communication interface; wherein,

[0030] The processor, memory, and communication interface communicate with each other;

[0031] The memory stores program instructions that can be executed by the processor. The processor calls the program instructions to execute the vehicle trajectory prediction and planning error control and collision avoidance method provided by any of the various implementations of the first aspect.

[0032] Fourthly, embodiments of the present invention provide a non-transitory computer-readable storage medium storing computer instructions that cause a computer to execute a vehicle trajectory prediction and planning error control and collision avoidance method provided by any of the various implementations of the first aspect.

[0033] The vehicle trajectory prediction and planning error control and collision avoidance method and device provided in this invention adopts a general error model (GEM) and an error injection framework to generate customized error trajectories, providing a tool for studying the relationship between vehicle trajectory prediction (VTP) accuracy and cooperative collision avoidance system (CCAS) performance. Experiments clearly show that VTP accuracy is negatively correlated with CCAS recall and positively correlated with precision. The proposed density-based warning strategy (DCWS) improves accuracy at various VTP accuracy levels. Furthermore, the quantitative analysis of the relationship between VTP accuracy and key CCAS indicators helps evaluate the impact of the VTP model on CCAS, further ensuring traffic safety. Attached Figure Description

[0034] To more clearly illustrate the technical solutions in the embodiments of the present invention or related technologies, the drawings used in the description of the embodiments or related technologies will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0035] Figure 1 This is a schematic diagram of the vehicle trajectory prediction and planning error control and collision avoidance method provided in an embodiment of the present invention;

[0036] Figure 2 This is a schematic diagram of the vehicle trajectory prediction and planning error control and collision avoidance device provided in an embodiment of the present invention;

[0037] Figure 3 A schematic diagram of the physical structure of an electronic device provided in an embodiment of the present invention;

[0038] Figure 4 A schematic diagram illustrating the controllability and diversity of the general error model GEM provided in the embodiments of the present invention;

[0039] Figure 5 The time range t for predicting VTP from vehicle trajectory provided in this embodiment of the invention f and final distance offset A schematic diagram of the generated transmission electron microscope image with distance offset d;

[0040] Figure 6 The time range t for predicting VTP from vehicle trajectory provided in this embodiment of the invention f and final heading deviation A schematic diagram of the generated heading offset error d from a transmission electron microscope (TEM) image;

[0041] Figure 7A schematic diagram of function M(k) and function P(k) provided in an embodiment of the present invention;

[0042] Figure 8 A schematic diagram illustrating the relationship between the spatial distance threshold dthd and the recall and precision of the Cooperative Collision Avoidance System (CCAS) provided in this embodiment of the invention;

[0043] Figure 9 A schematic diagram illustrating how the density-based collision warning strategy (DCWS) improves the accuracy of the cooperative collision avoidance system (CCAS) according to an embodiment of the present invention.

[0044] Figure 10 This is a schematic diagram illustrating the corresponding decrease in the CCAS recall capability of the cooperative collision avoidance system due to the improved accuracy provided in this embodiment of the invention. Detailed Implementation

[0045] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention. In addition, the technical features of the various embodiments or individual embodiments provided by the present invention can be arbitrarily combined to form feasible technical solutions. Such combinations are not constrained by the order of steps and / or structural composition patterns, but must be based on the ability of those skilled in the art to implement them. When the combination of technical solutions is contradictory or cannot be implemented, it should be considered that such combination of technical solutions does not exist and is not within the scope of protection claimed by the present invention. If there are step numbers in the following embodiments, they are only set for ease of explanation and the order between steps is not limited. The execution order of each step in the embodiments can be adaptively adjusted according to the understanding of those skilled in the art.

[0046] In traditional autonomous driving systems, trajectory prediction and planning errors often exhibit complex spatiotemporal coupling characteristics and are significantly affected by environmental disturbances and model limitations, making reliable collision risk assessment difficult in high-density traffic scenarios. To address these industry pain points, this invention first establishes a multi-directional deformation mapping mechanism to address the spatial deviation between the ground truth trajectory and the system's prediction results. By applying independent offsets in key directions such as front left, rear left, front right, and rear right, it realistically simulates the generation and superposition effects of directional errors in the real world. This mechanism not only precisely reflects the horizontal and vertical component decomposition of the offset but also provides a basic coordinate framework for subsequent error controllability.

[0047] After spatial offset modeling is completed, this invention introduces a nonlinear error growth scheme based on time evolution, mapping distance offset and heading offset to explosive growth and sublinear growth functions, respectively, to match the uneven rate of accuracy degradation in autonomous driving algorithms during long-term prediction. This method accurately characterizes the gradual spread of error with trajectory steps through adjustable parameters, making the evolution of the deviation between the simulated trajectory and the true value more closely resemble the dynamic characteristics of real perception and control closed loops.

[0048] To address the statistical distribution variations of the error itself, this invention further employs a multidimensional probability distribution model to define the final offset amplitude through sampling. Specifically, the distance offset is assumed to follow a normal distribution, while the heading offset is described by a folded normal distribution with bidirectional truncation characteristics. This probabilistic processing not only takes into account the uncertainty of the error amplitude but also achieves personalized settings for the error intensity under different system operating conditions through adaptive adjustment of the mean and variance.

[0049] After integrating the error characteristics across three dimensions—space, time, and probability—the proposed error hybrid model organically merges these three sub-models, generating a set of trajectories through multiple random sampling and temporal mapping. Each trajectory not only retains customized error characteristics but also forms a statistically representative sample set, ensuring comprehensive coverage of potential deviation scenarios in the simulation environment and providing a comprehensive and balanced input for subsequent collision warning strategy testing.

[0050] To examine the impact of different error levels on the collision detection system, this invention injects the generated trajectory into an open-source traffic micro-simulator and constructs a real-time injection framework. Within this framework, each error sample corresponds to a simulation experiment, with autonomous and human-driven vehicles operating under real traffic rules. The collision warning strategy is then tested and key performance indicators, such as false alarm rate, false negative rate, and response latency, are quantified under continuously iterative error scenarios. This approach overcomes the limitations of previous single-scenario testing, achieving a large-scale, efficient, and repeatable safety assessment process.

[0051] This invention employs a collaborative modeling approach, integrating spatial deformation, temporal evolution, and probability distribution, and injects the generated customized trajectory into a microscopic simulation environment in real time, thus perfecting the closed-loop solution from error injection to safety assessment. This technology not only provides the industry with a controllable and quantifiable error testing tool but also offers practical evidence for algorithm optimization and system upgrades, significantly improving the safety, reliability, and verifiability of autonomous driving systems in complex traffic scenarios.

[0052] This invention provides a method for vehicle trajectory prediction, planning error control, and collision avoidance. See [link to relevant documentation]. Figure 1The method includes: acquiring true trajectory data of the vehicle, the true trajectory data including position coordinates and heading angle at each moment; determining the spatial offset of the true trajectory data according to the Spatial Deformation Model (SDM), the spatial offset including distance offset and heading offset, specifically including: forward left offset, rear left offset, forward right offset, and rear right offset; determining the changes of distance offset and heading offset over time according to the Temporal Evolution Model (TEM); determining the probability distribution of the final distance offset and the second probability distribution of the final heading offset according to the Probability Distribution Model (PDM); fusing the spatial offset model, temporal evolution model, and probability distribution model through the Error Hybridization Model (EMM) to generate a trajectory that conforms to customized error characteristics; integrating the trajectory into the open-source traffic simulator SUMO to construct a real-time error injection framework and generate a customized trajectory with fully controllable error; and using different collision warning strategies to verify and quantify the key performance indicators of the collision detection system under different prediction / planning accuracies.

[0053] GEM achieves customized error characteristics through the synergy of a Spatial Deformation Model (SDM), a Temporal Evolution Model (TEM), a Probability Distribution Model (PDM), and an Error Hybrid Model (EMM). Specifically, SDM simulates the spatial offset of the trajectory (e.g., forward left, backward right), TEM describes the evolution of error over prediction time (distance offset grows exponentially, heading offset grows logarithmically), PDM characterizes the probabilistic properties of distance and heading offsets using folded normal and normal distributions respectively, and EMM integrates the above models to generate the trajectory.

[0054] In the real-time error injection framework, the SUMO simulator is used to construct hybrid traffic scenarios. Realistic trajectories are obtained by recording floating car data (FCD), and the scenario is then reproduced with customized errors injected. Experiments show that this framework can generate over 580,000 collision scenarios, validating the method's effectiveness. Furthermore, the controllability of the GEM allows for precise adjustment of parameters such as the mean and variance of the trajectory error, providing a tool for studying the relationship between VTP accuracy and the performance of cooperative collision avoidance systems (CCAS).

[0055] Based on the above method embodiments, as an optional embodiment, the vehicle trajectory prediction and planning error control and collision avoidance method provided in this embodiment of the invention, wherein determining the spatial offset of the true trajectory data according to the Spatial Deformation Model (SDM) includes:

[0056] Front left offset:

[0057] Left offset:

[0058] Front right offset:

[0059] Back to right offset:

[0060] in, For t i Left offset x-coordinate before time step; For t i The true value of the x-axis at time; For t i Time position offset vector; For t i Heading deviation at any moment; For t i True value of the course at any given moment; For t i Left offset of the ordinate before the time step; For t i True value of the ordinate at time; For t i Left offset x-coordinate after time step; For t i The ordinate shifts to the left after time step; For t i The x-coordinate shifted to the right before the given time; For t i The rightward offset of the ordinate before the time step; For t i The x-coordinate shifts to the right after time step; For t i The ordinate is shifted to the right after time step; sin represents the sine function; cos represents the cosine function.

[0061] Forward left offset: By decomposing the distance offset into x and y components, and combining the true value of the heading angle and the heading offset, the coordinates after the offset are calculated using sine and cosine functions, reflecting the trajectory deformation of the vehicle's front end offset to the left.

[0062] Rear left offset: By introducing the vehicle length parameter (dh) and decomposing the distance offset in reverse, the rear end of the vehicle is simulated to the left, which reflects the symmetry between the rear and front offsets.

[0063] Front right and back right offsets: Similar to left offsets, but with the y-components having opposite signs, reflecting the trajectory characteristics of rightward offsets.

[0064] The core of SDM (Self-Depth Measurement) is to decompose the error into components along the driving direction (u-axis) and the lateral direction (v-axis) by transforming the local vehicle coordinate system (uv-axis) and the global coordinate system (xy-axis), ensuring that the offset calculation conforms to the vehicle's kinematic characteristics. In the experiment, by adjusting... And distance offset, can generate diverse spatial deformation trajectories, such as Figure 4 As shown, the model's adaptability to different offset modes is verified.

[0065] Based on the above method embodiments, as an optional embodiment, the vehicle trajectory prediction and planning error control and collision avoidance method provided in this embodiment of the invention, wherein determining the changes in distance offset and heading offset over time according to the time evolution model TEM includes:

[0066]

[0067] Where E is the distance offset error explosion parameter; For the final distance offset; t f The time range for vehicle trajectory prediction (VTP) is defined; log represents the sign of the logarithmic function; G is the sublinear growth parameter for heading deviation error. This is the final heading offset.

[0068] Distance offset: Modeled using an exponential function, where the basis parameter E is determined by the final distance offset and the prediction time. This model reflects the characteristic of distance error accumulating and amplifying over time, such as... Figure 5 As shown, different and t f The distance offset curve under the combination shows a monotonically increasing trend.

[0069] Heading deviation: Sublinear growth described by a logarithmic function The basic parameter G is determined by the final heading offset. Confirmed. This model demonstrates the suppressive effect of road constraints on heading error, preventing the error from being amplified without limit, such as... Figure 6 As shown, the rate of increase in heading deviation slows down over time.

[0070] The TEM design is based on the actual VTP error characteristics: distance error is significantly affected by the cumulative effect, while heading error is constrained by road boundaries and grows more gradually. This model provides a dynamic error benchmark for subsequent time series trajectory generation.

[0071] Based on the above method embodiments, as an optional embodiment, the vehicle trajectory prediction and planning error control and collision avoidance method provided in this embodiment of the invention, wherein determining the probability distribution of the final distance offset and the probability distribution of the final heading offset according to the probability distribution model PDM includes:

[0072]

[0073] in, This represents the probability distribution of the final course deviation; π is the mathematical constant pi. σt represents the variance of the final heading deviation error; exp is an exponential function with base to the natural constant; μ α This represents the average of the final heading deviation error; The variance of the distance offset error; μ XThe mean of the final distance offset that conforms to a normal distribution; This represents the probability distribution of the final distance offset.

[0074] Final distance offset (probability distribution): A folded normal distribution (FND) is used, whose probability density function is determined by the mean and variance. FND is suitable for the non-negative property of distance error, determined by the parameter k(σ). X =kμ X The shape of the distribution can be controlled, where the ratio of the mean of the folded normal distribution to the mean of the final distance offset corresponding to the normal distribution is, for example... Figure 7 As shown in the upper part, this ensures that the statistical characteristics of the distance offset match the actual observations.

[0075] Final heading deviation (probability distribution): A normal distribution is adopted, and the probability density function reflects the symmetry of the heading deviation, which can be positive or negative. The mean and variance can be calibrated through experimental data.

[0076] The role of PDM is to provide a probabilistic basis for error injection. By sampling and generating error values ​​that conform to a customized distribution, the error characteristics of the trajectory are statistically representative. Experiments have verified that when k = 0.7 to 2.0, FND can accurately fit the distance error distribution of the actual VTP.

[0077] Based on the above method embodiments, as an optional embodiment, the vehicle trajectory prediction and planning error control and collision avoidance method provided in this embodiment of the invention, wherein the spatial offset model, temporal evolution model, and probability distribution model are fused through an error hybrid model (EMM) to generate a trajectory that conforms to customized error characteristics, includes: based on the expected final distance offset mean μ D Given parameter k, calculate the mean μ of the final distance offset that corresponds to a normal distribution. X =μ D / M(k) and standard deviation σ X =kμ X Where M(k) is the ratio of the mean of the folded normal distribution to the mean of the corresponding normal distribution; from the first normal distribution Mid-sampling final distance offset From the second normal distribution Mid-sample final heading offset Using the aforementioned time evolution model TEM, the time range [0, t] is calculated. f The distance offset sequence and heading offset sequence within the range; based on the distance offset sequence and heading offset sequence, the spatial deformation model (SDM) is used to generate multiple trajectories that conform to the customized error characteristics and form a trajectory.

[0078] Parameter calculation: Based on the expected final mean distance offset and parameter k, through μ X =μ D / M(k) and σ X =kμ X Determine the parameters of the normal distribution, where M(k) is determined by... Figure 7 The function curve for the upper part is obtained by looking up a table.

[0079] Error sampling: The final distance offset is obtained by sampling from the normal distribution and taking the absolute value. The final heading offset is obtained from the sampling.

[0080] Time series generation: Calculating [0, t] using TEM f Distance offset sequence d[0,t] within ] f ] and heading offset sequence α[0,t f ].

[0081] Trajectory generation: Combining the four spatial deformation modes of SDM, the error sequence is injected into the real trajectory to generate multiple trajectories with customized error characteristics.

[0082] The core of EMM is to achieve controllable error characteristics through a combination of probabilistic sampling and spatiotemporal models. In the experiment, when N=1000, the mean error between the generated trajectory FDE and the target value is less than 5% (e.g., Figure 4 This verified the accuracy of the fusion process.

[0083] Based on the above method embodiments, as an optional embodiment, the vehicle trajectory prediction and planning error control and collision avoidance method provided in this embodiment integrates the trajectory into the open-source traffic simulator SUMO, constructs a real-time error injection framework, and generates a customized trajectory with fully controllable error. This includes: generating a mixed traffic scenario, recording all floating car data (FCD) to obtain the real trajectory; reproducing the mixed traffic scenario, and loading the trajectory into the real-time error injection framework containing the open-source traffic simulator SUMO based on the real trajectory; wherein, the speed of the autonomous vehicle (CAV) is randomly sampled from a uniform distribution U (10 km / h, 30 km / h), and the human-driven vehicle (HDV) is controlled by the extended intelligent driver model (EIDM).

[0084] Hybrid Traffic Scenario Generation: Traffic flow of autonomous vehicles (CAVs) and high-density vehicles (HDVs) is controlled using a fixed random seed. CAV speeds are extracted from a uniform distribution U (10km / h, 30km / h), while HDVs are controlled by an Extended Intelligent Driver Model (EIDM) to simulate human driving behavior. Realistic Trajectory Acquisition: In the first stage, floating car data (FCD) files are recorded, capturing the position, speed, and other data of all vehicles as ground truth (GT) trajectories. Error Injection and Reproduction: In the second stage, the FCD files are loaded to reproduce the scenario. Error trajectories are generated based on a general error model (GEM), and real-world trajectories are injected through the interface of the open-source traffic simulation platform (SUMO) to achieve real-time error simulation. The key to the framework is ensuring scenario repeatability: by fixing the seed and controlling variables, collision detection results under different error configurations are comparable. In experiments, the framework supports traffic flow simulation of 1255 HDVs per hour, with a single experiment lasting 10 hours, generating 1048 collision data points, meeting the needs of large-scale validation.

[0085] Based on the above method embodiments, as an optional embodiment, the vehicle trajectory prediction and planning error control and collision avoidance method provided in this embodiment of the invention, which employs different collision warning strategies and verifies and quantifies the key performance indicators of the collision detection system under different prediction / planning accuracies, includes:

[0086] General Collision Warning Strategy (GCWS):

[0087]

[0088] Density-based collision warning strategy (DCWS):

[0089] w o +w r =t f

[0090] in, For t i The x-coordinate of vehicle A at that time; For t i The ordinate of vehicle A at that time; For t i The x-coordinate of vehicle B at that time; For t i The ordinate of vehicle B at time d; thd This is the spatial distance threshold; ∈ is the existence symbol; ∈ is the membership symbol; TRA A For the trajectory of vehicle A; TRA B For vehicle B's trajectory; N w The number of warnings observed in the window; Δt is the observation interval; w oFor the observation time window; ρ thd The warning density threshold; w r Reserve a time window; || ||2 is the 2-norm symbol.

[0091] The General Collision Warning Strategy (GCWS) is based on a spatial distance threshold (dthd). A warning is issued when the spatial distance between the predicted trajectories of two vehicles is less than dthd. Experiments show that the optimal dthd range is 2.25 meters to 2.75 meters (e.g., ...). Figure 8 At this point, both recall and precision exceed 80%, but a high false alarm rate remains. Density-based collision warning strategy (DCWS): Introduces an observation time window w o and warning density threshold ρ thd When the number of warnings N in the observation window per unit time w with w o The ratio exceeds ρ thd Warnings are triggered on time (w) o +w r =t f ).like Figure 9 and Figure 10 As shown, DCWS can improve accuracy by 10% to 40% at various VTP accuracy levels, when ρ thd At a value of 0.6, recall (greater than 90%) and precision (greater than 70%) can be balanced.

[0092] A comparison of the two strategies shows that the Global Collision Warning System (GCWS) is suitable for scenarios with extremely high safety requirements, while the Density-Based Collision Warning System (DCWS), by delaying judgment to reduce false alarms, is more suitable for scenarios with high comfort requirements. Experiments have verified that DCWS can improve the accuracy from 40% to over 80% in cases of high final displacement error (FDE) (3.5 to 4.0 meters).

[0093] The vehicle trajectory prediction and planning error control and collision avoidance method provided in this invention adopts a general error model (GEM) and an error injection framework to generate customized error trajectories, providing a tool for studying the relationship between vehicle trajectory prediction (VTP) accuracy and cooperative collision avoidance system (CCAS) performance. Experiments clearly show that VTP accuracy is negatively correlated with CCAS recall and positively correlated with precision. The proposed density-based warning strategy (DCWS) improves accuracy at various VTP accuracy levels. Furthermore, the quantitative analysis of the relationship between VTP accuracy and key CCAS indicators helps evaluate the impact of the VTP model on CCAS, further ensuring traffic safety.

[0094] The implementation of the various embodiments of this invention is based on programmed processing (i.e., software) using a device with processor functionality. Therefore, in practical engineering, the technical solutions and functions of the various embodiments of this invention can be encapsulated into various modules. Based on this reality, and building upon the above embodiments, this invention provides a vehicle trajectory prediction and planning error control and collision avoidance device, which is used to execute the vehicle trajectory prediction and planning error control and collision avoidance method in the above method embodiments. See also... Figure 2 The device includes: a first main module for acquiring true trajectory data of the vehicle, the true trajectory data including position coordinates and heading angle at each moment; a second main module for determining the spatial offset of the true trajectory data according to the Spatial Deformation Model (SDM), the spatial offset including distance offset and heading offset, specifically including: forward left offset, rear left offset, forward right offset, and rear right offset; a third main module for determining the changes of distance offset and heading offset over time according to the Temporal Evolution Model (TEM); a fourth main module for determining the probability distribution of the final distance offset and the probability distribution of the final heading offset according to the Probability Distribution Model (PDM); a fifth main module for fusing the spatial offset model, temporal evolution model, and probability distribution model through the Error Hybridization Model (EMM) to generate a trajectory that conforms to customized error characteristics; a sixth main module for integrating the trajectory into the open-source traffic simulator SUMO to build a real-time error injection framework and generate a customized trajectory with fully controllable error; and a seventh main module for verifying and quantifying the key performance indicators of the collision detection system under different prediction / planning accuracies using different collision warning strategies.

[0095] The vehicle trajectory prediction and planning error control and collision avoidance device provided in this embodiment of the invention adopts... Figure 2 Several modules in the model employ the general error model GEM and an error injection framework to generate customized error trajectories, providing a tool for studying the relationship between vehicle trajectory prediction (VTP) accuracy and cooperative collision avoidance system (CCAS) performance. Experiments clearly demonstrate that VTP accuracy is negatively correlated with CCAS recall and positively correlated with precision. The proposed density-based warning strategy DCWS improves accuracy at various VTP accuracy levels. Furthermore, the model quantitatively analyzes the relationship between VTP accuracy and key CCAS indicators, helping to evaluate the impact of the VTP model on CCAS and further ensuring traffic safety.

[0096] It should be noted that the apparatus in the device embodiments provided by the present invention can be used not only to implement the methods in the above method embodiments, but also to implement the methods in other method embodiments provided by the present invention. The difference lies only in setting corresponding functional modules (i.e., software). Its principle is basically the same as that of the above device embodiments provided by the present invention. As long as those skilled in the art, based on the above device embodiments and referring to the specific technical solutions in other method embodiments, obtain corresponding technical means and technical solutions composed of these technical means by combining technical features, and under the premise of ensuring the practicality of the technical solution, they can improve the apparatus in the above device embodiments to obtain corresponding device-type embodiments (i.e., software) for implementing the methods in other method-type embodiments. For example:

[0097] Based on the above-described device embodiments, as an optional embodiment, the vehicle trajectory prediction and planning error control and collision avoidance device provided in this embodiment of the invention further includes: a first sub-module, used to implement the determination of the spatial offset of the generated trajectory relative to the true trajectory data based on the spatial deformation model SDM, including:

[0098] Front left offset:

[0099] Left offset:

[0100] Front right offset:

[0101] Back to right offset:

[0102] in, For t i Left offset x-coordinate before time step; For t i The true value of the x-axis at time; For t i Time position offset vector; For t i Heading deviation at any moment; For t i True value of the course at any given moment; For t i Left offset of the ordinate before the time step; For t i True value of the ordinate at time; For t i Left offset x-coordinate after time step; For t i The ordinate shifts to the left after time step; For t i The x-coordinate shifted to the right before the given time; For t iThe rightward offset of the ordinate before the time step; For t i The x-coordinate shifts to the right after time step; For t i The ordinate is shifted to the right after time step; sin represents the sine function; cos represents the cosine function.

[0103] Based on the above-described device embodiments, as an optional embodiment, the vehicle trajectory prediction and planning error control and collision avoidance device provided in this embodiment of the invention further includes: a second submodule, used to implement the determination of the changes in distance offset and heading offset over time according to the time evolution model TEM, including:

[0104]

[0105] Where E is the distance offset error explosion parameter; For the final distance offset; t f The time range for vehicle trajectory prediction (VTP) is defined; log represents the sign of the logarithmic function; G is the sublinear growth parameter for heading deviation error. This is the final heading offset.

[0106] Based on the above-described device embodiments, as an optional embodiment, the vehicle trajectory prediction and planning error control and collision avoidance device provided in this embodiment of the invention further includes: a third submodule, used to implement the determination of the probability distribution of the final distance offset and the probability distribution of the final heading offset based on the probability distribution model PDM, including:

[0107]

[0108] in, This represents the probability distribution of the final course deviation; π is the mathematical constant pi. σt represents the variance of the final heading deviation error; exp is an exponential function with base to the natural constant; μ α This represents the average of the final heading deviation error; The variance of the distance offset error; μ X The mean of the final distance offset that conforms to a normal distribution; This represents the probability distribution of the final distance offset.

[0109] Based on the above-described device embodiments, as an optional embodiment, the vehicle trajectory prediction and planning error control and collision avoidance device provided in this embodiment of the invention further includes: a fourth sub-module, used to implement the generation of a trajectory conforming to customized error characteristics by fusing the spatial offset model, temporal evolution model, and probability distribution model through the error hybridization model EMM, including: based on the expected final distance offset mean μ DGiven parameter k, calculate the mean μ of the final distance offset that corresponds to a normal distribution. X =μ D / M(k) and standard deviation σ X =kμ X Where M(k) is the ratio of the mean of the folded normal distribution to the mean of the corresponding normal distribution; from the first normal distribution Mid-sampling final distance offset From the second normal distribution Mid-sample final heading offset Using the aforementioned time evolution model TEM, the time range [0, t] is calculated. f The distance offset sequence and heading offset sequence within the range; based on the distance offset sequence and heading offset sequence, the spatial deformation model (SDM) is used to generate multiple trajectories that conform to the customized error characteristics and form a trajectory.

[0110] Based on the above-described device embodiments, as an optional embodiment, the vehicle trajectory prediction and planning error control and collision avoidance device provided in this embodiment of the invention further includes: a fifth sub-module, used to integrate the trajectory into the open-source traffic simulator SUMO, construct a real-time error injection framework, and generate a customized trajectory with fully controllable error, including: generating a mixed traffic scenario, recording all floating car data FCD to obtain the real trajectory; reproducing the mixed traffic scenario, and loading the trajectory into the real-time error injection framework containing the open-source traffic simulator SUMO based on the real trajectory; wherein, the speed of the autonomous vehicle CAV is randomly sampled from a uniform distribution U (10 km / h, 30 km / h), and the human-driven vehicle HDV is controlled by the extended intelligent driver model EIDM.

[0111] Based on the above-described device embodiments, as an optional embodiment, the vehicle trajectory prediction and planning error control and collision avoidance device provided in this embodiment of the invention further includes: a sixth sub-module, used to implement the key performance indicators of the collision detection system under different collision warning strategies and different VTP accuracies, including:

[0112] General Collision Warning Strategy (GCWS):

[0113]

[0114] Density-based collision warning strategy (DCWS):

[0115] w o +w r =t f

[0116] in, For t iThe x-coordinate of vehicle A at that time; For t i The ordinate of vehicle A at that time; For t i The x-coordinate of vehicle B at that time; For t i The ordinate of vehicle B at time d; thd This is the spatial distance threshold; ∈ is the existence symbol; ∈ is the membership symbol; TRA A For the trajectory of vehicle A; TRA B For vehicle B's trajectory; N w The number of warnings observed in the window; Δt is the observation interval; w o For the observation time window; ρ thd The warning density threshold; w r Reserve a time window; || ||2 is the 2-norm symbol.

[0117] The method in this embodiment of the invention is implemented using an electronic device; therefore, it is necessary to introduce the relevant electronic device. For this purpose, this embodiment of the invention provides an electronic device, such as... Figure 3 As shown, the electronic device includes at least one processor, a communications interface, at least one memory, and a communications bus, wherein the at least one processor, the communications interface, and the at least one memory communicate with each other via the communications bus. The at least one processor can invoke logical instructions stored in the at least one memory to execute all or part of the steps of the methods provided in the foregoing method embodiments.

[0118] Furthermore, when the logical instructions in at least one of the aforementioned memories can be implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to related technologies, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various method embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0119] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0120] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the parts that contribute to the related technology, can be embodied in the form of software products. These computer software products can be stored in computer-readable storage media, such as ROM / RAM, magnetic disks, optical disks, etc., and include several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0121] Example 1

[0122] In a scenario of an urban ring elevated road, the system first uses a LiDAR and camera fusion perception module to collect real-time ground truth trajectory data of vehicles, including recording the vehicle's two-dimensional position coordinates and heading angle every 100ms. Then, a spatial deformation model is applied to apply distance and heading offsets of 0.5m and 1° in the front left, rear left, front right, and rear right directions, respectively. A time evolution model is used to increase the distance offset exponentially to 2m and the heading offset logarithmically to 3°, and the final offsets are generated by sampling based on standard normal distribution and folded normal distribution.

[0123] Next, the aforementioned offset features were fused using an error fusion model to generate a trajectory containing 50 trajectories, which was then injected into the SUMO simulator to reproduce actual traffic flow. When the autonomous vehicle was running at a speed of 20 km / h, tests were conducted on general collision warning and density-based collision warning strategies. The results showed that the false alarm rate was 3.2% and the false alarm rate was 1.8% under the general strategy; while the false alarm rate was 2.7% and the false alarm rate was 2.1% under the density-based strategy, validating the effectiveness of controllable error injection and multi-strategy evaluation.

[0124] Example 2

[0125] In a suburban two-lane national highway scenario, after collecting the true vehicle trajectory, a spatial deformation model was applied to apply distance offsets of 1m and 2m and heading offsets of 2° and 5° for different vehicle speeds (10km / h and 40km / h). The time evolution model evolved the two sets of offset values ​​over 5 seconds, with the distance offset increasing exponentially to the target value using E = 1.2, and the heading offset increasing logarithmically to the target value using G = 0.8. The probability distribution model employed dynamically adjusted variance to accommodate the local amplification effect of road curves on error intensity.

[0126] A hybrid traffic simulation involving ten human-driven vehicles and three autonomous vehicles was constructed in SUMO, with 100 simulation experiments executed after real-time trajectory injection. Based on the simulation results, the warning performance at different speeds was statistically analyzed: in the 10km / h scenario, the false alarm rate of the general warning strategy was <1%, and the false alarm rate of the density warning strategy was <1.5%; in the 40km / h scenario, the false alarm rate of the general strategy was <2%, and the false alarm rate of the density strategy was <2.3%. This example demonstrates that the model parameters can be flexibly adjusted according to vehicle speed and road type.

[0127] Example 3

[0128] In complex urban intersection scenarios, for four-way intersections controlled by traffic lights, a large number of ground-value trajectories are collected, and an offset distribution is fitted based on historical operational data. The spatial deformation model applies distance offsets of 0.8m and 1.2m in the front right and rear left directions, respectively, and a heading offset of ±3° in both directions. The temporal evolution model schedules the offset peak to occur at the moment of traffic light switching, better reflecting the characteristics of a sharp increase in error during braking and acceleration. The probability distribution model captures the degree of error dispersion under different lane-changing scenarios by estimating the variance in real time.

[0129] After trajectory injection into the SUMO and CARLA joint simulation platform, more than ten real signal timing schemes were introduced, and collision warning assessment was initiated. In 500 simulation cycles, the response delay distribution of each strategy was output: the average response delay for the general warning strategy was 85ms, and the average response delay for the density-based strategy was 110ms. This embodiment verifies the applicability of the model in signal control scenarios and provides data support for the real-time deployment of warning strategies.

[0130] Example 4

[0131] In a scenario combining long straight sections and gentle curves on a highway, the true trajectory at 120 km / h was acquired, and distance offset was sampled in segments ranging from 0.5m to 3m, and heading offset in segments ranging from 1° to 8°. The time evolution model utilizes multi-stage parameter settings to maintain a low-speed error increase on straight sections, rapidly increasing to a peak on gentle curves before smoothly decreasing; the probability distribution model dynamically adjusts the variance based on the curve radius to ensure that the error is densely distributed in high-risk road sections.

[0132] The generated trajectory was injected into SUMO, and virtual floating cars were set up in each lane for multi-lane parallel testing. False alarm and false negative data were compared between the general and density-based warning strategies: on straight sections, the general strategy had a false alarm rate of 2.5% and a false negative rate of 1.3%; on gentle curves, the general strategy had a false alarm rate of 3.8% and a false negative rate of 2.6%; the density-based strategy improved the false alarm rate by approximately 0.4 percentage points. This embodiment demonstrates the feasibility of error injection and evaluation in high-speed guidance scenarios.

[0133] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of the present invention. Based on this understanding, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions marked in the blocks may occur in a different order than those shown in the drawings. For example, two consecutive blocks may actually be executed substantially in parallel, or sometimes in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or action, or using a combination of dedicated hardware and computer instructions.

[0134] It should be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element. Any expressions such as "predetermined threshold," "preset threshold," etc., without specifying a particular value, can be determined by those skilled in the art through simple experimentation or appropriate adjustments.

[0135] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for vehicle trajectory prediction and planning error control and collision avoidance, characterized in that, include: a. Obtain true vehicle trajectory data and use the position coordinates and heading angle at each moment as a reference; b. Use the spatial deformation model to determine the distance offset and heading offset of the generated trajectory relative to the true trajectory, and apply the offsets in the four directions of front left, back left, front right, and back right respectively; c. Use a time evolution model to determine the variation patterns of the distance offset and heading offset over time; d. Use probability distribution models to determine the probability distribution of the final distance offset and the probability distribution of the final heading offset; e. By fusing the spatial deformation model, temporal evolution model, and probability distribution model through an error hybridization model, a trajectory conforming to customized error characteristics is generated; f. Integrate the trajectory into a traffic simulator and generate a custom trajectory with controllable error within a real-time error injection framework; g. Evaluate collision detection performance under different prediction or planning accuracies using collision warning strategies.

2. The method according to claim 1, characterized in that, In step b, the amount of distance offset applied in the four directions is determined by the offset vector and the heading offset, and the offset vector is decomposed along the vehicle's lateral and longitudinal coordinates.

3. The method according to claim 1, characterized in that, In step c, the distance offset error grows exponentially with respect to the explosion parameter, while the heading offset error grows logarithmically with respect to the sublinear parameter, and the growth process of both depends on the prediction or planning duration.

4. The method according to claim 1, characterized in that, In step d, the final distance offset follows a normal distribution, and the final heading offset follows a folded normal distribution. The mean and variance are set by the expected error level.

5. The method according to claim 1, characterized in that, In step e, the error hybrid model obtains a complete distance offset sequence and a heading offset sequence by sampling the final distance offset and the final heading offset, and uses this to generate multiple trajectories that conform to the error characteristics to form the trajectory.

6. The method according to claim 1, characterized in that, In step f, when the traffic simulator reproduces the mixed traffic scenario, it samples the speed of autonomous vehicles in a uniform distribution of ten to thirty kilometers per hour, and controls human-driven vehicles according to an extended intelligent driver model.

7. A trajectory-based collision warning assessment method, characterized in that: a. Within the set observation window, calculate the minimum Euclidean distance between the trajectories of the two vehicles and compare it with the spatial distance threshold to trigger a general collision warning; b. Count the number of warnings within the same observation window and compare them with the warning density threshold to trigger density-based collision warnings; c. Combine the time reservation window to determine the collision risk level, and output the false alarm rate, false negative rate and response delay of the collision detection system.

8. A vehicle trajectory prediction and planning error control and collision avoidance device, characterized in that, include: The first module is used to acquire true vehicle trajectory data; The second module is used to calculate the distance offset and heading offset based on the spatial deformation model; The third module is used to generate the change of offset over time based on the time evolution model; The fourth module is used to obtain the probability distribution of the final offset based on the probability distribution model; The fifth module is used to generate trajectories through an error mixture model; The sixth module is used to inject the trajectory into the traffic simulator and generate a customized trajectory; The seventh module is used to perform collision warning assessments and output performance metrics.

9. The apparatus according to claim 8, characterized in that, The first through seventh modules are implemented by the processor through computer instructions in memory and interact with the external traffic simulation platform through a network interface.

10. A computer-readable storage medium having a computer program stored thereon, the computer program, when executed by a processor, causing the processor to perform the method according to any one of claims 1 to 7.

Images